Control method and device for energy storage thermal management system

Abstract

This application provides a control method and device for an energy storage thermal management system. The device includes an energy storage thermal management system, a computer storage medium, and a computer program product. The control method of the energy storage thermal management system includes: acquiring historical operating data and current input parameters of the thermal management unit; obtaining a cooling capacity-energy consumption prediction model based on the historical operating data; predicting the required cooling capacity of the battery cells based on the battery cell temperature prediction model; and obtaining a target control strategy for the adjustable input parameters of the thermal management unit based on the cooling capacity-energy consumption prediction model, the required cooling capacity, and the current input parameters, with the goal of minimizing energy consumption, to control the operation of the thermal management unit. The historical operating data includes historical input parameters and corresponding cooling capacity and energy consumption. This application can reduce the energy consumption of the energy storage thermal management system.

Description

Technical Field

[0001] This application relates to the field of energy storage technology, and in particular to a control method and equipment for an energy storage thermal management system, the equipment including an energy storage thermal management system, a computer storage medium and a computer program product. Background Technology

[0002] The cells of an energy storage system can store and utilize electrical energy through cyclic charging and discharging. As the application scenarios of energy storage systems become increasingly complex, existing energy storage systems are usually equipped with energy storage thermal management systems. These existing energy storage thermal management systems can only perform thermal management based on preset fixed modes, resulting in high energy consumption. Summary of the Invention

[0003] This application provides a control method and device for an energy storage thermal management system. The device includes an energy storage thermal management system, a computer storage medium, and a computer program product, which can reduce the energy consumption of the energy storage thermal management system.

[0004] To address the aforementioned technical problems, this application provides a control method for an energy storage thermal management system. This method includes: acquiring historical operating data and current input parameters of the thermal management unit; obtaining a cooling capacity-energy consumption prediction model based on the historical operating data; predicting the required cooling capacity of the battery cells based on a cell temperature prediction model; and obtaining a target control strategy for the thermal management unit, with the goal of minimizing energy consumption, based on the cooling capacity-energy consumption prediction model, the required cooling capacity, and the current input parameters, to control the operation of the thermal management unit. The historical operating data includes historical input parameters and corresponding cooling capacity and energy consumption.

[0005] To address the aforementioned technical problems, this application provides an energy storage thermal management system, including a memory and a processor. The memory stores program data, which can be executed by the processor to implement the control method described above.

[0006] To address the aforementioned technical problems, this application further provides a computer storage medium. The computer storage medium stores program instructions, which are executed by a processor to implement the control method described above.

[0007] To address the aforementioned technical problems, this application further provides a computer program product. The computer program product includes computer program instructions that enable a computer to implement the aforementioned control method.

[0008] The beneficial effects of this application are as follows: The control method of the energy storage thermal management system of this application includes: acquiring historical operating data and current input parameters of the thermal management unit; obtaining a cooling capacity-energy consumption prediction model based on the historical operating data; predicting the required cooling capacity of the battery cells based on the battery cell temperature prediction model; and obtaining a target control strategy for the adjustable input parameters of the thermal management unit based on the cooling capacity-energy consumption prediction model, the required cooling capacity, and the current input parameters, with the minimum energy consumption as the optimization objective, to control the operation of the thermal management unit; wherein, the historical operating data includes historical input parameters and corresponding cooling capacity and energy consumption. The above method can obtain a cooling capacity-energy consumption prediction model by acquiring historical operating data of the thermal management unit, and then obtain a target control strategy, so that the target control strategy can better fit the actual operating conditions, enabling the energy storage thermal management system to better adapt to the actual operating conditions during thermal management and improve the thermal management effect; and predicting the required cooling capacity of the battery cells based on the battery cell prediction model and determining the target control strategy based on the required cooling capacity can improve the response speed of the energy storage thermal management system, thereby improving the thermal management effect; and obtaining a target control strategy with the minimum energy consumption as the optimization objective can effectively reduce the energy consumption of the energy storage thermal management system. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart illustrating an embodiment of the charge / discharge prediction method of this application; Figure 2 yes Figure 1 Step S13 in the embodiment is a flowchart of an embodiment; Figure 3 yes Figure 2 Step S23 in the embodiment is a flowchart of an embodiment; Figure 4 yes Figure 3 Step S31 in the embodiment is a flowchart of an embodiment; Figure 5 yes Figure 4 Step S42 in the embodiment is a flowchart of an embodiment; Figure 6 yes Figure 3 Step S32 in the embodiment is a flowchart of an embodiment; Figure 7 yes Figure 6 Step S62 in the embodiment is a flowchart of an embodiment; Figure 8 yes Figure 7Step S73 in the embodiment is a flowchart of an embodiment; Figure 9 This is a schematic flowchart of another embodiment of the charge / discharge prediction method of this application; Figure 10 This is a schematic flowchart of an embodiment of the cell temperature prediction method of this application; Figure 11 yes Figure 10 A flowchart illustrating step S102 in the embodiment; Figure 12 yes Figure 10 A flowchart illustrating another embodiment of step S102 in the present invention; Figure 13 This is a flowchart illustrating an embodiment of the control method for the energy storage thermal management system of this application; Figure 14 yes Figure 13 Step A12 in the embodiment is a flowchart of an embodiment; Figure 15 yes Figure 13 A flowchart illustrating another embodiment of step A12 in the examples; Figure 16 yes Figure 13 A flowchart of another embodiment of step A12 in the examples; Figure 17 yes Figure 13 A flowchart of step A12 in another embodiment; Figure 18 yes Figure 17 Step C24 in the embodiment is a flowchart of an embodiment; Figure 19 yes Figure 13 A flowchart illustrating step A12 in one embodiment; Figure 20 This is a flowchart illustrating another embodiment of the control method for the energy storage thermal management system of this application; Figure 21 This is a flowchart illustrating yet another embodiment of the control method for the energy storage thermal management system of this application; Figure 22 This is a schematic diagram of the charge and discharge power curve of a charge and discharge event in one embodiment of this application; Figure 23 This is a schematic diagram of the structure of an embodiment of the computer storage medium of this application. Detailed Implementation

[0010] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0011] The terms "first," "second," etc., used in this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. It should be understood that, when used in this specification, the term "comprising" indicates the presence of the described feature, integral, step, operation, element, and / or component, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof. It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this specification, unless the context clearly indicates otherwise, the singular forms "a," "an," and "the" are intended to include the plural forms. It should also be further understood that the term "and / or," as used in this specification, refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0012] As used in this specification, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determination" or "if [the described condition or event] is detected" may be interpreted, depending on the context, as "once determination," "in response to determination," "once [the described condition or event] is detected," or "in response to detection of [the described condition or event]."

[0013] It should be noted that when one element is fixed to another element, this includes fixing the element directly to the other element or fixing the element to the other element through at least one other intermediate element. When one element is connected to another element, this includes connecting the element directly to the other element or connecting the element to the other element through at least one other intermediate element.

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0015] Energy storage system cells can store and utilize electrical energy through cyclic charging and discharging. As energy storage system applications become increasingly complex, existing systems typically include various management systems such as thermal management systems and cell management systems. These systems are used to collect status data from components like cells to formulate corresponding control strategies, thereby ensuring the safe operation of the energy storage system. How to further improve the control accuracy of management systems in energy storage systems is one of the key issues of concern to those skilled in the art.

[0016] This application first proposes a charge and discharge prediction method for energy storage thermal management systems, such as... Figure 1 As shown, the charge / discharge prediction method includes steps S11 to S13.

[0017] Step S11: Obtain the current data of the battery cell.

[0018] Current data of a battery cell can include its current state of health (SOH), current state of charge (SOC), current temperature, current operating mode, current charge / discharge event, and current charge / discharge power curve. This data can reflect the working status of the battery cell to a certain extent, so obtaining its current data helps to make more accurate predictions of future charge / discharge power.

[0019] Among them, the charge and discharge power curve refers to the curve of how the charge and discharge power of the battery cell changes over time.

[0020] Step S12: Obtain the historical charge and discharge event library of the battery cell.

[0021] In some embodiments, the historical charge and discharge event database stores records of charge and discharge events of the battery cell at different historical periods. Each historical charge and discharge event includes the charge and discharge power curve at the time of the event and the corresponding state parameters, such as the health status, state of charge, and ambient temperature of the battery cell at that time.

[0022] In some embodiments, the historical charge / discharge event library includes at least 30 days of historical data.

[0023] In some embodiments, a charge / discharge event includes a waiting period, a first charge / discharge period, a rest period, and a second charge / discharge period.

[0024] During the first charge / discharge period, the battery cell is either charging or discharging; during the second charge / discharge period, the battery cell is either charging or discharging. (See also...) Figure 22 , Figure 22The charge / discharge power curves for a single charge / discharge event are shown. Negative charge / discharge power corresponds to the discharge period, while positive charge / discharge power corresponds to the charging period. This marks the beginning of the event. t1 and t2 are the start and end times of the first charging and discharging period, respectively, and t3 and t4 are the start and end times of the second charging and discharging period, respectively. The waiting time is the duration from the start of the event to the start of the first charging / discharging period. The duration of the first charging / discharging period and the corresponding charging / discharging power, i.e., the first charging / discharging period; The duration from the end of the first charging / discharging period to the beginning of the second charging / discharging period is the resting period. This refers to the duration of the second charging / discharging period and the corresponding charging / discharging power.

[0025] In one application scenario, within a historical charge / discharge event, the battery cell is in a charging state during the first charge / discharge period and in a discharging state during the second charge / discharge period. Defining a charge / discharge event as including at least two charge / discharge periods more closely reflects the daily usage scenarios of energy storage systems. For example, in some practical applications, the energy storage system charges at fixed intervals, then remains idle for a period before discharging at a fixed interval, and then remains idle for another period before resuming charging at the fixed charging interval. This multi-period division of charge / discharge events more realistically reflects the actual operation of the battery cell; and when conducting comprehensive analysis of similar events, this division provides more accurate and relevant basic data.

[0026] In some embodiments, the full charge and discharge power curve of the battery cell within a preset time period is obtained, and the full charge and discharge power curve is divided according to a preset rule to determine the charging and discharging period and the non-charging and discharging period. Then, multiple historical charging and discharging events and their corresponding historical scene tags can be obtained based on the preset rule.

[0027] For example, in some embodiments, a charge / discharge power threshold and a minimum charge / discharge duration are set. When the absolute value of the charge / discharge power is not less than a specific charge / discharge power threshold, and the duration of such a situation is not less than the minimum charge / discharge duration, the time period in which such a situation occurs is determined to be a charge / discharge period. As another example, in some embodiments, a charge / discharge power threshold and a minimum static duration are set. When the absolute value of the charge / discharge power is less than a specific charge / discharge power threshold, and the duration of such a situation is not less than the minimum static duration, the time period in which such a situation occurs is determined to be a non-charge / discharge period. As yet another example, in some embodiments, a hysteresis threshold can also be set. Only when the absolute value of the charge / discharge power exceeds the range of the hysteresis threshold is it determined to be an exit from a charge / discharge period, thus achieving a stabilization effect. As yet another example, in some embodiments, when the charge / discharge power of two adjacent charge / discharge periods is both positive or both negative, the two charge / discharge periods are merged into one charge / discharge period when the static interval between them is less than a specific threshold.

[0028] In other embodiments, depending on the use case and prediction strategy, a charge / discharge event may also be set to include 1, 3, 4, or other charge / discharge periods.

[0029] Step S13: Based on current data and the historical charge and discharge event library, predict the predicted charge and discharge power curve of the battery cell.

[0030] The charge / discharge power curve is a curve showing how the charge / discharge power of a battery cell changes over time. The time span for predicting the charge / discharge power curve is not limited.

[0031] The charge and discharge prediction method in steps S11 to S13 can use the current data of the battery cell and the historical charge and discharge event library of the battery cell to predict the predicted charge and discharge power curve of the battery cell. The predicted charge and discharge power curve can characterize the working state of the battery cell in the future. This helps various management systems in the energy storage system to start the corresponding control strategies in advance based on the predicted charge and discharge power curve before the working state of the battery cell changes, thereby improving the response speed and control accuracy of the management system and improving the safety of the energy storage system.

[0032] Taking an energy storage thermal management system as an example, the system can employ appropriate thermal management strategies to control the thermal performance of the battery cells. The heat generated during the charging and discharging process of the battery cells can cause drastic temperature changes. In some embodiments, a predicted charging and discharging power curve is obtained based on step S13. This predicted power curve helps to obtain a predicted current curve for the battery cells. The predicted current curve can then predict the first heat generation power curve of the battery cells in the future. This allows the energy storage thermal management system to initiate corresponding pre-emptive control measures based on the first heat generation power curve before the actual temperature change of the battery cells, improving response speed and enhancing the safety of battery cell use.

[0033] In some embodiments, the historical charge-discharge event database includes scene tags corresponding to historical charge-discharge events. Historical charge-discharge events can be retrieved based on these scene tags to obtain similar historical events to the current charge-discharge event. A predicted charge-discharge power curve is then obtained based on the similar charge-discharge power curves corresponding to these historical similar events. An example is provided below.

[0034] In some embodiments, step S13 can be performed as follows: Figure 2 The method shown is implemented in the following way, specifically including steps S21 to S23.

[0035] Step S21: Based on the current data, obtain the current scene tag and current charge / discharge power curve corresponding to the current charge / discharge event of the battery cell.

[0036] In some embodiments, a charge / discharge event has a corresponding charge / discharge power curve. For example, a current charge / discharge event has a current charge / discharge power curve corresponding to the current charge / discharge event, and a historical charge / discharge event has a historical charge / discharge power curve corresponding to the historical charge / discharge event.

[0037] The current charge / discharge power curve refers to the curve showing the actual charge / discharge power of the battery cell over time, from the start of the current charge / discharge event to the current moment. The historical charge / discharge power curve refers to the curve showing the actual charge / discharge power of the battery cell over time, from the start of a historical charge / discharge event to the end of the previous event.

[0038] In some embodiments, a charge / discharge event has a corresponding scene tag. For example, a current charge / discharge event may have a corresponding current scene tag, and a historical charge / discharge event may have a corresponding historical scene tag. In some embodiments, before step S21 or step S22, the following step may be added: classifying historical charge / discharge events in the historical charge / discharge event library by scene, and setting different scene tags for historical charge / discharge events in different scenes.

[0039] In some embodiments, scenario tags include photovoltaic (PV) power storage, peak-valley arbitrage, backup power plan, emergency plan, dispatch instructions, and frequency regulation events. For example, in the PV power storage scenario, the charging period of the battery cell overlaps with the peak output period of PV, and the outline of the charging power curve is related to the changes in PV power. In the peak-valley arbitrage scenario, the charging period of the battery cell mainly occurs during the valley price period, and the discharging period mainly occurs during the peak price period. In the backup power plan or emergency plan scenario, when a power outage / islanding signal is detected, or when the battery cell's state of charge protection strategy is triggered, the battery cell's discharge power is in a stable power supply state. In the dispatch instructions or frequency regulation events scenario, there are dispatch instructions, or the charging and discharging power exhibits high-frequency up-and-down adjustment characteristics.

[0040] In some embodiments, the current scenario label of the current charging and discharging event can be obtained based on the planning or control strategy in the Energy Management System (EMS) of the energy storage system.

[0041] There are several ways to obtain scene tags for historical charging and discharging events.

[0042] In some embodiments, historical charging and discharging events are structurally grouped based on preset rules and relevant parameters of historical charging and discharging events to obtain a scene label for each historical charging and discharging event. For example, in some embodiments, historical charging and discharging events are structurally grouped based on at least one of the following factors: EMS mode position, grid connection / off-grid status, electricity price period, photovoltaic power output, dispatch instructions, and the shape of the charging and discharging power curve, to determine their scene labels.

[0043] In some embodiments, statistical features of historical charge-discharge events can be used to train a classifier to output corresponding scene labels. For example, in some embodiments, statistical features such as peak charge-discharge power, duration, occurrence time, slope of the charge-discharge power curve, and correlation between the charge-discharge power curve and electricity price, photovoltaic modules, and load are extracted from the historical charge-discharge event database for each historical charge-discharge event. Based on these statistical features, a classifier is trained to output scene labels and confidence scores corresponding to the historical charge-discharge events. In some embodiments, when the confidence score of a scene label is lower than a confidence threshold, the above-mentioned structured grouping method can be used to redetermine the scene label of the corresponding event.

[0044] Step S22: Based on the current scene label, obtain historical similar events and corresponding similar charge and discharge power curves from the historical charge and discharge event library.

[0045] In some embodiments, historical charge-discharge events with the same tag are obtained from the historical charge-discharge event database based on the current scene tag as historical similar events. Each historical charge-discharge event has a corresponding historical charge-discharge power curve, that is, the actual charge-discharge power curve of the cell within the time period corresponding to the historical charge-discharge event; in some embodiments, the historical charge-discharge power curve corresponding to the historical similar event is determined as the similar charge-discharge power curve.

[0046] In other embodiments, if the current scenario label cannot be obtained, a candidate label set can be obtained based on electricity price time schedules, day-ahead electricity price forecasts, photovoltaic forecasts, load forecasts, grid connection constraints, dispatch signals, etc. , These are candidate scene labels. The corresponding probability weights are used; based on the candidate tag set, historical charging and discharging events corresponding to candidate scene tags are retrieved from the historical charging and discharging event database and used as historical similar events. At this time, multiple historical similar events corresponding to multiple candidate scene tags can be obtained. In some embodiments, the probability weights corresponding to each historical similar event can be used to correct the corresponding historical charging and discharging power curves, and the processed power curves are determined as similar charging and discharging power curves corresponding to historical similar events.

[0047] Step S23: Obtain the predicted charge / discharge power curve based on similar charge / discharge power curves and the current charge / discharge power curve.

[0048] In some embodiments, the predicted charge / discharge power curve may be predicted only for the period before the current charge / discharge event occurs; in other embodiments, the predicted charge / discharge power curve may also be predicted from the current moment to any future moment. For example, at the end of the current charge / discharge event, the prediction of the next future charge / discharge event is immediately initiated, and the charge / discharge power curve corresponding to the next charge / discharge event is predicted.

[0049] The above method, when using a historical charge / discharge event database to predict the predicted charge / discharge power curve of a battery cell, can classify historical charge / discharge events and obtain similar historical events based on the event classification. It then uses the corresponding charge / discharge power curves of these similar historical events to obtain the predicted charge / discharge power curve. Therefore, it can more accurately capture the power variation patterns under different types of charge / discharge scenarios, thereby improving prediction accuracy. For example, when the current charge / discharge event is a peak-valley arbitrage scenario, all similar historical events with the same tags are selected from the historical charge / discharge event database. By combining the initial state of the current charge / discharge power curve and current electricity price information with the similar charge / discharge power curves of these similar historical events, the future charge / discharge power variation trend can be predicted more accurately.

[0050] In some embodiments, step S23 can be performed as follows: Figure 3 The method shown is implemented in the following way, specifically including steps S31 to S32.

[0051] Step S31: Based on the current data and the historical charging and discharging event database, obtain the event weights corresponding to similar historical events.

[0052] Event weights can reflect, to some extent, the similarity between a historical similar event and the current charge / discharge event. Event weights can be determined by the similarity between the current data of the battery cell and the relevant parameters corresponding to historical charge / discharge events.

[0053] In some embodiments, step S31 can be performed as follows: Figure 4 The method shown is implemented in the following way, specifically including steps S41 to S42.

[0054] Step S41: Based on the current data and the historical charging and discharging event database, obtain at least one of the following: curve similarity, state similarity, event freshness, and event quality factor corresponding to similar historical events.

[0055] Curve similarity refers to the similarity of curve profiles. In some embodiments, the dynamic time warping algorithm can be used to normalize similar charge-discharge power curves and the current charge-discharge power curve to obtain curve similarity.

[0056] State similarity refers to the similarity of state parameters related to the operating state of a battery cell. State parameters may include the cell's health status, state of charge, ambient temperature, current time characteristics, and current strategy mode. State similarity can be determined by comparing the current parameters with corresponding parameters from similar historical events.

[0057] Event freshness characterizes the temporal distance between similar historical events and the current moment, and can be represented by a freshness decay term. For example, the smaller the temporal distance between the occurrence time of similar historical events and the current moment, the higher the event freshness value.

[0058] The event quality factor characterizes the quality of relevant data for similar historical events in the historical charge and discharge event database, such as data missing, data anomalies, and data interruptions. When data is missing, anomalies, or interruptions occur, the event quality factor value will decrease.

[0059] Step S42: Calculate the event weight corresponding to historical similar events based on at least one of curve similarity, state similarity, event freshness, and event quality factor.

[0060] For example, when determining event weights, curve similarity, state similarity, event freshness, and event quality factor can be considered simultaneously. This approach can comprehensively measure the similarity between historical similar events and current charging / discharging events from multiple dimensions, enabling the calculated event weights to more accurately reflect the similarity between a given historical similar event and the current charging / discharging event.

[0061] In some embodiments, step S42 can be performed as follows: Figure 5 The method shown is implemented in the following way, specifically including steps S51 to S52.

[0062] Step S51: Obtain the first relation.

[0063] The first relation is a relation that reflects the relationship between event weight and curve similarity, state similarity, event freshness, and event quality factor.

[0064] Step S52: Calculate the event weight based on the first relation, curve similarity, state similarity, event freshness, and event quality factor.

[0065] The first relation satisfies:

[0066] The first relation is the formula for calculating the weights. Wherein, The event weight corresponding to the i-th historical similar event; This is the current charge / discharge power curve. The current state parameter; , These are the historical charge / discharge power curves and curve similarity corresponding to the i-th historical similar event, respectively. , Let be the historical state parameters and state similarity corresponding to the i-th historical similar event, respectively. , Let be the event freshness and event quality factor corresponding to the i-th historical similar event, respectively; where, p is the set of all historically similar events. j , Sets The historical charge / discharge power curve and curve similarity corresponding to the j-th historical similar event; x j D x (x cur x j ) represent the historical state parameters and state similarity corresponding to the j-th historical similar event, respectively. , and are the event freshness and event quality factor corresponding to the j-th historical similar event, respectively; where, The first preset parameter, This is the second preset parameter.

[0067] Where, exp( () is an exponential function with the natural constant e as the base, used to perform a nonlinear transformation on distance metrics to generate normalized weight coefficients.

[0068] in, The DTW distance can be used to characterize curve similarity; the current charge / discharge power curve is calculated using the dynamic time warping algorithm. Comparison with the i-th historical charge / discharge power curve The DTW distance can more accurately reflect the curve similarity between the two. The smaller the value, the greater the event weight corresponding to the i-th historical charging and discharging event.

[0069] Similarly, Current charge / discharge power curve Comparison with the j-th historical charge / discharge power curve DTW distance.

[0070] in, for and The distance metric between states reflects their similarity. It can be calculated using distance metrics such as Euclidean distance and Mahalanobis distance. . The smaller the value, the greater the event weight corresponding to the i-th historical charging and discharging event.

[0071] Similarly, for and The distance metric between them can be improved in a similar way.

[0072] in, , This is a sample freshness decay term, which can be used to characterize event similarity. In some embodiments, , These represent the time intervals between the i-th historical charge / discharge event and the current charge / discharge event, and the j-th historical charge / discharge event and the current charge / discharge event, respectively. Time interval The longer the duration, the smaller the event weight corresponding to the i-th historical charging and discharging event.

[0073] In some embodiments, ≥0, ≥0. When the data for the i-th historical similar event is complete, without anomalies, and uninterrupted, The value is 1; when there is data loss, anomaly, or interruption, The value is less than 1. The specific value can be set according to the degree of data loss, abnormality or interruption. For example, the value is 0.8 when there is slight loss and 0.3 when there is severe abnormality. The smaller the value, the smaller the event weight corresponding to the i-th historical charging and discharging event.

[0074] In some embodiments, ≥0, ≥0, ≥0; It can be used as a sensitivity adjustment factor to adjust the sensitivity of DTW distance to event weights. The larger the value, the greater the impact of the curve similarity on the event weight. It can be used as a sensitivity adjustment factor for regulation. Sensitivity to event weights The larger the value, the greater the impact of the state similarity on the event weight. It can be used as a sensitivity adjustment factor for regulation. Sensitivity to event weights The larger the value, the greater the impact of event freshness on event weight. This can be adjusted based on different application scenarios. , , The value of .

[0075] Step S32: Obtain the predicted charge / discharge power curve based on similar charge / discharge power curves and corresponding event weights, and the current charge / discharge power curve.

[0076] In the above method, when using similar charge-discharge power curves to obtain the predicted charge-discharge power curve, setting corresponding event weights for each similar charge-discharge power curve can yield a predicted charge-discharge power curve that is more in line with the actual situation, thereby improving the accuracy of the prediction.

[0077] In some embodiments, step S32 can be performed as follows: Figure 6 The method shown is implemented in the following way, specifically including steps S61 to S62.

[0078] Step S61: Based on the current charge and discharge power curve, normalize the similar charge and discharge power curve using the dynamic time warping algorithm to obtain the corresponding normalized curve.

[0079] Step S62: Obtain the predicted charge / discharge power curve based on the normalized curve and the corresponding event weights.

[0080] In the above method, when obtaining the predicted charge / discharge power curve using similar charge / discharge power curves, the similar charge / discharge power curves are first normalized. Then, the predicted charge / discharge power curve is obtained using the normalized curve and the corresponding event weights. This can, to a certain extent, reduce the prediction deviation caused by differences in time scale, power amplitude, etc., between different historical similar events, and improve the accuracy and reliability of the predicted charge / discharge power curve. For example, in an application scenario, the current charge / discharge event has occurred up to the end of the first charge / discharge period, i.e., the waiting period and the first charge / discharge period have been completed. The duration of the waiting period plus the first charge / discharge period in the similar charge / discharge power curve of a historical similar event is different from the duration of the current charge / discharge event. The similar charge / discharge power curve can be stretched or compressed using a dynamic time warping algorithm to make the duration of the waiting period plus the first charge / discharge period in the similar charge / discharge power curve the same in length as the duration of the current charge / discharge power curve.

[0081] In some embodiments, the predicted charging and discharging power curve can also be obtained based on similar charging and discharging power curves by predicting the duration of the waiting period. For example, in an application scenario, the current charging and discharging power curve is always 0, that is, the current charging and discharging event is still in the waiting period. Step S32 can also be implemented by the following steps: obtaining the event weight of each historical similar event; obtaining the duration of the waiting period of each historical similar event based on the similar charging and discharging power curve; and calculating the weighted average of the event weight and the duration of the waiting period to obtain the predicted duration of the waiting period.

[0082] The duration of the waiting period is the time from the start of a charge / discharge event to the start of the first charge / discharge phase of that event. Therefore, the predicted duration of the waiting period can be used to predict the start time of the first charge / discharge phase in the current charge / discharge event. The predicted durations of other periods can also be obtained using the same method, which will not be elaborated further. After obtaining the durations of all periods in the current charge / discharge event, the charge / discharge power within the corresponding periods can be further obtained based on similar charge / discharge power curves. For example, event weights can be used to perform weighted fusion of all similar charge / discharge power curves, or the similar charge / discharge power curves can be normalized first, and then weighted fusion can be performed using event weights to obtain the charge / discharge power curves within the charge / discharge periods. This allows us to obtain the predicted charge / discharge power curves corresponding to the periods before the current charge / discharge event has occurred.

[0083] In some embodiments, step S62 can be performed as follows: Figure 7 The method shown is implemented in the following way, specifically including steps S71 to S73.

[0084] Step S71: Calculate the weighted average curve based on the normalized curve and the corresponding event weights.

[0085] Step S72: Obtain the current state parameters of the battery cell based on the current data of the battery cell.

[0086] The status parameters can be referred to in the above embodiments and will not be repeated here. The current status parameters are those that can characterize the current working state of the battery cell, such as the current health status of the battery cell, the current state of charge, and the current ambient temperature.

[0087] Step S73: Correct the weighted average curve based on the current state parameters to obtain the predicted charge and discharge power curve.

[0088] In the above method, when obtaining the predicted charge / discharge power curve based on the normalized curve and the corresponding event weights, the weighted average curve is corrected using the current state parameters. This allows for dynamic adjustment of the prediction results based on the current actual operating state of the battery cell, making the predicted charge / discharge power curve more closely match the current performance level and operating conditions of the battery cell. For example, if the current health state of the battery cell is lower than that of the battery cell in similar historical events, its charge / discharge capacity will be correspondingly weakened. By incorporating the current health state parameters into the correction process, the power amplitude of the weighted average curve can be appropriately reduced, minimizing the prediction deviation caused by battery cell aging. Furthermore, if the current ambient temperature is significantly higher than the ambient temperature during similar historical events, considering the impact of high temperature on the battery cell's charge / discharge efficiency, the charge / discharge rate of the weighted average curve can be corrected using the current ambient temperature data, reducing the prediction deviation caused by temperature changes.

[0089] In some embodiments, step S73 can be performed as follows: Figure 8 The method shown is implemented in the following way, specifically including steps S81 to S82.

[0090] Step S81: Obtain the second relation.

[0091] The second relationship is a relationship that reflects the relationship between the predicted charge and discharge power curve and the current state parameters and weighted average curve.

[0092] Step S82: Obtain the predicted charge and discharge power curve based on the second relation, the weighted average curve, and the current state parameters.

[0093] The second relation satisfies:

[0094] in, The predicted charge / discharge power curve of the battery cell. For the current state parameters, This refers to the current health status of the battery cell. This represents the current state of charge of the battery cell. The current ambient temperature of the battery cell; For weighted average curves, , These are the event weights and normalized curves corresponding to the i-th historical similar event, respectively. This is the lower limit of power. This is the upper limit of power. This represents the current temperature of the battery cell.

[0095] Among them, the lower limit of power and power limit These are all the current temperatures of the battery cells. Current health status of the battery cells The function.

[0096] The clip function is a limiting function. It restricts the calculated power value to a range defined by a lower and upper power limit.

[0097] In the above method, after correcting the weighted average curve using current state parameters, a clipping function is used to further adjust the curve, making the predicted charge / discharge power curve closer to the actual situation. For example, when the current health state of the battery cell is low, its maximum charge / discharge power will decrease. In this case, the values ​​of the lower and upper power limits are adjusted accordingly, and the clip function will limit the calculated power value within the range defined by the lower and upper power limits. This dual correction can further improve the prediction accuracy. For another example, in an energy storage thermal management system, the resistance of the battery cell is related to its current health state (SOH) and current temperature (T). Therefore, the current health state (SOH) and current temperature (T) can affect the heat generation power of the battery cell by influencing its resistance. Incorporating parameters such as the current health state (SOH) and current temperature (T) during power prediction can effectively improve the prediction accuracy when predicting heat generation power using predicted charge / discharge power.

[0098] In some embodiments, this application further proposes a charge / discharge prediction method, such as... Figure 9 As shown, it includes steps S91 to S97.

[0099] Step S91: Obtain the current data of the battery cell.

[0100] The specific implementation of step S91 can be referred to step S11, and will not be repeated here.

[0101] Step S92: Obtain the historical charge and discharge event library of the battery cell.

[0102] The specific implementation of step S92 can be referred to step S12, and will not be repeated here.

[0103] Step S93: Based on current data and the historical charge and discharge event library, predict the predicted charge and discharge power curve of the battery cell.

[0104] The specific implementation of step S93 can be referred to step S13, and will not be repeated here.

[0105] Step S94: Obtain the predicted value of the charge and discharge power at the current moment based on the predicted charge and discharge power curve of the previous moment.

[0106] The previous moment is prior to the current moment, and the predicted charge / discharge power curve of the previous moment includes the predicted value of the charge / discharge power at the current moment. In some embodiments, each moment has a current charge / discharge power curve and a predicted charge / discharge power curve that predicts future charge / discharge power. For example, taking the current moment as an example, the current charge / discharge power curve at least includes a curve showing the actual value of the charge / discharge power changing over time from the start of the current charge / discharge event to the current moment, and the predicted charge / discharge power curve at least includes the predicted value of the charge / discharge power at the next moment after the current moment.

[0107] Step S95: Obtain the actual value of the charging and discharging power at the current moment.

[0108] Step S96: Calculate the prediction error based on the predicted value and the actual value.

[0109] Step S97: In response to the prediction error meeting the first preset condition, re-predict the predicted charge and discharge power curve corresponding to the current time based on the actual value.

[0110] For example, in some embodiments, the relative error between the predicted value and the actual value is calculated, and this relative error is used as the prediction error. The relative error can be calculated with reference to Expression 1-1, which satisfies:

[0111] in, This is a relative error. This is the actual power. To predict power, This is the fourth preset parameter.

[0112] In some embodiments, the first preset condition is that the prediction error is greater than an error threshold, or the first preset condition is that the number of times the prediction error is greater than the error threshold is greater than a frequency threshold. In some embodiments, the first preset condition is that the prediction error is greater than the error threshold, and the number of times the prediction error is greater than the error threshold is greater than a frequency threshold.

[0113] In some embodiments, the first preset condition is that the prediction error is greater than the error threshold, and the number of times the prediction error is greater than the error threshold within the sliding window is greater than the number threshold.

[0114] Taking relative error as an example, in one application scenario, the error threshold in the first preset condition is 20%, and the number of times threshold is 3. When the relative error... >20%, and occurring 4 times consecutively If the error rate exceeds 20%, a re-prediction is deemed necessary. In another application scenario, the preset duration of the sliding window is 10 minutes. If, within these 10 minutes, the number of times the relative error exceeds the error threshold exceeds the number of occurrences threshold, a re-prediction mechanism is triggered. In other application scenarios, the error threshold can also be 5%, 10%, 15%, 25%, or 30%, and the number of occurrences threshold can be 1, 2, 4, 5, 6, 7, or 8, which can be set according to actual needs.

[0115] In some embodiments, the error threshold and the number of times threshold can be configured such that the larger the error threshold, the smaller the number of times threshold.

[0116] In some embodiments, when it is determined that a re-prediction is needed, the current charge-discharge power curve is updated based on the actual charge-discharge power value at the current moment, and the process returns to step S93. The updated current charge-discharge power curve is then used to re-filter historical similar events from the historical charge-discharge event database, thereby recalculating the event weights and predicting the charge-discharge power curve, achieving dynamic correction of the prediction results. This closed-loop feedback mechanism can promptly correct prediction inaccuracies caused by initial prediction deviations or sudden changes in external conditions, improving prediction accuracy.

[0117] In some embodiments, when it is determined that re-prediction is needed, the current charge / discharge power curve is updated based on the actual charge / discharge power value at the current moment, and the process returns to step S93. Specifically, with For the current moment, the nearest window (L>0) The actual charge / discharge power curve corresponding to this event is re-segmented according to the judgment rules of a charge / discharge event, determining which period of the event the current moment falls into: the waiting period, the first charge / discharge period, the static period, or the second charge / discharge period. Then, the consumed time and remaining prediction window of the current charge / discharge event are re-determined. Based on the control method in the above embodiment, the event tag of the current charge / discharge event is obtained again, and historical similar events are re-filtered from the historical charge / discharge event database. Based on the re-obtained historical similar events, the predicted charge / discharge power curve of the battery cell at subsequent moments is re-predicted, overwriting the original prediction data and recording the reasons for the re-prediction, such as plan deviation, strategy switching, etc., and the event quality factor in the above embodiment can be updated based on the current actual value.

[0118] This application further proposes a method for predicting the cell temperature in an energy storage thermal management system, such as... Figure 10 As shown, the cell temperature prediction method includes steps S101 to S104.

[0119] Step S101: Obtain the predicted charge / discharge power curve.

[0120] In some embodiments, the predicted charge and discharge current curve of the battery cell at a future time can be predicted based on the predicted charge and discharge power curve of the battery cell, thereby predicting the first heat generation power curve of the battery cell at a future time, and then the temperature of the battery cell at a future time can be predicted by combining the current temperature of the battery cell.

[0121] In some embodiments, a predicted charge / discharge power curve is obtained based on the charge / discharge prediction method of any of the above embodiments. The meaning and principle of the predicted charge / discharge power curve can be referred to the above embodiments, and will not be repeated here. The current data includes at least the current health state, current state of charge, and current temperature of the battery cell.

[0122] Step S102: Based on the current health status, current state of charge, and predicted charge / discharge power curve, predict the first heat generation power curve of the battery cell.

[0123] The current health status and current state of charge of the battery cell also affect the basic heat generation of the battery cell. Therefore, considering the current health status and current state of charge of the battery cell when predicting the first heat generation power curve of the battery cell helps to improve the accuracy of the prediction.

[0124] In some embodiments, step S102 can be performed as follows: Figure 11 The method shown is implemented in the following way, specifically including steps S111 to S113.

[0125] Step S111: Obtain the aging correction coefficient of the battery cell based on the current health status.

[0126] The aging correction factor is used to quantify the impact of cell aging on heat generation power. For example, as the health of a cell declines, its internal resistance typically increases, which leads to increased heat generation at the same current. Therefore, in some embodiments, the lower the current health of the cell, the larger the aging correction factor, and the higher the predicted heat generation power.

[0127] Step S112: Obtain the charge deviation coefficient of the cell based on the current state of charge.

[0128] The current state of charge (SPC) of a battery cell affects its internal resistance characteristics, which in turn affects the cell's heat generation power. The charge deviation coefficient is used to quantify the impact of the cell's current SPC on its heat generation power.

[0129] Step S113: Obtain the second heat generation power curve of the battery cell based on the predicted charge and discharge power curve.

[0130] The predicted charge / discharge power curve of a battery cell can predict its predicted charge / discharge current curve, and thus its second heat generation power curve. An aging correction factor and a charge deviation factor are used to correct this second heat generation power curve to obtain the first heat generation power curve. For example, the predicted charge / discharge current curve can be calculated based on the predicted charge / discharge power curve and the battery cell's voltage parameters.

[0131] Step S114: Obtain the third relational expression, and predict the first heat generation power curve of the battery cell based on the third relational expression, aging correction coefficient, charge deviation coefficient, and second heat generation power curve.

[0132] The third relation satisfies:

[0133] in, The heat generation capacity of the battery cell is represented by SOH, which represents the current health status of the battery cell. This is the aging correction factor for the battery cell, and SOC is the current state of charge of the battery cell. This is the charge deviation coefficient of the battery cell. This is the basic heat generation power of the battery cell.

[0134] In some embodiments, the heat generation power of the battery cell can be calculated with reference to the following expressions 1-2.

[0135]

[0136] Where Pcell_base is the basic heat generation power of the battery cell, I is the charging and discharging current of the battery cell, k1 is the fourth preset parameter, k2 is the fifth preset parameter, and k3 is the sixth preset parameter.

[0137] In some embodiments, k1, k2, and k3 can be calibrated through standard cyclic testing. This setting can take into account the influence of the cell's internal resistance, the thermodynamic hysteresis effect of the cell's material, and the entropy change heat of the cell during charging and discharging on the heat generation power when calculating the cell's heat generation, thereby improving the accuracy of cell temperature prediction.

[0138] Step S103: Obtain the current ambient heat exchange power.

[0139] The current ambient heat exchange power refers to the amount of heat exchanged between the battery cell and the external environment per unit time. For example, the greater the temperature difference between the current ambient temperature and the current temperature of the battery cell, the greater the absolute value of the ambient heat exchange power. When the ambient temperature is higher than the battery cell temperature, the battery cell absorbs heat from the environment, and the ambient heat exchange power is positive. When the ambient temperature is lower than the battery cell temperature, the battery cell releases heat to the environment, and the ambient heat exchange power is negative.

[0140] In some embodiments, the ambient heat exchange power includes leakage heat power and radiant heat power. For example, in one application scenario, the battery cells are installed inside a container, and the leakage heat power is calculated using expressions 1-3. Calculate the radiant heat power using expression 1-4. .

[0141]

[0142] in, U is the heat leakage power of the container, U is the heat transfer coefficient of the container body. In some embodiments, the heat transfer coefficient of the container body is typically 0.5 W / m²·K-1.5 W / m²·K, and A is the effective heat dissipation area of ​​the container body. The temperature of the inner surface of the chamber. This refers to the outer surface temperature of the enclosure.

[0143]

[0144] in, Let σ be the radiative thermal power, and σ be the Stefan-Boltzmann constant, with values ​​ranging from 1 to 10. ε is the surface emissivity of the enclosure. In some embodiments, the value of ε ranges from 0.8 to 0.95. A is the effective heat dissipation area of ​​the enclosure. S(t) is the solar radiation correction factor. In some embodiments, the solar radiation correction factor S(t) can be calculated from the latitude and longitude, date and time of the cell location. This refers to the outer surface temperature of the enclosure. The ambient temperature.

[0145] Step S104: Predict the cell temperature based on the current temperature, the first heat generation power curve, and the ambient heat exchange power.

[0146] For example, in some embodiments, the predicted heat generation power curve of the battery cell can be predicted by using the first heat generation power curve of the battery cell and the heat exchange power of the environment. Based on the predicted heat generation power curve of the battery cell, the predicted heat generation of the battery cell at a future time can be predicted. By combining the current temperature of the battery cell with the predicted heat generation of the battery cell, the temperature of the battery cell at a future time can be predicted, and the predicted temperature curve of the battery cell can also be obtained.

[0147] In some embodiments, when predicting the predicted heat generation power curve of a battery cell using its first heat generation power curve and ambient heat exchange power, a sliding window mean filtering method can be used to process the relevant data used in the calculation process. This setting can effectively smooth data fluctuations and reduce the interference of abnormal data on the prediction results. In some embodiments, the predicted heat generation power curve can also be updated every 60 seconds to improve the timeliness and accuracy of the prediction.

[0148] In the above method, considering the cell's aging correction factor, charge deviation factor, and ambient heat exchange power when predicting cell temperature can significantly improve the accuracy of temperature prediction. For example, when the cell's state of heat (SOH) is low, the aging correction factor α(SOH) increases, and the calculated heat generation power Pcell increases accordingly. If the cell absorbs heat from the external environment at this time, and the ambient heat exchange power is positive, then the predicted cell temperature will rise, and the predicted temperature curve will have a higher slope.

[0149] In other embodiments, step S102 can be performed as follows: Figure 12 The method shown is implemented in the following way, specifically including steps S121 to S127.

[0150] Step S121: Obtain the historical charge-discharge cycle count of the battery cell.

[0151] Generally, the higher the historical charge-discharge cycle count of a battery cell, the higher its aging level and the greater its aging correction factor. In some embodiments, a battery cell completing one charge and one discharge cycle sequentially can be considered as completing one cycle. Specifically, a charge cycle can be defined as completed when the battery cell reaches a state of charge of 80% or higher, and a discharge cycle can be defined as completed when the battery cell reaches a state of charge of 20% or lower. These thresholds can be adjusted according to actual usage requirements.

[0152] Step S122: Obtain the cumulative temperature difference value of the historical over-temperature events of the battery cell.

[0153] The energy storage system has a temperature threshold for the battery cells under normal operating conditions. When the current temperature of the battery cell exceeds the temperature threshold, it is recorded as an over-temperature event. Each over-temperature event corresponds to a cumulative temperature value. For example, in a historical over-temperature event, if the temperature of the battery cell exceeds the temperature threshold by 5°C, then the temperature difference with the corresponding historical over-temperature event is 5°C.

[0154] In some embodiments, the temperature difference of all historical overheating events within a specific time period tracing back from the current moment is obtained, and the cumulative value of the absolute value of the temperature difference is calculated to obtain the cumulative value of the temperature difference.

[0155] Step S123: Obtain the current health status of the battery cell based on the cumulative temperature difference and the number of historical charge-discharge cycles.

[0156] In some embodiments, the current health status of the battery cell can be calculated based on expressions 1-5.

[0157]

[0158] in, This refers to the current health status of the battery cell. k8 is the seventh preset parameter, N is the eighth preset parameter, and N is the number of charge / discharge cycles. This represents the cumulative temperature difference from historical overheating events. In some embodiments, ≥0, ≥0. As the number of charge-discharge cycles of the battery cell increases and over-temperature events accumulate, the health of the battery cell will decline.

[0159] Step S124: Obtain the aging correction coefficient of the battery cell based on the current health status.

[0160] The specific implementation of step S124 can be referred to step S111, and will not be repeated here.

[0161] Step S125: Obtain the charge deviation coefficient of the cell based on the current state of charge.

[0162] The specific implementation of step S125 can be referred to step S112, and will not be repeated here.

[0163] Step S126: Obtain the second heat generation power curve of the battery cell based on the predicted charge and discharge power curve.

[0164] The specific implementation of step S126 can be referred to step S113, and will not be repeated here.

[0165] Step S127: Obtain the third relation, and predict the first heat generation power curve of the cell based on the third relation, aging correction coefficient, charge deviation coefficient, and second heat generation power curve.

[0166] The specific implementation of step S127 can be referred to step S114, and will not be repeated here.

[0167] In the above setup, when predicting the first heat generation power curve of a battery cell using its current health status, the current health status of the cell is first obtained based on historical over-temperature events. This setup can take into account the aging damage caused by abnormal temperature fluctuations during actual use, which can improve the accuracy of the aging correction coefficient. For example, if a battery cell has experienced multiple over-temperature events with large temperature differences, its cumulative temperature difference value is large, and the SOH value calculated according to expression 1-5 will decrease accordingly, which in turn leads to an increase in the aging correction coefficient α(SOH), resulting in an increase in the predicted heat generation power Pcell, which is more in line with the actual heat generation situation.

[0168] In some embodiments, this application further proposes a charge-discharge prediction model, constructed using the charge-discharge prediction method described in any of the above embodiments, configured to: receive current data of the battery cell and a historical charge-discharge event library, and output the predicted charge-discharge power curve of the battery cell.

[0169] In some embodiments, this application further proposes a cell heat generation prediction model, configured to: receive the predicted charge / discharge power curve of the cell and the current ambient heat exchange power, and output the predicted heat generation power curve of the cell.

[0170] The specific implementation method for predicting the predicted heat generation power curve of the battery cell using the predicted charge and discharge power curve of the battery cell and the current ambient heat exchange power can refer to the above embodiments. For example, the first heat generation power curve of the battery cell is predicted using the predicted charge and discharge power curve, and then the predicted heat generation power curve of the battery cell is obtained based on the first heat generation power curve and the ambient heat exchange power. This will not be elaborated here.

[0171] In some embodiments, this application further proposes a cell temperature prediction model, constructed using the cell temperature prediction method described in any of the above embodiments, configured to: receive the predicted charge / discharge power curve of the cell, the current ambient heat exchange power, and the current temperature of the cell, and output the predicted temperature of the cell. Specific implementation methods for the cell temperature prediction method can be found in the above embodiments and will not be repeated here.

[0172] The cells of an energy storage system can store and utilize electrical energy through cyclic charging and discharging. As the application scenarios of energy storage systems become increasingly complex, existing energy storage systems are usually equipped with energy storage thermal management systems. These existing energy storage thermal management systems can only perform thermal management based on preset fixed modes, resulting in high energy consumption.

[0173] This application further proposes a control method for an energy storage thermal management system, such as... Figure 13 As shown, the control method includes steps A11 to A14.

[0174] Step A11: Obtain historical operating data and current input parameters of the thermal management unit.

[0175] Current input parameters refer to the input parameters of the thermal management unit at the current moment or under the current operating conditions, such as the current environmental parameters and current operating parameters, such as the current ambient temperature, supply liquid temperature, return liquid temperature, compressor speed, water pump flow rate, valve opening, fan speed, etc.

[0176] Step A12: Obtain a cooling capacity-energy consumption prediction model based on historical operating data.

[0177] Different input parameters correspond to different cooling capacities and energy consumptions. Historical operating data includes the historical input parameters of the thermal management unit and their corresponding cooling capacities and energy consumptions. Therefore, a cooling capacity-energy consumption prediction model can be constructed based on historical operating data. This model can predict the energy consumption and cooling capacity of the thermal management unit under specific input parameters.

[0178] To improve the prediction accuracy of the cooling capacity-energy consumption prediction model, in some embodiments, the input parameters of the thermal management unit include at least environmental parameters and operating parameters. Historical input parameters are the input parameters of the thermal management unit prior to the current moment, representing the input parameters of the thermal management unit at historical points in time. These historical input parameters can be used to train the cooling capacity-energy consumption prediction model. For example, the model can be trained based on historical environmental parameters and historical operating parameters, along with the corresponding cooling capacity and energy consumption, from historical operating data to obtain the cooling capacity-energy consumption prediction model.

[0179] Environmental parameters, which are external conditions during the operation of the thermal management unit, significantly impact the heat load and heat dissipation efficiency of the energy storage system. For example, ambient temperature affects the heat transfer from the external environment to the energy storage system and the heat dissipation from the energy storage system to the external environment. When changes in ambient temperature reduce the temperature difference between the energy storage system and the outside environment, the natural heat dissipation capacity of the energy storage system decreases, requiring the thermal management unit to provide stronger cooling capacity to maintain the cell temperature within a suitable range. In some embodiments, environmental parameters include at least ambient temperature. Ambient temperature is a significant environmental factor affecting the heat load and heat dissipation efficiency of the energy storage system. Among numerous environmental parameters, compared to humidity and wind speed, ambient temperature has a higher impact on cell self-heating, energy storage system heat dissipation, and the cooling / heating efficiency of the thermal management unit. In some application scenarios, higher ambient temperatures increase cell heat generation, reduce the cooling efficiency of the thermal management unit, and increase the energy consumption of the thermal management unit. Therefore, referencing ambient temperature when obtaining the cooling capacity-energy consumption prediction model can improve the accuracy of the model's predictions of cooling capacity and energy consumption.

[0180] In some embodiments, environmental parameters may also include the latitude and longitude of the location of the energy storage system, the date and time, etc.

[0181] Operating parameters are the internal system parameters of the thermal management unit during operation. In some embodiments, the thermal management unit is a liquid-cooled unit, which includes at least a refrigerant circulation loop and a coolant circulation loop. The liquid-cooled unit includes at least a compressor assembly, an evaporator assembly, a condenser assembly, a water pump assembly, and a cold plate assembly. The compressor assembly is located in the refrigerant circulation loop to pump refrigerant, and the water pump assembly is located in the coolant circulation loop to pump coolant. The coolant exchanges heat with energy storage elements such as battery cells through the coolant circulation loop, and the refrigerant exchanges heat with the coolant through the refrigerant circulation loop. In some embodiments, the energy storage elements of the energy storage system are located on the cold plate assembly, which is located in the coolant circulation loop, and the coolant flows through the cold plate assembly to exchange heat with the energy storage elements such as battery cells.

[0182] In some embodiments, the liquid cooling unit further includes a fan assembly, which is disposed on or near the condenser assembly and can drive airflow through the condenser assembly to remove the heat released by the refrigerant at the condenser assembly and accelerate the condensation and heat release process of the refrigerant.

[0183] In some embodiments, the liquid-cooled unit also includes a throttling component such as an electronic expansion valve. This throttling component is located in the refrigerant circulation loop and is used to throttle and depressurize the refrigerant flowing from the condenser assembly, reducing its pressure to the evaporation pressure of the evaporator assembly. The opening degree of the throttling component can affect state parameters such as the subcooling degree of the refrigerant as it flows out of the throttling component, thereby affecting its heat exchange efficiency within the evaporator assembly.

[0184] In some embodiments, the liquid cooling unit further includes a regulating valve disposed in the coolant circulation loop for regulating the circulation state of the coolant. In some embodiments, the regulating valve includes a three-way regulating valve capable of regulating the temperature of the coolant flowing into the cold plate assembly, i.e., regulating the supply temperature.

[0185] In some embodiments, the operating parameters include at least one of the following: supply liquid temperature, return liquid temperature, compressor speed, water pump flow rate, valve opening, fan speed, high pressure value on the high-pressure side of the refrigerant circulation loop, low pressure value on the low-pressure side of the refrigerant circulation loop, evaporation temperature, and condensation temperature.

[0186] Taking a liquid-cooled unit as an example, the supply temperature refers to the temperature of the coolant at the inlet of the cold plate assembly; the return temperature refers to the temperature of the coolant at the outlet of the cold plate assembly. The water pump flow rate reflects the circulating flow rate of the coolant. In some embodiments, the valve opening degree includes at least one of the opening degree of the aforementioned throttling component and the opening degree of the regulating valve. Specifically, the high-pressure side of the refrigerant circulation loop refers to the region where the refrigerant is under high pressure between the outlet of the compressor assembly and the inlet of the condenser assembly, while the low-pressure side refers to the region where the refrigerant is under low pressure between the inlet of the compressor assembly and the outlet of the evaporator assembly. In the refrigerant circulation loop, a first pressure detection point is set between the outlet of the compressor assembly and the inlet of the condenser assembly, and a second pressure monitoring point is set between the inlet of the compressor assembly and the outlet of the evaporator assembly. The pressure value detected by the first pressure detection point is taken as the high-pressure value on the high-pressure side, and the pressure value detected by the second pressure detection point is taken as the low-pressure value on the low-pressure side. During normal system operation, the pressure value detected by the first pressure detection point is significantly higher than the pressure value detected by the second pressure detection point.

[0187] The supply and return temperatures of the coolant can reflect the heat exchange effect between the coolant and the battery cells to a certain extent. Parameters such as compressor speed, water pump flow rate, valve opening, and fan speed can affect the refrigerant circulation efficiency and coolant circulation efficiency of the thermal management unit to a certain extent. Valve opening can also be used to adjust the flow distribution of different loops to adapt to the heat dissipation needs of different areas within the energy storage system. Parameters such as the high-pressure value on the high-pressure side of the refrigerant circulation loop, the low-pressure value on the low-pressure side of the refrigerant circulation loop, evaporation temperature, and condensation temperature can reflect the actual circulation of the refrigerant in the refrigerant circulation loop and the actual operating conditions of components such as the compressor assembly, evaporator assembly, condenser assembly, and throttling assembly in the thermal management unit. These operating parameters directly affect the heat exchange efficiency and workload of the thermal management unit, and thus the cooling capacity and energy consumption of the thermal management system. For example, in a certain application scenario under refrigeration conditions, a lower supply temperature results in stronger cooling capacity, but may lead to increased energy consumption and system overcooling; a higher compressor speed results in greater cooling capacity, but also increases energy consumption. Therefore, referring to the above parameters when obtaining the cooling capacity-energy consumption prediction model can improve the accuracy of the model's prediction of cooling capacity and energy consumption.

[0188] In some embodiments, the liquid supply temperature is adjusted within a range of 18℃-30℃, and in other embodiments, it can be set to 15℃-40℃, etc. In some embodiments, the compressor speed is adjusted within a range of 30%-100%, and in other embodiments, it can be set to 10%-100%, 20%-90%, 20%-100%, etc. In some embodiments, the opening of the solenoid valve is controlled in a closed-loop manner using a proportional-integral-derivative (PID) control algorithm to control the water pump flow rate. For example, a PID controller is used to control the solenoid valve.

[0189] In some embodiments, historical operating data also includes the operating conditions and corresponding control modes corresponding to the historical input parameters. This data can, to a certain extent, reflect the cooling capacity and energy consumption of the thermal management unit under different operating conditions and control modes.

[0190] Step A13: Predict the required cooling capacity of the battery cell based on the cell temperature prediction model.

[0191] In some embodiments, the cell temperature prediction model can be obtained by referring to the construction method of the cell temperature prediction model in the above embodiments, which will not be repeated here.

[0192] In some embodiments, the predicted temperature of the battery cell at a future time is obtained based on a cell temperature prediction model, a temperature threshold for the battery cell is obtained, and the required cooling capacity of the battery cell at a future time is calculated based on the difference between the temperature threshold and the predicted temperature. In other embodiments, the required cooling capacity of the battery cell can also be calculated in other ways.

[0193] A temperature threshold is a temperature range that ensures the battery cell operates safely. In some embodiments, the temperature threshold may include at least one of a maximum temperature threshold and a minimum temperature threshold. In some embodiments, the temperature threshold value can be determined based on factors such as the type of battery cell, its chemical properties, the usage scenario, and safety standards. For example, in one application scenario, when the predicted temperature exceeds the temperature threshold range, the thermal management unit activates the corresponding control strategy in advance to provide the corresponding cooling capacity and control the battery cell temperature within a safe range. When the predicted temperature is less than the minimum temperature threshold, the required cooling capacity is less than 0, and the thermal management unit starts the heating program; when the predicted temperature is greater than the maximum temperature threshold, the required cooling capacity is greater than 0, and the thermal management unit starts the cooling program.

[0194] Step A14: With the goal of minimizing energy consumption, a target control strategy is developed based on the cooling capacity-energy consumption prediction model, the required cooling capacity, and the current input parameters to obtain the adjustable input parameters of the thermal management unit, thereby controlling the operation of the thermal management unit.

[0195] In some embodiments, the cooling capacity-energy consumption prediction model is configured to receive input parameters to be measured and output predicted cooling capacity and predicted energy consumption. The input parameters to be measured refer to the input parameters of the thermal management unit under the measured operating condition or at the measured time, such as environmental and operational parameters at the measured time.

[0196] In some embodiments, the cooling capacity-energy consumption prediction model has adjustable input parameters, which are typically controllable variables among the operating parameters of the thermal management unit, and the current input parameters include the adjustable input parameters. For example, in some embodiments, the adjustable input parameters include liquid supply temperature, compressor speed, water pump flow rate, valve opening degree, and fan speed. These operating parameters can quickly change the cooling capacity and energy consumption of the thermal management unit. A control strategy that obtains the adjustable input parameters helps to control the operating state of the thermal management unit based on the control strategy.

[0197] The control strategy is composed of specific values ​​of adjustable input parameters. When controlling the thermal management unit, adjusting the adjustable input parameters to the values ​​determined by the control strategy enables the thermal management unit to achieve specific cooling capacity and energy consumption targets. To improve the safety of battery cell usage, when the required cooling capacity is greater than 0, the thermal management unit needs to provide cooling, and the cooling capacity of the thermal management unit needs to be greater than or equal to the required cooling capacity of the battery cell. However, among multiple feasible control strategies that satisfy the requirement of cooling capacity being greater than or equal to the required cooling capacity, the energy consumption of the thermal management unit corresponding to each control strategy usually differs. The above embodiment utilizes a cooling capacity-energy consumption prediction model to establish the correspondence between cooling capacity, energy consumption, and input parameters. Therefore, with the minimum energy consumption as the optimization objective and the predicted cooling capacity being greater than or equal to the required cooling capacity of the battery cell as the constraint, solving the cooling capacity-energy consumption prediction model can yield the target control strategy with adjustable input parameters. Controlling the thermal management unit based on this target control strategy can significantly reduce the operating energy consumption of the thermal management unit while meeting the required cooling capacity.

[0198] In some embodiments, optimization algorithms can be used to solve the cooling capacity-energy consumption prediction model, such as genetic algorithms, particle swarm optimization algorithms, simulated annealing algorithms, ant colony optimization algorithms, or gradient descent methods.

[0199] In some embodiments, the cooling capacity-energy consumption prediction model can output predicted cooling capacity and predicted energy efficiency ratio (EER). The predicted energy consumption can be calculated using these two values. Taking an example where the required cooling capacity is greater than 0 and a thermal management unit is needed for cooling, in one application scenario, when using an optimization algorithm to solve the problem, the constraints are set as follows: The optimization objective is:

[0200] in, This is the predicted cooling capacity. The required cooling capacity for the battery cells This is the predicted energy consumption value. This is the predicted energy efficiency ratio. For ambient temperature, For the liquid supply temperature, This refers to the compressor speed. The above method can predict the cooling capacity and energy consumption of the thermal management unit during the period t0-t4.

[0201] In one application scenario, t0 is the start time of a charge / discharge event, and t4 is the end time of a charge / discharge event. The meaning and implementation details of a charge / discharge event can be found in the above embodiments and will not be repeated here. In another application scenario, t4 represents the current time when the controller executes the above control method once, and t4 represents the prediction time domain. At any future discrete moment within the range, , ,and In some embodiments, the prediction time domain The time interval is 1 hour, and k is an integer. It represents the time interval between two adjacent discrete moments.

[0202] The above method can obtain a cooling capacity-energy consumption prediction model by acquiring historical operating data of the thermal management unit, and then obtain a target control strategy. This makes the target control strategy more in line with actual operating conditions, enabling the energy storage thermal management system to better adapt to actual operating conditions and improve thermal management effectiveness. Furthermore, predicting the required cooling capacity of the battery cells based on the cell prediction model and determining the target control strategy based on the required cooling capacity can improve the response speed of the energy storage thermal management system, thereby improving thermal management effectiveness. Moreover, obtaining a target control strategy with the lowest energy consumption as the optimization objective can effectively reduce the energy consumption of the energy storage thermal management system.

[0203] In some embodiments, when the battery's required cooling capacity is less than 0, it indicates that the predicted temperature of the battery cell at a future time is less than the minimum temperature threshold, and the heating program of the thermal management unit needs to be activated. In this case, a target control strategy under minimum energy consumption can be solved using a method similar to that described for the cooling condition. Correspondingly, the constraint is that the predicted cooling capacity of the thermal management unit is less than or equal to the required cooling capacity (i.e., the heating capacity is greater than or equal to the required heating capacity) to ensure that the heating capacity meets the battery cell's temperature rise requirement. For example, if the required cooling capacity is -500W, then the predicted cooling capacity values ​​are -600W, -560W, -510W, etc., which can meet the heating requirement. Alternatively, the constraint can be set as follows: the predicted heating capacity of the thermal management unit is greater than or equal to the battery cell's required heating capacity, where the required heating capacity is determined by the absolute value of the required cooling capacity, and the predicted heating capacity is determined by the absolute value of the predicted cooling capacity.

[0204] In some embodiments, the cooling capacity-energy consumption prediction model includes a cooling capacity prediction sub-model for predicting cooling capacity and an energy consumption prediction sub-model for predicting energy consumption. In some embodiments, the cooling capacity prediction sub-model is trained separately. In some embodiments, the energy consumption prediction sub-model is trained separately.

[0205] Decomposing the cooling capacity-energy consumption prediction model into a cooling capacity prediction sub-model and an energy consumption prediction sub-model enables targeted prediction and optimization of these two indicators. The cooling capacity prediction sub-model focuses on predicting the cooling capacity of the thermal management unit, and its input parameters can include ambient temperature, predicted cell temperature, temperature threshold, and operating parameters of the thermal management unit. The energy consumption prediction sub-model focuses on predicting the energy consumption of the thermal management unit. This decomposition design makes the model prediction results more targeted. For example, when training the cooling capacity prediction sub-model, the focus can be on the heat exchange and energy conversion processes directly related to cooling capacity; while when training the energy consumption prediction sub-model, more emphasis can be placed on the power characteristics of each component and the power variation under operating conditions.

[0206] In other embodiments, a cooling capacity prediction sub-model or an energy consumption prediction sub-model can also be obtained through joint training. Training data comes from historical operational data. Specifically, the training objectives of the cooling capacity prediction sub-model and the energy consumption prediction sub-model can be correlated. For example, during training, both cooling capacity prediction error and energy consumption prediction error can be considered simultaneously, and their importance can be balanced by setting weights. For instance, during model training, the loss functions of the cooling capacity prediction task and the energy consumption prediction task can be combined to construct a multi-task joint loss function L. total =α·L Q +β·L E , where L Q For the loss of the cooling capacity prediction sub-model, L E The loss function for the energy consumption prediction sub-model is α, where α and β are adjustable weighting coefficients. During training, based on historical operating data from the same batch, the cooling capacity prediction error and energy consumption prediction error are calculated simultaneously, and the result is determined according to the multi-task joint loss function L. total The shared parameters or individual parameters of the two sub-models are updated through backpropagation. This approach enables the cooling capacity prediction sub-model and the energy consumption prediction sub-model to co-optimize during the learning process, thereby improving the coordination and generalization ability of the cooling capacity-energy consumption prediction model. For example, during training, if the actual cooling capacity corresponding to a certain training data is high while the actual energy consumption is low, but the cooling capacity prediction sub-model outputs a low cooling capacity prediction value, the energy consumption prediction sub-model will calculate an energy consumption prediction value based on this prediction value that is closer to the actual energy consumption. Through joint training, the multi-task joint loss function will simultaneously penalize the cooling capacity prediction bias and the energy consumption prediction bias of the training data, driving the model to adjust its parameters so that the prediction results more accurately fit the actual values ​​corresponding to each sub-model.

[0207] In some embodiments, the cooling capacity prediction sub-model includes a cooling capacity prediction function. .

[0208] Here, 'x' represents the input parameters of the cooling capacity prediction sub-model, which may include parameters related to the cooling capacity of the thermal management unit, such as ambient temperature, liquid supply temperature, liquid return temperature, compressor speed, water pump flow rate, valve opening, and fan speed. By inputting these parameters into the cooling capacity prediction sub-model, the predicted cooling capacity that the thermal management unit can provide under a specific combination of input parameters can be obtained. For example, when the ambient temperature rises, the cooling capacity prediction sub-model can determine from historical operating data that a higher compressor speed or a lower liquid supply temperature is required to achieve the same cooling capacity, and output the corresponding predicted cooling capacity value accordingly.

[0209] In some embodiments, the energy consumption prediction sub-model includes an energy consumption prediction function. or energy efficiency ratio prediction function Here, y represents the input parameters of the energy consumption prediction sub-model, which may include parameters such as compressor speed, water pump flow rate, and fan speed. Inputting these parameters into the energy consumption prediction sub-model yields the predicted energy consumption or energy efficiency ratio of the thermal management unit under a specific combination of input parameters. The predicted cooling capacity is then obtained through the cooling capacity prediction sub-model. Then, if the energy consumption prediction sub-model outputs a predicted energy efficiency ratio... Then it can be determined according to the relation. = / Calculate the corresponding predicted energy consumption value.

[0210] In some embodiments, a deep learning model is trained based on historical operating data to obtain a cooling capacity prediction sub-model or an energy consumption prediction sub-model. For example, a multi-layer perceptron (MLP) can be used, which can learn the complex nonlinear relationship between input parameters and cooling capacity through multiple hidden layers, and is particularly suitable for processing tabular data. Another example is a recurrent neural network (RNN) or a long short-term memory (LSTM) network. RNNs can introduce recurrent connections, process sequential data, and capture dependencies in the time dimension. LSTM networks can introduce gating mechanisms to improve long-term dependency problems and can be used to process historical operating data with long-term temporal correlations.

[0211] In some embodiments, the thermal management unit includes operating conditions and control modes; historical operating data also includes operating conditions and corresponding control modes corresponding to historical input parameters. For example, each set of historical input parameters, cooling capacity, and energy consumption in the historical operating data corresponds to a specific operating condition and a specific control mode.

[0212] In some embodiments, the thermal management unit can operate under one or more operating conditions. For example, the operating conditions include at least one of environmental operating conditions and load operating conditions. The environmental operating conditions may include high-temperature operating conditions, normal-temperature operating conditions, and low-temperature operating conditions. The load operating conditions may include high-load operating conditions, medium-load operating conditions, or low-load operating conditions. In some embodiments, the thermal management unit can operate under one or more control modes, such as standby mode, self-circulation mode, cooling mode, and heating mode. Historical operating data includes not only historical input parameters, cooling capacity, and energy consumption, but also corresponding operating condition tags and control mode tags.

[0213] In some embodiments, the thermal management unit is a liquid-cooled unit. In some embodiments, standby mode means that the thermal management unit does not actively perform cooling or heating operations. Standby mode is activated when the cell temperature is within a suitable range and there is no significant thermal management demand. In some embodiments, self-circulation mode means that the thermal management unit uses a water pump to drive the coolant to circulate between the cell and the heat exchanger, using natural convection or forced convection to exchange heat with the external environment to achieve initial regulation of the cell temperature. Self-circulation mode is activated when the cell temperature fluctuation is small or the system's cooling / heating demand is low. In some embodiments, cooling mode means that the core components of the thermal management unit, such as the compressor assembly, condenser assembly, evaporator assembly, and water pump assembly, are activated. Through the phase change cycle of the refrigerant and the heat exchange between the refrigerant and the coolant, the heat released by the cell is absorbed and discharged to the external environment, thereby reducing the cell temperature. In some embodiments, the heating mode refers to the start-up of core components of the thermal management unit, such as the compressor assembly, condenser assembly, evaporator assembly, electric heating element, and water pump assembly. The refrigerant absorbs heat from the external environment through phase change circulation and releases it to the coolant, which then transfers the heat to the battery cell to raise the battery cell temperature to a safe range.

[0214] In some embodiments, when a cooling capacity-energy consumption prediction model is obtained based on historical operating data, the cooling capacity-energy consumption prediction model is configured to: receive the input parameters to be measured, the operating condition to be measured, and the control mode to be measured, and output the predicted cooling capacity and predicted energy consumption values ​​corresponding to the operating condition to be measured and the control mode to be measured.

[0215] When training the cooling capacity-energy consumption prediction model, incorporating operating conditions and control modes as input parameters allows the model to learn the specific correspondence between input parameters and cooling capacity and energy consumption under different operating conditions and control modes. This improves the model's prediction accuracy and adaptability under complex and variable operating conditions. For example, in the cooling mode under high temperature and high load conditions, it is identified that the high ambient temperature and large heat generation of the battery cells require higher compressor speeds and larger water pump flow rates to meet the cooling demand, and the energy consumption of the thermal management unit also increases. In contrast, in the self-circulation mode under normal temperature and low load conditions, only a lower water pump flow rate is needed to maintain stable battery cell temperature, and the energy consumption of the thermal management unit remains at a low level. In this way, the cooling capacity-energy consumption prediction model can predict the cooling capacity and energy consumption of the thermal management unit under different combinations of adjustable input parameters based on the current operating conditions and control modes. This facilitates subsequent optimization algorithms to solve for the target control strategy that satisfies the minimum energy consumption and the constraints.

[0216] In some embodiments, step A12 can be performed as follows: Figure 14 The method shown is implemented in the following way, specifically including steps A21 to A22.

[0217] Step A21: Divide the historical operation data into multiple sample subsets according to the control mode.

[0218] Each set of input parameters, cooling capacity, energy consumption, and operating conditions corresponds to a control mode. Historical operating data is categorized according to the control mode label, forming multiple sample subsets. All sample data within each subset belong to the same control mode.

[0219] For example, in some embodiments, historical operating data is divided into a standby mode sample subset, a self-circulation mode sample subset, a cooling mode sample subset, and a heating mode sample subset.

[0220] Step A22: Train the mode sub-models corresponding to the control mode based on the sample subsets.

[0221] The cooling capacity-energy consumption prediction model is configured to: receive the input parameters to be tested and the control mode to be tested, route to the mode sub-model corresponding to the control mode to be tested based on the control mode to be tested, and output the corresponding cooling capacity prediction value and energy consumption prediction value.

[0222] For example, in some embodiments, multiple mode sub-models corresponding to different control modes are trained based on sample subsets, and all mode sub-models constitute a cooling capacity-energy consumption prediction model.

[0223] For example, the cooling capacity-energy consumption prediction model includes a standby mode sub-model, a self-circulation mode sub-model, a cooling mode sub-model, and a heating mode sub-model.

[0224] In some embodiments, each mode sub-model learns only the specific correspondence between input parameters and cooling capacity and energy consumption under a single control mode, and is able to predict the cooling capacity and energy consumption values ​​under that control mode based on the input parameters to be measured. In some embodiments, each mode sub-model includes a cooling capacity prediction sub-model and an energy consumption prediction sub-model for the corresponding mode. The functional principles of the cooling capacity prediction sub-model and the energy consumption prediction sub-model can be referred to the above embodiments, and will not be repeated here.

[0225] During the model usage phase, when the cooling capacity-energy consumption prediction model receives a prediction request containing the input parameters to be measured and the control mode to be measured, it first performs a routing decision based on the control mode to be measured, and automatically selects the mode sub-model corresponding to that control mode to process the input data. The selected mode sub-model can output cooling capacity prediction values ​​and energy consumption prediction values ​​that match the control mode based on the input data. For example, when the control mode to be measured is cooling mode, the model is routed to the cooling mode sub-model, which processes the input parameters to be measured and outputs the corresponding cooling capacity prediction values ​​and energy consumption prediction values.

[0226] In some embodiments, as time progresses, when a new set of historical operating data is added to the thermal management unit, the parameters of the corresponding sub-mode can be adjusted through incremental learning based on the control mode of the new data to update the cooling capacity-energy consumption prediction model.

[0227] In some embodiments, when a new control mode is added to the thermal management unit, only historical operating data under that control mode needs to be collected to train a corresponding new mode sub-model and connect it to the routing framework, without retraining or adjusting the existing mode sub-model. This setup not only improves the prediction accuracy of the cooling capacity-energy consumption prediction model for specific control modes but also enhances the model's scalability.

[0228] In other embodiments, referring to steps A21 to A22, historical operating data can be divided into multiple sample subsets based on control modes and operating conditions, with each sample subset corresponding to a combination of operating conditions and control modes. Corresponding operating condition-mode sub-models are trained based on each sample subset. All operating condition-mode sub-models constitute a cooling capacity-energy consumption prediction model. In some embodiments, each operating condition-mode sub-model includes a cooling capacity prediction sub-model and an energy consumption prediction sub-model under the corresponding operating condition-control mode combination. The functional principles of the cooling capacity prediction sub-model and the energy consumption prediction sub-model can be referred to the above embodiments and will not be repeated here.

[0229] For example, the cooling capacity-energy consumption prediction model includes sub-models for high temperature and high load conditions - standby mode, high temperature and high load conditions - self-circulation mode, high temperature and high load conditions - cooling mode, high temperature and high load conditions - heating mode, normal temperature and high load conditions - standby mode, normal temperature and high load conditions - self-circulation mode, normal temperature and high load conditions - cooling mode, normal temperature and high load conditions - heating mode, low temperature and high load conditions - standby mode, and low temperature and high load conditions - self-circulation mode. Sub-models for various operating conditions include: cyclic mode, low temperature high load - cooling mode, low temperature high load - heating mode, high temperature medium load - standby mode, high temperature medium load - self-circulation mode, high temperature medium load - cooling mode, high temperature medium load - heating mode, high temperature low load - standby mode, high temperature low load - self-circulation mode, high temperature low load - cooling mode, and high temperature low load - heating mode.

[0230] In other embodiments, a unified modeling approach can also be adopted, that is, a single overall prediction model is trained on historical operating data of all control modes, without distinguishing between mode sub-models or operating condition-mode sub-models.

[0231] In some embodiments, step A12 can be performed as follows: Figure 15 The method shown is implemented in the following way, specifically including steps A31 to A33.

[0232] Step A31: Obtain a cooling capacity-energy consumption prediction model based on the original operating data.

[0233] Historical operational data includes both raw operational data and updated operational data. In some embodiments, the updated operational data is generated more recently than the current time.

[0234] In some embodiments, raw operating data refers to operating data collected by the thermal management unit during the initial operation phase or a specific historical period, including key information such as input parameters, actual cooling capacity, and actual energy consumption under different operating conditions and control modes. This data can provide preliminary learning data for the model. For example, historical operating data from the first six months after the thermal management unit is put into use can be collected as raw operating data.

[0235] Step A32: Obtain historical running data within the sliding window as updated running data.

[0236] In some embodiments, a sliding window is obtained by tracing back a specific period of time from the current moment. In some embodiments, the duration of the sliding window can be set according to the operating characteristics and data update frequency of the thermal management unit, for example, it can be set to 2 days, 5 days, 10 days, 30 days, 60 days or 90 days.

[0237] In some embodiments, the duration of the sliding window can also be determined based on the amount of data within it. For example, starting from the end of the sliding window at the last model update and ending at the current time, when the accumulated update data within the sliding window reaches a preset data volume threshold, an update instruction is triggered, and step A33 is executed.

[0238] In some embodiments, when the prediction error of the cooling capacity-energy consumption prediction model exceeds the prediction error threshold a certain number of times within the sliding window, an update instruction is triggered, and step A33 is executed.

[0239] By dynamically selecting updated operating data through a sliding window, the cooling capacity-energy consumption prediction model can be updated using the latest data from the thermal management unit. This improves the accuracy of model predictions and mitigates the problem of prediction accuracy drifting over time due to outdated operating data. For example, in one application scenario, the thermal management unit experiences component aging and performance degradation during operation, resulting in higher energy consumption for the same cooling capacity. Recent updated operating data reflects this change, and updating the model based on this data improves the model's prediction accuracy.

[0240] In other embodiments, the duration of the sliding window can be adjusted according to the amount of data. For example, when the operating data collection frequency of the thermal management unit is high, a shorter sliding window duration can be set to improve the timeliness of the data; when the collection frequency is low, the sliding window duration can be appropriately extended to increase the amount of data used for model updates.

[0241] Step A33: Based on the updated operating data, adjust the parameters of the cooling capacity-energy consumption prediction model through incremental learning to update the cooling capacity-energy consumption prediction model.

[0242] In some embodiments, incremental updates can be performed based on historical running data within a sliding window using the recursive least squares method.

[0243] Adjusting the parameters of a cooling capacity-energy consumption prediction model through incremental learning involves modifying the model parameters using newly updated operational data, rather than retraining the entire model. This approach effectively reduces computational resource consumption and the time required for model updates. For example, using gradient-based incremental learning, after each acquisition of updated operational data, only the updated data is used to fine-tune the model parameters, enabling the model to quickly adapt to the latest operating status of the thermal management unit.

[0244] In some embodiments, a parameter freezing strategy can also be adopted to freeze the general feature parameters learned in the model and update only the top-level parameters related to the updated running data, thereby further improving the efficiency of model updates.

[0245] This dynamic update mechanism enables the cooling capacity-energy consumption prediction model to maintain high prediction accuracy, which helps to obtain a target control strategy that is more in line with the actual operating conditions of the thermal management unit in the subsequent solution process, so as to achieve the effect of reducing energy consumption and achieving the required cooling capacity.

[0246] In other embodiments, the control method described in other embodiments of this application may also be used to retrain the model using historical running data within a sliding window.

[0247] In some embodiments, step A12 can be performed as follows: Figure 16 The method shown is implemented by steps C11 to C12.

[0248] Step C11: Construct physically inspired features based on historical operational data.

[0249] Physically inspired features are features with clear physical meaning. In some embodiments, physically inspired features are constructed using training data based on thermodynamic and fluid dynamic relationships. Specifically, input parameters and corresponding cooling capacity and energy consumption for specific time periods can be collected from historical operating data as training data for the model. For example, data sequences of the changes over time for various input parameters of the energy storage thermal management system, such as ambient temperature, supply liquid temperature, return liquid temperature, compressor speed, water pump flow rate, valve opening, fan speed, high pressure value on the high-pressure side of the refrigerant circulation loop, low pressure value on the low-pressure side of the refrigerant circulation loop, evaporation temperature, and condensation temperature, can be collected for a specific historical period.

[0250] In some embodiments, the physical heuristic features include at least the hot-side load features, the refrigeration cycle temperature rise features, the actuator power-related features, and the interaction features.

[0251] In some embodiments, the hot-side load characteristics include the supply-return fluid temperature difference ΔT, where ΔT is the supply-return fluid temperature. Subtract the supply temperature The difference obtained.

[0252] In some embodiments, the hot-side load characteristic also includes load proxy Q. proxy In some embodiments, the load proxy quantity Q proxy For the water pump flow rate F pump A function of the product of the supply and return fluid temperature difference ΔT. For example, the load agent quantity Q. proxy Satisfying expression 2-1: Q proxy =c p·ρ·F pump ·ΔT………2-1 Among them, c p ρ is the specific heat capacity of the coolant, and c is the density of the coolant. In some embodiments, c can be determined based on the actual usage scenario. p The values ​​of ρ and Q. Load proxy quantity Q. proxy It can more accurately reflect the heat generation of energy storage components.

[0253] In some embodiments, the temperature rise characteristic of the refrigeration cycle includes a first temperature rise characteristic lift, which is the difference between the condensing temperature and the evaporating temperature, and can reflect the impact of environmental changes on the system's energy efficiency ratio. The condensing temperature corresponds to the condensing pressure, and the evaporating temperature corresponds to the evaporating pressure. In some embodiments, the first temperature rise characteristic lift can be obtained from the condensing pressure and the evaporating pressure.

[0254] In some embodiments, the refrigeration cycle temperature rise characteristic includes a second temperature rise characteristic, which is the first temperature rise characteristic lift multiplied by the ambient temperature T. amb The product of . When the ambient temperature T amb As the temperature rises, the first temperature rise characteristic (lift) increases, causing the energy storage thermal management system to consume more power to maintain the same cooling capacity, resulting in a decrease in the energy efficiency ratio of the energy storage thermal management system. Therefore, the second temperature rise characteristic can be used to reflect the energy efficiency degradation under high-temperature environments.

[0255] In some embodiments, actuator power-related characteristics include those related to compressor speed (RPM). comp Related power terms, such as RPM comp (RPM) comp ) 2 (RPM) comp ) 3 It can be used to reflect the power consumption of the compressor.

[0256] In some embodiments, actuator power-related characteristics include those related to pump flow rate F. pump Related power terms, such as F pump (F) pump ) 2 (F) pump ) 3 It can be used to reflect the power consumption of the water pump.

[0257] In some embodiments, actuator power-related characteristics include those related to fan speed (RPM). fan Related power terms, such as RPM fan (RPM) fan ) 2 (RPM) fan ) 3It can be used to reflect the power consumption of the wind turbine.

[0258] In some embodiments, the interaction feature includes the pump flow rate F pump The product term with the temperature difference ΔT between the supply and return fluid (F) pump ·ΔT). Taking a liquid-cooled unit as an example, the water pump flow rate F pump The temperature difference ΔT between the supply and return liquid has a coupled effect on the heat exchange efficiency. When the pump flow rate F... pump If the temperature difference between the supply and return liquids is too large, and the temperature difference ΔT is too small, the heat exchange efficiency will decrease to some extent; if the temperature difference ΔT is too large, the flow rate will be insufficient, and the energy storage thermal management system will be prone to overload risk. (F) pump ΔT can reflect the pump flow rate F to a certain extent. pump The coupled effect of the supply-return liquid temperature difference ΔT on heat exchange efficiency can improve the problem of large prediction bias caused by simply fitting a single raw temperature or flow rate data.

[0259] In some embodiments, the interaction feature includes compressor speed (RPM). comp The product term with the first temperature rise characteristic lift (RPM) comp •lift). Different first temperature rise characteristics correspond to different compressor speeds. For example, when the ambient temperature rises, the first temperature rise characteristic lift increases, and the compressor needs to operate at a higher speed to maintain heat exchange efficiency. Product term (RPM) comp •lift can reflect the compressor speed (RPM) to some extent. comp The coupling effect of the first temperature rise feature lift on system energy consumption is improved, thus mitigating the problem of large prediction bias caused by simply fitting a single raw speed data or raw temperature data.

[0260] In some embodiments, the interactive features include a liquid supply temperature setpoint T. supply,set With load proxy volume Q proxy The product term (T) supply,set ·Q proxy Among them, the liquid supply temperature setpoint T supply,set The target supply temperature is adjustable and differs from the actual supply temperature in the flow path. Load capacity Q proxy It is the water pump flow rate F pump A function of the product of the supply and return fluid temperature difference ΔT, where the supply and return fluid temperature difference ΔT is based on the actual return fluid temperature. Subtract the actual supply temperature The data obtained.

[0261] Taking cooling mode as an example, when the load of the energy storage thermal management system changes, such as the load agent Q... proxyTo increase the temperature, the liquid supply temperature setpoint T needs to be adjusted. supply,set To match the load situation. When the load proxy quantity Q proxy When increasing the temperature, the liquid supply temperature setpoint T needs to be reduced. supply,set This aims to improve cooling efficiency and address the issue of decreased system energy efficiency ratio caused by excessively high liquid supply temperature. When the load agent Q... proxy When the temperature is reduced, the setpoint T for the liquid supply temperature can be increased. supply,set To reduce energy consumption and improve the effect of liquid supply temperature The problem of wasted cooling capacity due to excessively low temperatures. This is addressed by using a product term (T). supply,set ·Q proxy As a physically inspired feature, it can better reflect the relationship between energy consumption and heat exchange, and improve the prediction accuracy of the cooling capacity-energy consumption prediction model.

[0262] In some embodiments, the training data is first preprocessed, and then the preprocessed training data is used to construct physically inspired features.

[0263] For example, in some embodiments, data preprocessing includes time alignment of training data with different sampling frequencies to a uniform sampling period. In other embodiments, data preprocessing includes completing training data with missing data and removing outlier data. For example, in some embodiments, training data with quality anomalies due to communication packet loss, sensor erratic behavior, or the existence of physically impossible values ​​(such as sustained negative temperature differences, pressure exceeding limits, etc.) is removed. In still other embodiments, data preprocessing includes processing extreme values ​​in the training data, such as using quantile pruning or robust loss functions, to improve the accuracy of model training.

[0264] Step C12: Construct a cooling capacity-energy consumption prediction model based on physical heuristic features using linear regression.

[0265] For example, in some embodiments, a modeling framework based on multiple linear regression is used. First, the aforementioned physical heuristic features are constructed based on thermodynamic and fluid dynamic relationships. Then, using the physical heuristic features as independent variables, multiple linear regression is employed for fitting, and model parameters such as regression coefficients and biases are solved to construct a cooling capacity-energy consumption prediction model. Unlike simply fitting raw sampled data from historical operating data linearly, this method can explicitly encode key nonlinear relationships in the energy storage thermal management system by constructing physical heuristic features, thereby achieving a linearized expression of the model and maintaining a linear structure in the model's parameter space. The linear structure facilitates online model updates; for example, algorithms such as recursive least squares can be used to update model parameters online, improving the adaptability of the cooling capacity-energy consumption prediction model. The linear structure makes it easy to introduce constraints; for example, regularization terms such as ridge regression or physical inequality constraints can be easily added to improve the model's generalization ability and physical consistency. Furthermore, there is often a direct and quantifiable correspondence between model parameters and physical heuristic features in the linear structure, which can improve the interpretability of the model's decision-making process. Therefore, the cooling capacity-energy consumption prediction model obtained in this way can improve the prediction accuracy to a certain extent and effectively improve the problem of insufficient model interpretability caused by complex nonlinear implicit fitting of common black box models.

[0266] In some embodiments, when constructing a cooling capacity-energy consumption prediction model using linear regression, at least one of the following optimization methods may be included. For example, regularized training may be employed, adding a regularization term to the loss function to address feature collinearity or achieve feature sparsity. Another example is selecting a specific subset of features from the aforementioned physically inspired features for modeling based on the Akaike information criterion (AIC), Bayesian information criterion (BIC), recursive feature elimination methods, or stepwise regression. Yet another example is using K-fold cross-validation to divide the training and validation sets, and using this method to determine the regularization coefficients and relevant parameters in the model. Finally, physical consistency constraints may be applied, imposing sign or monotonicity constraints on some regression coefficients; for example, setting the pump flow rate F... pump The power term (F) pump ) 3 The coefficient is a non-negative number, and the compressor speed is in RPM. comp Power term (RPM) comp ) 3 The coefficients are non-negative to reduce the likelihood of the model producing predictions that defy physical laws.

[0267] In some embodiments, step A12 can be performed as follows: Figure 17The method shown is implemented by steps C21 to C24.

[0268] Step C21: Obtain training data based on historical running data.

[0269] In some embodiments, input parameters and corresponding cooling capacity and energy consumption for a specific period of time are first collected from historical operating data as training data for the model. For example, data sequences of the changes of various input parameters such as ambient temperature, liquid supply temperature, liquid return temperature, compressor speed, water pump flow rate, valve opening, fan speed, high pressure value on the high-pressure side of the refrigerant circulation loop, low pressure value on the low-pressure side of the refrigerant circulation loop, evaporation temperature, and condensation temperature over time are collected within a specific historical period.

[0270] Step C22: Identify the steady-state segment data in the training data that is in the steady-state period and the transition segment data that is in the transition period based on preset rules.

[0271] In practical applications, when an energy storage thermal management system operates stably under a specific operating condition or control mode, its operating and environmental parameters are relatively stable, and their changes over time are relatively gradual; this period is the steady-state period of the energy storage thermal management system. When the energy storage thermal management system faces situations such as mode switching, operating condition switching, sudden load changes, or start-up and shutdown, its operating and environmental parameters will change abruptly due to the state transition; this period is the transitional period of the energy storage thermal management system. By using preset rules to divide the training data into steady-state data that is easy to fit and transitional data that is more difficult to fit, it is easier to select different data segments based on different training objectives during subsequent model training, thereby improving the accuracy of model training.

[0272] For example, in some embodiments, the corresponding data sequence is divided into steady-state data and transitional data based on the rate of change threshold and minimum duration of the training data. Taking compressor speed as an example, if the rate of change of the compressor speed at the current moment is less than or equal to the compressor speed rate of change threshold, and the duration of this state exceeds the preset minimum duration, the corresponding data segment is marked as steady-state data; otherwise, it is marked as transitional data.

[0273] Specifically, during transitional periods such as compressor start-up / stop, load switching, or control mode changes, the compressor speed typically changes rapidly, with the rate of change often exceeding a preset threshold and the duration of the change being short. Such data segments can be identified as transitional period data using the method described above. Conversely, during steady-state periods of compressor operation, the speed fluctuations are gentle, the rate of change remains below the preset threshold, and the stable state lasts for a longer period. Such data segments can be identified as steady-state period data using the method described above.

[0274] Step C23: Train the basic prediction model using the linear regression method based on the steady-state data.

[0275] The training data includes input parameters such as ambient temperature and compressor speed, along with their corresponding cooling capacity and energy consumption. This data represents historical operational data, and the physical heuristics constructed from the training data also have corresponding cooling capacity and energy consumption. In some embodiments, training data at different sampling times are labeled according to a division into steady-state and transitional states. A basic prediction model can be trained using linear regression based on the physical heuristics corresponding to the steady-state data.

[0276] In some embodiments, the basic prediction model includes a cooling capacity basic prediction model and an energy consumption basic prediction model. Step C23 can be implemented as follows: the cooling capacity basic prediction model and the energy consumption basic prediction model are trained respectively using a linear regression method based on the physical heuristic features corresponding to the steady-state data.

[0277] In some embodiments, the basic prediction model for cooling capacity, trained using a linear regression method based on the physical heuristic features corresponding to the steady-state data, satisfies expression 2-2: Q cool =β0+Σβ k ·φ k (x)………2-2 Where, φ k (x) is a physically inspired feature, β k β0 is the regression coefficient corresponding to the physical heuristic feature, and β0 is the bias term of the linear equation.

[0278] In some embodiments, the energy consumption prediction model is trained using a linear regression method based on the physical heuristic features corresponding to the steady-state data, satisfying expression 2-3: E=γ0+Σγ k ·ψ k (x)………2-3 Where, ψ k (x) is a physically inspired feature, γ k γ is the regression coefficient corresponding to the physical heuristic feature, and γ0 is the bias term of the linear equation.

[0279] In some embodiments, the predicted cooling capacity and energy consumption of the energy storage thermal management system during steady-state periods can be directly predicted using a basic prediction model. In some embodiments, the predicted cooling capacity of the energy storage thermal management system during steady-state periods can be directly predicted using a basic cooling capacity prediction model. In some embodiments, the predicted energy consumption of the energy storage thermal management system during steady-state periods can be directly predicted using a basic energy consumption prediction model.

[0280] In some embodiments, step C23 can be implemented as follows: based on the physical heuristic features corresponding to the steady-state segment data, the optimal parameters are solved using ordinary least squares or ridge regression to obtain the basic prediction model.

[0281] For example, ordinary least squares can be used to solve for the optimal model parameters with the objective of minimizing the sum of squared prediction errors. In some embodiments, the model parameters include regression coefficients and bias terms.

[0282] For example, ridge regression can be used to reduce the collinearity among multiple physical heuristic features and improve the model's generalization ability.

[0283] In some embodiments, to improve the model's fitting accuracy for key operating conditions such as high-load conditions and near-temperature control boundary conditions, step C23 can be implemented as follows: obtain operating condition labels corresponding to the steady-state segment data, where the operating condition labels include at least key operating conditions and ordinary operating conditions; set different training weights for the steady-state segment data corresponding to different operating condition labels, wherein the training weights for the steady-state segment data corresponding to key operating conditions are greater than the training weights for the steady-state segment data corresponding to ordinary operating conditions; construct physical heuristic features based on the steady-state segment data; and train the basic prediction model using a weighted linear regression method based on the physical heuristic features and their corresponding training weights.

[0284] In some embodiments, critical operating conditions include at least high-load operating conditions and near-temperature control boundary operating conditions. In some embodiments, weighted least squares method is used for weighted linear regression fitting to improve the prediction accuracy of the model under particularly critical operating conditions.

[0285] In other embodiments, the basic prediction model can also be trained using the above-described linear regression method based on the physical heuristics corresponding to all training data, without distinguishing between steady-state and transitional data.

[0286] Step C24: Based on the transition state data, revise the basic prediction model to obtain the cooling capacity-energy consumption prediction model.

[0287] In some embodiments, step C24 can be performed as follows: Figure 18 The method shown is implemented by steps C31 to C33.

[0288] Step C31: Obtain the prediction residuals of the basic prediction model during the transition period.

[0289] In practical applications, due to equipment lag and thermal inertia, the predictions made by the basic prediction model during the transition period may contain systematic biases. Based on the transition period data, the basic prediction model can be used to obtain predicted values ​​for cooling capacity and energy consumption. These predicted values ​​are then compared with the actual values ​​for cooling capacity and energy consumption to obtain the prediction residuals. For example, in some embodiments, transition period data and the corresponding actual values ​​for cooling capacity and energy consumption are obtained; the basic prediction model is used to predict the transition period data to obtain basic predicted values; and the difference between the actual values ​​and the basic predicted values ​​is used to obtain the prediction residuals.

[0290] Step C32: Train the correction model based on the physical heuristic features corresponding to the predicted residuals and transition state data.

[0291] In some embodiments, a fixed modified model is obtained by training an autoregressive exogenous model based on the prediction residuals. In some embodiments, a dynamic modified model is obtained by online training using recursive least squares based on the prediction residuals, whereby the recursive least squares method can continuously update the modified model based on incremental updates.

[0292] In some embodiments, the correction model includes a cooling capacity correction model and an energy consumption correction model, and the cooling capacity correction model e can be trained separately using the methods described above. Q(t) With energy consumption correction model e E(t) .

[0293] Step C33: Obtain the cooling capacity-energy consumption prediction model based on the basic prediction model and the modified model.

[0294] For example, a cooling capacity-energy consumption prediction model includes a cooling capacity prediction sub-model and an energy consumption prediction sub-model, wherein the cooling capacity prediction sub-model... Satisfying expression 2-4: =Q cool +e Q(t) ………2-4 Among them, Q cool Satisfying expression 2-2, e Q(t) This is a model for correcting cooling capacity.

[0295] Energy consumption prediction sub-model Satisfying expression 2-5: =E+e E(t) ………2-5 Where E satisfies expression 2-3, e E(t) This is an energy consumption correction model.

[0296] The steady-state period of an energy storage thermal management system is typically longer than its transitional period. If steady-state and transitional data are not distinguished, the trained cooling capacity-energy consumption model is prone to greater prediction errors during the transitional period. By using the method described above—training a basic prediction model using steady-state data and a modified model using transitional data—combining the basic and modified models to obtain the cooling capacity-energy consumption model can improve the accuracy of the cooling capacity-energy consumption model for the transitional period. For example, the cooling capacity prediction sub-model and energy consumption prediction sub-model obtained based on expressions 2-4 and 2-5, respectively, ensure that the main part of the model has a linear structure, and the dynamic response capability of the model can be improved through the modified model.

[0297] In some embodiments, the cooling capacity-energy consumption prediction model includes a routing layer and a prediction layer. The prediction layer includes multiple sub-models. In some embodiments, the prediction layer includes multiple cooling capacity prediction sub-models and energy consumption prediction sub-models corresponding to different operating condition-control mode combinations. The cooling capacity-energy consumption prediction model is configured to perform prediction by selecting the corresponding sub-model based on the current control mode and operating condition through the routing layer. In some embodiments, the prediction layer includes multiple cooling capacity prediction sub-models and energy consumption prediction sub-models corresponding to different operating conditions. The cooling capacity-energy consumption prediction model is configured to perform prediction by selecting the corresponding sub-model based on the current operating condition through the routing layer. In some embodiments, the prediction layer includes multiple cooling capacity prediction sub-models and energy consumption prediction sub-models corresponding to different control modes. The cooling capacity-energy consumption prediction model is configured to perform prediction by selecting the corresponding sub-model based on the current control mode through the routing layer.

[0298] In this application, the training methods of other embodiments can be used to train the models individually to obtain different multiple cooling capacity prediction sub-models and multiple energy consumption prediction sub-models corresponding to different operating condition-control mode combinations. When using the cooling capacity-energy consumption prediction model for prediction, when the cooling capacity-energy consumption prediction model receives a prediction request containing the input parameters to be tested and the operating condition-control mode combination to be tested, it first performs a routing judgment based on the operating condition-control mode combination to be tested, and automatically selects the sub-model corresponding to the operating condition-control mode combination to process the input data. For example, when the combination of the test condition and control mode is high load condition-standby mode, the routing layer first selects the cooling capacity prediction sub-model and energy consumption prediction sub-model corresponding to the high load condition-standby mode. Then, the basic prediction value of cooling capacity is output using the basic prediction model of cooling capacity, and the basic prediction value of energy consumption is output using the basic prediction model of energy consumption. Then, the correction value of cooling capacity is output using the correction model of cooling capacity, and the correction value of energy consumption is output using the correction model of energy consumption. Thus, the predicted value of cooling capacity = the basic prediction value of cooling capacity + the correction value of cooling capacity, and the predicted value of energy consumption = the basic prediction value of energy consumption + the correction value of energy consumption.

[0299] In other embodiments, after obtaining the cooling capacity-energy consumption prediction model, for data in the steady state period, the output of the basic prediction model of the cooling capacity-energy consumption prediction model can be used as the corresponding cooling capacity prediction value and energy consumption prediction value; for data in the transition state period, the sum of the output of the basic prediction model and the output of the dynamic correction model can be used as the corresponding cooling capacity prediction value and energy consumption prediction value.

[0300] In some embodiments, step A12 can be performed as follows: Figure 19 The method shown is implemented by steps C41 to C45.

[0301] Step C41: Obtain training data based on historical running data.

[0302] The specific implementation of step C41 can be referred to the above embodiments, and will not be repeated here.

[0303] Step C42: Identify the steady-state segment data in the training data that is in the steady-state period and the transition segment data that is in the transition period based on preset rules.

[0304] The specific implementation of step C42 can be referred to the above embodiments, and will not be repeated here.

[0305] Step C43: Construct physically inspired features based on training data.

[0306] The specific implementation method for constructing physically inspired features based on training data can be referred to the above embodiments, and will not be repeated here. The training data includes steady-state segment data and transitional segment data. The physically inspired features constructed based on the steady-state segment data correspond to the steady-state time period, and the physically inspired features constructed based on the transitional segment data correspond to the transitional time period. In some embodiments, steps C42 and C43 are executed sequentially; in some embodiments, steps C43 and C42 are executed sequentially.

[0307] Step C44: Assign different training weights to the training data at different time periods.

[0308] In some embodiments, based on the method for identifying steady-state data and transitional data in the above embodiments, each sampling moment in the data sequence corresponding to input parameters such as compressor speed can be marked as a steady-state period or a transitional period, thereby determining the steady-state data and the transitional data.

[0309] In some embodiments, the duration of the steady-state period of the energy storage thermal management system is typically longer than the duration of its transitional period, and the proportion of steady-state data is typically greater than the proportion of transitional data. In some embodiments, setting the training weight of transitional data to be greater than the training weight of steady-state data can improve the prediction accuracy of the cooling capacity-energy consumption prediction model for the energy storage thermal management system during the transitional period.

[0310] Step C45: Based on the physical heuristic features and the corresponding training weights, a weighted linear regression method is used to train and obtain the cooling capacity-energy consumption prediction model.

[0311] A cooling capacity-energy consumption prediction model was trained using a weighted linear regression method. Adjusting the corresponding training weights can specifically improve the model's prediction accuracy for cooling capacity and energy consumption at different times.

[0312] In some embodiments, the above method can be used to train the cooling capacity prediction sub-model and the energy consumption prediction sub-model separately, and the cooling capacity prediction sub-model and the energy consumption prediction sub-model can be combined to form a cooling capacity-energy consumption prediction model.

[0313] In some embodiments, the cooling capacity-energy consumption prediction model includes a routing layer and a prediction layer. The prediction layer includes multiple sub-models. In some embodiments, the prediction layer includes multiple cooling capacity prediction sub-models and energy consumption prediction sub-models corresponding to different operating condition-control mode combinations. The cooling capacity-energy consumption prediction model is configured to select the corresponding sub-model for prediction based on the current control mode and operating condition through the routing layer. In some embodiments, the prediction layer includes multiple cooling capacity prediction sub-models and energy consumption prediction sub-models corresponding to different operating conditions. The cooling capacity-energy consumption prediction model is configured to select the corresponding sub-model for prediction based on the current operating condition through the routing layer. In some embodiments, the prediction layer includes multiple cooling capacity prediction sub-models and energy consumption prediction sub-models corresponding to different control modes. The cooling capacity-energy consumption prediction model is configured to select the corresponding sub-model for prediction based on the current control mode through the routing layer. The multiple cooling capacity prediction sub-models and multiple energy consumption prediction sub-models under the corresponding implementation can be trained separately with reference to the training methods of other embodiments of this application.

[0314] In some embodiments, the prediction layer includes a basic prediction layer and a correction layer. The basic prediction layer includes the basic prediction model described in the above embodiments, and the correction layer includes the correction model described above.

[0315] In some embodiments, taking the cooling capacity-energy consumption prediction model as an example, which includes multiple cooling capacity prediction sub-models and energy consumption prediction sub-models combined with different operating conditions-control modes, the routing decision method of the routing layer is to directly obtain the operating condition-control mode combination at the current moment from the energy management system. In some embodiments, the routing decision method of the routing layer is to determine the operating condition-control mode combination at the current moment based on preset rules and current operating data.

[0316] Taking a cooling capacity-energy consumption prediction model that includes multiple cooling capacity prediction sub-models and energy consumption prediction sub-models combined with different operating conditions and control modes as an example, where each cooling capacity prediction sub-model includes a basic cooling capacity prediction model and a cooling capacity correction model, and each energy consumption prediction sub-model includes a basic energy consumption prediction model and an energy consumption correction model, the prediction process of the cooling capacity-energy consumption prediction model is illustrated. First, the cooling capacity-energy consumption prediction model receives a prediction request containing the input parameters to be measured and the combination of operating conditions and control modes to be measured. Then, the routing layer performs routing decisions based on the combination of operating conditions and control modes to be measured, selecting the sub-model in the prediction layer corresponding to that combination of operating conditions and control modes to process the input data. For example, when the test condition-control mode combination is high load condition-standby mode, the routing layer first selects the cooling capacity prediction sub-model and energy consumption prediction sub-model corresponding to the high load condition-standby mode in the prediction layer. Then, the basic prediction value of cooling capacity is output using the basic prediction model of cooling capacity, and the basic prediction value of energy consumption is output using the basic prediction model of energy consumption. Then, the correction value of cooling capacity is output using the correction model of cooling capacity, and the correction value of energy consumption is output using the correction model of energy consumption. Thus, the predicted value of cooling capacity = the basic prediction value of cooling capacity + the correction value of cooling capacity, and the predicted value of energy consumption = the basic prediction value of energy consumption + the correction value of energy consumption.

[0317] In other embodiments, after obtaining the cooling capacity-energy consumption prediction model, for data in the steady state period, the output of the basic prediction model of the cooling capacity-energy consumption prediction model can be used as the corresponding cooling capacity prediction value and energy consumption prediction value; for data in the transition state period, the sum of the output of the basic prediction model and the output of the dynamic correction model can be used as the corresponding cooling capacity prediction value and energy consumption prediction value.

[0318] The cells of an energy storage system can store and utilize electrical energy through cyclic charging and discharging. As the application scenarios of energy storage systems become increasingly complex, existing energy storage systems are usually equipped with energy storage thermal management systems. However, existing energy storage thermal management systems can only perform thermal management based on preset fixed modes, and their effectiveness needs to be improved.

[0319] This application further proposes a control method for an energy storage thermal management system, such as... Figure 20 As shown, the cleaning method includes steps B11 to B14.

[0320] Step B11: Obtain the temperature threshold corresponding to the current state of the battery cell from the dynamic temperature threshold library.

[0321] The dynamic temperature threshold library stores multiple temperature thresholds, with different temperature thresholds corresponding to different cell states. In some embodiments, the dynamic temperature threshold library includes preset expression relationships; in other embodiments, the dynamic temperature threshold library includes a data table, and the specific implementation is not limited.

[0322] In some embodiments, the temperature threshold includes at least one of a maximum temperature threshold and a minimum temperature threshold.

[0323] For example, in some embodiments, the temperature threshold includes a maximum temperature threshold, which is the upper limit of the cell's temperature during normal operation. The maximum temperature threshold differs depending on the cell's state, and the dynamic temperature threshold library includes multiple different maximum temperature thresholds. In some embodiments, the maximum temperature threshold may also include a maximum charging temperature threshold and a maximum discharging temperature threshold to differentiate temperature control strategies during charging and discharging processes, improving temperature control accuracy and thus enhancing the cell's safety. Specifically, the maximum charging temperature threshold is the upper limit of the cell's temperature during charging, and the maximum discharging temperature threshold is the upper limit of the cell's temperature during discharging.

[0324] For example, in some embodiments, the temperature threshold includes a minimum temperature threshold, which is the lower limit of the battery cell's temperature during normal operation. The minimum temperature threshold varies depending on the battery cell's state, and the dynamic temperature threshold library includes multiple different minimum temperature thresholds. In some embodiments, the minimum temperature threshold can be further subdivided into a minimum charging temperature threshold and a minimum discharging temperature threshold to differentiate temperature control strategies during charging and discharging processes, improve temperature control accuracy, and thus enhance the battery cell's safety. Specifically, the minimum charging temperature threshold is the lower limit of the battery cell's temperature during charging, and the minimum discharging temperature threshold is the lower limit of the battery cell's temperature during discharging.

[0325] For example, in some embodiments, the temperature threshold includes a minimum temperature threshold and a maximum temperature threshold, which can define the temperature range of the battery cell when it is working normally; the safe operating temperature range of the battery cell is different when it is in different states, and the dynamic temperature threshold library includes multiple different minimum temperature thresholds and multiple different maximum temperature thresholds.

[0326] Selecting the corresponding temperature threshold from the dynamic temperature threshold library based on the current state of the battery cell enables temperature control to better match the actual working needs of the battery cell, improves the safety of battery cell operation, and optimizes the thermal management effect of the energy storage thermal management system.

[0327] In one application scenario, if the current temperature of the battery cell exceeds the maximum charging temperature threshold, charging can be disabled or the charging rate limited, and the thermal management unit can be controlled to start cooling. In another application scenario, if the predicted temperature of the battery cell exceeds the maximum charging temperature threshold, the required cooling capacity can be obtained based on the predicted temperature and the temperature threshold. Based on the required cooling capacity, a target control strategy can be determined to control the thermal management unit to start cooling in advance to prevent the predicted temperature of the battery cell from exceeding the maximum charging temperature threshold in the future.

[0328] In one application scenario, if the current temperature or preset temperature of the battery cell exceeds the maximum discharge temperature threshold, the discharge can be prohibited or the discharge rate limited, or the thermal management unit can be controlled to start cooling, as described above. Further details are omitted.

[0329] In one application scenario, if the current temperature of the battery cell is lower than the minimum charging temperature threshold, charging can be disabled or the charging rate limited, and the thermal management unit can be controlled to start heating. In another application scenario, if the predicted temperature of the battery cell is lower than the minimum charging temperature threshold, the required heating capacity can be obtained based on the predicted temperature and the temperature threshold. Based on the required heating capacity, a target control strategy can be determined to control the thermal management unit to start heating in advance to prevent the predicted temperature of the battery cell from falling below the minimum charging temperature threshold in the future. The method for obtaining the required heating capacity and the corresponding target control strategy can be referred to the above embodiments, and will not be repeated here.

[0330] In one application scenario, if the current or preset temperature of the battery cell is lower than the minimum discharge temperature threshold, discharge can be prohibited or the discharge rate limited, or the thermal management unit can be controlled to start heating, as described above. Further details are omitted.

[0331] In some embodiments, the current state of the battery cell includes at least one of the current state of charge and the current state of health.

[0332] The heat generation, aging, and internal resistance characteristics of a battery cell are related to its current state of charge (SOC) and current health status. For example, the current SOC affects the rate of internal chemical reactions and heat generation characteristics, and the cell's sensitivity to temperature and tolerance vary under different SOCs. Specifically, the internal resistance and heat generation characteristics differ in the high SOC range (e.g., above 90%), low SOC range (e.g., below 20%), and medium SOC range (e.g., between 30% and 80%). In the medium SOC range, the ohmic and polarization resistances are relatively small, irreversible heat generation is relatively small, and the entropy change of the electrochemical reaction within the cell is small, resulting in minimal impact on reversible heat generation. In the low SOC range, both ohmic and polarization resistances increase significantly, irreversible heat generation rises significantly, and the entropy change of the electrochemical reaction within the cell is large, resulting in a significant impact on reversible heat generation. In the high charge state range, the polarization resistance of the battery cell increases significantly, irreversible heat generation increases, and side reactions such as electrolyte oxidation are enhanced.

[0333] Therefore, determining the temperature threshold of a battery cell based on its current state of charge and current health status helps to more accurately match the safe operating boundaries of the cell under different conditions. For example, when the current health status of a battery cell is low, its high-temperature resistance may decrease. In this case, dynamically adjusting the maximum temperature threshold of the cell can reduce further damage to the cell caused by high temperatures. Similarly, under a high state of charge, the risk of thermal runaway increases. In this case, dynamically adjusting the temperature threshold of the cell can improve its operational safety.

[0334] In some embodiments, a maximum temperature threshold is set based on the current health status of the battery cell.

[0335] The health status of a battery cell directly reflects its performance degradation. As the cell's usage time and charge-discharge cycle count increase, a series of physical and chemical changes occur within the cell, such as the loss of active materials, electrolyte decomposition, and diaphragm aging. These changes lead to a decrease in the cell's thermal stability. When a cell's health status is low, its high-temperature resistance weakens, and under the same temperature conditions, aging cells have a relatively higher risk of thermal runaway. Therefore, setting a maximum temperature threshold based on the cell's current health status allows for dynamic adjustment of the safe operating temperature limit according to the cell's actual aging degree. For example, a new cell with a health status close to 100% can be allowed to operate within a relatively high temperature range; as the cell ages, lowering the maximum temperature threshold effectively mitigates the risk of accelerated aging due to high temperatures, enabling more precise temperature control in response to the cell's actual operating state and improving operational safety.

[0336] In some embodiments, the dynamic temperature threshold library includes a first preset relationship, which satisfies expression 3-1:

[0337] in, The maximum temperature threshold, ζ represents the first basic temperature threshold of the battery cell, SOH represents the current health status of the battery cell, and ζ represents the temperature derating factor.

[0338] In this embodiment, the maximum temperature threshold is the upper limit of the temperature of the battery cell during safe operation, including the charging and discharging processes of the battery cell.

[0339] In some embodiments, the first base temperature threshold is the upper limit of the temperature at which the cell operates under standard conditions, such as when the cell is at 100% SOH. The first base temperature threshold can be determined based on the type and material of the cell.

[0340] In some embodiments, the temperature derating factor ζ ≥ 0. For example, when ζ is 0.1℃ / % and the current state of health (SOH) of the cell is 80%, the maximum temperature threshold Tmax = Tbase - 0.1 × (100% - 80%) = Tbase - 2℃, meaning the maximum temperature threshold is reduced by 2℃, which better meets the thermal safety requirements of an aging cell with an SOH of 80%. This method can dynamically adjust the maximum temperature threshold according to the actual health state of the cell, reducing the temperature control risk caused by sudden changes in health state and improving the operational stability of the energy storage thermal management system. In other embodiments, the first preset relationship can also be non-linear, for example, using an exponential function or a piecewise function, to more accurately match the temperature tolerance characteristics of the cell under different health state ranges. For example, when SOH is higher than 80%, the temperature derating factor is smaller; when SOH is lower than 80%, the temperature derating factor increases; this can provide stricter temperature limits when the cell's health state is severely degraded, further improving the cell's operational safety.

[0341] In some embodiments, a minimum charging temperature threshold is set based on the cell's current state of charge (SOC). In one application scenario, at low temperatures, the rate of lithium-ion diffusion and electrochemical reaction within the cell decreases significantly. If charging is performed under these conditions, lithium-ion deposition is highly likely to occur at the negative electrode. If the cell is at a higher SOC, this deposition will be further exacerbated. To improve the cell's safety, a minimum charging temperature threshold can be set based on the cell's current SOC. For example, when the SOC exceeds a certain value, the minimum charging temperature threshold increases with the SOC. For instance, at a current SOC of 90%, the minimum charging temperature threshold is increased by 5°C or even higher to suppress lithium plating side reactions, thereby improving battery operational safety.

[0342] In some embodiments, a minimum charging temperature threshold is set based on the current health state of the battery cell. For example, when the battery cell's health state is low, its low-temperature performance may further deteriorate. In this case, the minimum charging temperature threshold can be appropriately increased to reduce the damage to aging battery cells caused by low-temperature charging. In some embodiments, the minimum charging temperature threshold is increased accordingly as the current health state of the battery cell deteriorates.

[0343] In some embodiments, the dynamic temperature threshold library can also comprehensively consider the cell's current state of charge and current health state to set the minimum charging temperature threshold. For example, by establishing a multivariable functional relationship between the two and the minimum charging temperature threshold, the minimum charging temperature threshold can more comprehensively reflect the cell's low-temperature charging safety requirements under complex conditions.

[0344] In some embodiments, the dynamic temperature threshold library includes a second preset relationship, which satisfies expression 3-2:

[0345] in, The minimum charging temperature threshold, This is the second basic temperature threshold for the battery cell. SOH represents the current health state of the battery cell, and SOC represents the current state of charge of the battery cell. The reference state of charge of the battery cell is a1, which is the first preset coefficient, and a2 is the second preset coefficient.

[0346] In some embodiments, the second base temperature threshold This is the lower limit of the temperature range when the battery cell is charged under standard conditions, for example, when the battery cell is at 100% SOH and 50% SOC. When set to 80%.

[0347] In some embodiments, a1≥0, a2≥0. When a1≥0, the minimum charging temperature threshold is reached when the current state of health (SOH) of the battery cell decreases. This will increase accordingly. Cells with deteriorated health are less tolerant of low temperatures; increasing the minimum charging temperature threshold can reduce damage to the cells from low-temperature charging. a2≥0, when the cell's current state of charge (SOC) is higher than the reference SOC. At that time, minimum charging temperature threshold The minimum charging temperature threshold increases with the percentage exceeding the limit, further ensuring charging safety and suppressing lithium plating risk at high SOC levels. This setting dynamically and precisely adjusts the minimum charging temperature threshold based on the cell's current health and state of charge, effectively improving charging safety and lifespan under different conditions.

[0348] In other embodiments, the second preset relationship may also be in the form of a piecewise function, etc., to more accurately match the temperature tolerance characteristics of the battery cell under different states of charge or different health states.

[0349] In some embodiments, a minimum discharge temperature threshold is set based on the current health status of the battery cell.

[0350] A decline in the health of a battery cell, i.e., the cell exhibiting signs of aging, increases its ohmic and polarization resistance. In low-temperature environments, the cell's internal resistance also increases. When an aging cell discharges at low temperatures, this cumulative effect can cause a significant voltage drop at the start of discharge, potentially triggering undervoltage protection instantaneously. Therefore, setting a higher minimum discharge temperature threshold for aging cells can improve their safety. For example, if the minimum discharge temperature threshold for a cell with a state of harmlessness (SOH) of 100% is -20℃, then setting it to -10℃ for a cell with an SOH of 80% would be preferable.

[0351] In some embodiments, the dynamic temperature threshold library includes a third preset relationship, which satisfies expression 3-3:

[0352] in, The minimum discharge temperature threshold. This is the third basic temperature threshold for the battery cell, and SOH represents the current health status of the battery cell. This is the third preset coefficient.

[0353] In some embodiments, the third base temperature threshold This is the lower limit of the temperature at which the battery cell discharges under standard conditions. For example, when the battery cell is at 100% SOH.

[0354] In some embodiments, a3 ≥ 0. The minimum discharge temperature threshold is set when the current state of health (SOH) of the battery cell decreases. Increase. For example, if a3 is 0.1℃ / %, when SOH decreases from 100% to 80%, the minimum discharge temperature threshold is... +2℃, meaning the minimum discharge temperature threshold is increased by 2℃.

[0355] In other embodiments, the third preset relationship may also take other functional forms, such as piecewise functions, depending on the specific type and aging characteristics of the battery cell.

[0356] Step B12: Predict the required cooling capacity of the battery cell based on the cell temperature prediction model and temperature threshold.

[0357] In some embodiments, the cell temperature prediction model can be obtained by referring to the construction method of the cell temperature prediction model in the above embodiments, which will not be repeated here.

[0358] In some embodiments, the predicted temperature of the battery cell at future times is obtained based on a battery cell temperature prediction model, and the required cooling capacity of the battery cell at future times is determined based on a temperature threshold corresponding to the current state of the battery cell and the predicted temperature.

[0359] For example, the difference between the temperature threshold and the predicted temperature is calculated, and the required cooling capacity of the battery cell at future times is calculated based on this difference. For instance, in one application scenario, the temperature threshold includes a maximum temperature threshold and a minimum temperature threshold. It is determined whether the predicted temperature exceeds the temperature range between the minimum and maximum temperature thresholds. If the predicted temperature is higher than the maximum temperature threshold, the required cooling capacity is calculated based on the difference between (predicted temperature and maximum temperature threshold). If the predicted temperature is lower than the minimum temperature threshold, the required heating capacity is calculated based on the difference between (minimum temperature threshold and predicted temperature), or the required cooling capacity is less than 0 based on the difference between (predicted temperature and minimum temperature threshold), and the required heating capacity is determined based on the absolute value of the required cooling capacity.

[0360] In one application scenario, when the predicted temperature exceeds the temperature threshold, the thermal management unit activates the corresponding control strategy in advance to provide the corresponding cooling capacity and keep the cell temperature within a safe range. When the predicted temperature is below the minimum temperature threshold and the required cooling capacity is less than 0, the thermal management unit starts the heating program; when the predicted temperature is above the maximum temperature threshold and the required cooling capacity is greater than 0, the thermal management unit starts the cooling program.

[0361] Step B13: Obtain the cooling capacity-energy consumption prediction model and current input parameters of the thermal management unit.

[0362] In some embodiments, the cooling capacity-energy consumption prediction model of the thermal management unit can be obtained by referring to the setting method of the cooling capacity-energy consumption prediction model in the above embodiments, for example, the cooling capacity-energy consumption prediction model can be obtained based on historical operating data, which will not be elaborated here.

[0363] Current input parameters refer to the input parameters of the thermal management unit at the current moment or under the current operating conditions, such as current environmental and operating parameters, including current ambient temperature, supply liquid temperature, return liquid temperature, compressor speed, water pump flow rate, valve opening, and fan speed. For details, please refer to the above embodiments; further elaboration is not provided here.

[0364] Step B14: With the goal of minimizing energy consumption, a target control strategy is adopted based on the demand cooling capacity, the cooling capacity-energy consumption prediction model, and the current input parameters to obtain the adjustable input parameters of the thermal management unit, so as to control the operation of the thermal management unit.

[0365] The specific implementation of step B14 can be referred to step A14 above, and will not be repeated here.

[0366] The above method can obtain the corresponding temperature threshold based on the current state of the battery cell, and then obtain the corresponding target control strategy. This enables the energy storage thermal management system to better adapt to the actual state of the battery cell when performing thermal management, improve the thermal management effect, and thus improve the operational safety of the battery cell.

[0367] Furthermore, the above-mentioned setup can improve thermal management performance, effectively controlling the temperature difference between different areas of the battery cells within the energy storage system, as well as the temperature difference between different battery cells. In one application scenario, this setup can keep the difference between the highest and lowest internal temperatures of the container housing the battery cells below 3°C, effectively reducing the aging rate of the battery cells and extending their lifespan. In another application scenario, this setup can effectively reduce the average annual energy consumption of the thermal management unit by 10%-20%.

[0368] In other embodiments, the temperature threshold can also be set to vary with the service life of the battery cell. For example, the maximum temperature threshold can be set to decrease as the service life of the battery cell increases. In one application scenario, the maximum charging temperature threshold and the maximum discharging temperature threshold of the battery cell are equal, and as the service life of the battery cell increases, the maximum charging temperature threshold and the maximum discharging temperature threshold are set to decrease by 0.5°C per year. For example, a brand new battery cell has a maximum charging temperature threshold and a maximum discharging temperature threshold of 35°C in the first year, and a maximum charging temperature threshold and a maximum discharging temperature threshold of 34.5°C in the second year.

[0369] In some embodiments, this application further proposes a control method for an energy storage thermal management system, such as... Figure 21 As shown, it includes steps B21 to B27.

[0370] Step B21: Obtain the historical charge-discharge cycle count of the battery cell.

[0371] The specific implementation of step B21 can be found in step S121, and will not be repeated here.

[0372] Step B22: Obtain the cumulative temperature difference value of the historical over-temperature events of the battery cell.

[0373] The specific implementation of step B22 can be found in step S122, and will not be repeated here.

[0374] Step B23: Determine the current health status of the battery cell based on the cumulative temperature difference and the number of historical charge-discharge cycles.

[0375] The specific implementation of step B23 can be found in step S123, and will not be repeated here.

[0376] Step B24: Obtain the temperature threshold corresponding to the current state of the battery cell from the dynamic temperature threshold library.

[0377] The specific implementation of step B24 can be referred to step B11, and will not be repeated here.

[0378] Step B25: Predict the required cooling capacity of the battery cell based on the cell temperature prediction model and temperature threshold.

[0379] The specific implementation method of step B25 can be referred to step B12, and will not be repeated here.

[0380] Step B26: Obtain the cooling capacity-energy consumption prediction model and current input parameters of the thermal management unit.

[0381] The specific implementation method of step B26 can be referred to step B13, and will not be repeated here.

[0382] Step B27: With the goal of minimizing energy consumption, a target control strategy is adopted based on the demand cooling capacity, the cooling capacity-energy consumption prediction model, and the current input parameters to obtain the adjustable input parameters of the thermal management unit, so as to control the operation of the thermal management unit.

[0383] The specific implementation method of step B27 can be referred to step B14, and will not be repeated here.

[0384] In some embodiments, the current input parameters include at least the operating parameters of the thermal management unit and the ambient temperature.

[0385] Among them, environmental parameters are external conditions during the operation of the thermal management unit, which have a significant impact on the heat load and heat dissipation efficiency of the energy storage system. The selection of environmental parameters and their working principles can be referred to the above embodiments, and will not be repeated here.

[0386] Operating parameters are internal system parameters of the thermal management unit during operation. For example, in some embodiments, operating parameters include at least one of the following: supply liquid temperature, return liquid temperature, compressor speed, water pump flow rate, valve opening, fan speed, high pressure on the high-pressure side of the refrigerant circulation loop, low pressure on the low-pressure side of the refrigerant circulation loop, evaporation temperature, and condensation temperature. The selection and operating principle of operating parameters can be referred to the above embodiments, and will not be repeated here.

[0387] In some embodiments, the adjustable input parameters include at least one of the following: liquid supply temperature, compressor speed, water pump flow rate, valve opening degree, and fan speed.

[0388] In some embodiments, the cooling capacity-energy consumption prediction model has adjustable input parameters, which are typically controllable variables among the operating parameters of the thermal management unit, and the current input parameters include the adjustable input parameters. For example, in some embodiments, the adjustable input parameters include liquid supply temperature, compressor speed, water pump flow rate, valve opening degree, and fan speed. These operating parameters can quickly change the cooling capacity and energy consumption of the thermal management unit. A control strategy that obtains the adjustable input parameters helps to control the operating state of the thermal management unit based on the control strategy. The control strategy is composed of the specific values ​​of the adjustable input parameters. When controlling the thermal management unit to operate, adjusting the adjustable input parameters to the values ​​determined by the control strategy can control the thermal management unit to achieve specific cooling capacity and energy consumption targets. The selection, setting, and operating principle of the adjustable input parameters can be referred to the above embodiments, and will not be repeated here.

[0389] In some embodiments, a negative temperature coefficient (NTC) temperature sensor is mounted on the surface of the battery cell, and a pressure sensor is installed in the thermal management unit. A controller is used to implement any of the control methods, charge / discharge prediction methods, and battery cell temperature prediction methods described above. In some embodiments, current input parameters are obtained based on various sensors of the thermal management unit, and historical operating data is stored in a memory. In some embodiments, the EMS system of the thermal management unit includes a controller that uses the above-described charge / discharge prediction methods to predict future charge / discharge events.

[0390] This application further proposes an energy storage thermal management system, which includes a memory and a processor. The memory is used to store program data. In some embodiments, the program data can be executed by the processor to implement at least one of the charge / discharge prediction method, the cell temperature prediction method, and the control method of any of the above embodiments.

[0391] The specific implementation methods for the charge / discharge prediction method and the cell temperature prediction method can be referred to the above embodiments, and will not be repeated here.

[0392] This application further proposes a computer storage medium. For example... Figure 23 As shown, Figure 23 This is a schematic diagram of a computer storage medium according to an embodiment of the present application. The computer storage medium 10 stores program instructions 11, which are executed by a processor to implement at least one of the charge / discharge prediction method, the cell temperature prediction method, and the control method of any of the above embodiments.

[0393] Specifically, program instructions 11 can form a program file and be stored in the aforementioned storage medium as a software product, so that an electronic device (which may be a personal computer, server, or network device, etc.) or processor can execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, or terminal devices such as computers, servers, mobile phones, and tablets.

[0394] In this embodiment, the computer storage medium 10 can be, but is not limited to, a USB flash drive, SD card, PD optical drive, portable hard drive, large-capacity floppy drive, flash memory, multimedia memory card, server, etc.

[0395] This application further proposes a computer program product or computer program, which includes computer instructions that enable a computer to implement at least one of the charge / discharge prediction method, the cell temperature prediction method, and the control method of any of the above embodiments.

[0396] In some embodiments, the computer instructions are stored in a computer storage medium. In some embodiments, the processor of the electronic device reads the computer instructions from the computer storage medium, executes the computer instructions, and causes the electronic device to perform the steps in the above-described method embodiments.

[0397] Furthermore, if the aforementioned functions are implemented as software functions and sold or used as independent products, they can be stored in a mobile terminal-readable storage medium. That is, this application also provides a storage device storing program data, which can be executed to implement the methods of the above embodiments. This storage device can be, for example, a USB flash drive, an optical disc, or a server. In other words, this application can be embodied in the form of a software product, which includes several instructions to cause a smart terminal to execute all or part of the steps of any embodiment of the method in this application.

[0398] Unlike existing technologies, the charge / discharge prediction method of this application includes: acquiring the current data of the battery cell; acquiring a historical charge / discharge event database of the battery cell; and predicting the predicted charge / discharge power curve of the battery cell based on the current data and the historical charge / discharge event database. This method can utilize the current data of the battery cell and the historical charge / discharge event database to predict the predicted charge / discharge power curve of the battery cell. The predicted charge / discharge power curve can characterize the operating state of the battery cell in the future. This helps various management systems in the energy storage system to proactively initiate corresponding control strategies based on the predicted charge / discharge power curve before the operating state of the battery cell changes, thereby improving the response speed and control accuracy of the management system and enhancing the safety of the energy storage system.

[0399] The control method of the energy storage thermal management system of this application includes: acquiring historical operating data and current input parameters of the thermal management unit; obtaining a cooling capacity-energy consumption prediction model based on the historical operating data; predicting the required cooling capacity of the battery cells based on the battery cell temperature prediction model; and obtaining a target control strategy for the adjustable input parameters of the thermal management unit based on the cooling capacity-energy consumption prediction model, the required cooling capacity, and the current input parameters, with the minimum energy consumption as the optimization objective, to control the operation of the thermal management unit; wherein, the historical operating data includes historical input parameters and corresponding cooling capacity and energy consumption. The above method can obtain a cooling capacity-energy consumption prediction model by acquiring historical operating data of the thermal management unit, and then obtain a target control strategy, making the target control strategy more closely aligned with actual operating conditions, enabling the energy storage thermal management system to better adapt to actual operating conditions and improve thermal management effectiveness; furthermore, predicting the required cooling capacity of the battery cells based on the battery cell prediction model and determining the target control strategy based on the required cooling capacity can improve the response speed of the energy storage thermal management system, thereby improving thermal management effectiveness; and obtaining a target control strategy with the minimum energy consumption as the optimization objective can effectively reduce the energy consumption of the energy storage thermal management system.

[0400] The control method of this application includes obtaining the temperature threshold corresponding to the current state of the battery cell from a dynamic temperature threshold library; predicting the required cooling capacity of the battery cell based on the battery cell temperature prediction model and the temperature threshold; obtaining the cooling capacity-energy consumption prediction model and current input parameters of the thermal management unit; and obtaining a target control strategy for the adjustable input parameters of the thermal management unit based on the required cooling capacity, the cooling capacity-energy consumption prediction model, and the current input parameters, with the minimum energy consumption as the optimization objective, to control the operation of the thermal management unit. This method can obtain the corresponding temperature threshold based on the current state of the battery cell, and thus obtain the corresponding target control strategy. This enables the energy storage thermal management system to better adapt to the actual state of the battery cell during thermal management, thereby improving the thermal management effect.

[0401] In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. For example, the apparatus implementations described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0402] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, the functional units in the various embodiments of this application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units.

[0403] Any process or method description in the flowchart or otherwise herein can be understood as representing an apparatus, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order according to the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0404] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (which may be a personal computer, server, network device, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0405] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A control method for an energy storage thermal management system, characterized in that, include: Obtain historical operating data and current input parameters of the thermal management unit; A cooling capacity-energy consumption prediction model is obtained based on the historical operating data. Predict the required cooling capacity of battery cells based on a cell temperature prediction model. With the goal of minimizing energy consumption, a target control strategy is adopted to obtain the adjustable input parameters of the thermal management unit based on the cooling capacity-energy consumption prediction model, the required cooling capacity, and the current input parameters, so as to control the operation of the thermal management unit. The historical operating data includes historical input parameters and corresponding cooling capacity and energy consumption.

2. The control method according to claim 1, characterized in that, The historical input parameters include at least historical environmental parameters and historical operating parameters.

3. The control method according to claim 2, characterized in that, The historical environmental parameters include at least the historical environmental temperature; And / or, the historical operating parameters include at least one of the following: supply liquid temperature, return liquid temperature, compressor speed, water pump flow rate, valve opening, fan speed, high pressure value on the high-pressure side of the refrigerant circulation loop, low pressure value on the low-pressure side of the refrigerant circulation loop, evaporation temperature, and condensation temperature.

4. The control method according to claim 1, characterized in that, The cooling capacity-energy consumption prediction model includes a cooling capacity prediction sub-model and an energy consumption prediction sub-model, which are obtained by training separately.

5. The control method according to claim 4, characterized in that, The cooling capacity prediction sub-model includes a cooling capacity prediction function, and / or the energy consumption prediction sub-model includes an energy consumption prediction function or an energy efficiency ratio prediction function.

6. The control method according to claim 1, characterized in that, The thermal management unit is equipped with operating conditions and control modes; the historical operating data also includes the operating conditions and corresponding control modes corresponding to the historical input parameters. The cooling capacity-energy consumption prediction model is configured to: receive the input parameters to be tested, the operating condition to be tested, and the control mode to be tested, and output the predicted cooling capacity and energy consumption values ​​corresponding to the operating condition to be tested and the control mode to be tested.

7. The control method according to claim 6, characterized in that, The operating conditions include at least one of environmental operating conditions and load operating conditions, and / or the control modes include at least one of standby mode, self-circulation mode, cooling mode, and heating mode.

8. The control method according to claim 6, characterized in that, The steps for obtaining the cooling capacity-energy consumption prediction model based on the historical operating data include: The historical operating data is divided into multiple sample subsets according to the control mode described above; Based on the sample subset, train the mode sub-models corresponding to the control modes respectively; The cooling capacity-energy consumption prediction model is configured to: receive the input parameters to be tested and the control mode to be tested, route to the mode sub-model corresponding to the control mode to be tested based on the control mode to be tested, and output the corresponding cooling capacity prediction value and energy consumption prediction value.

9. The control method according to claim 1, characterized in that, The historical operating data includes raw operating data and updated operating data. The step of obtaining a cooling capacity-energy consumption prediction model based on the historical operating data includes: A cooling capacity-energy consumption prediction model was obtained based on the original operating data; The historical running data within the sliding window is used as the updated running data; Based on the updated operating data, the parameters of the cooling capacity-energy consumption prediction model are adjusted through incremental learning to update the cooling capacity-energy consumption prediction model.

10. An energy storage thermal management system, characterized in that, It includes a memory and a processor, the memory being used to store program data, the program data being executable by the processor to implement the control method according to any one of claims 1 to 9.

11. A computer storage medium, characterized in that, It stores program instructions that are executed by a processor to implement the control method according to any one of claims 1 to 9.

12. A computer program product, characterized in that, It includes computer program instructions that cause a computer to implement the control method according to any one of claims 1 to 9.

Images