Battery connection state detection method, device and equipment

Abstract

The invention provides a battery connection state detection method, device and equipment, and relates to the technical field of electrochemical energy storage. The method comprises the following steps: acquiring various measurement data at a battery joint; wherein the various measurement data comprises voltage data, temperature data and connection impedance data; inputting the voltage data, the temperature data and the connection impedance data into a trained connection state classification model to obtain a connection state classification result of the battery connection part; wherein the connection state classification model is obtained by training based on historical voltage data, historical temperature data and historical connection impedance data of a plurality of battery joints and corresponding historical connection state classification results; based on the connection state classification result, evaluating the connection state of the battery joint to obtain a connection state evaluation result; and according to a connection state evaluation result, performing graded early warning so as to perform corresponding operation and maintenance. According to the invention, accurate detection of the connection state of the battery is realized, and the maintenance timeliness is improved through graded early warning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electrochemical energy storage, and in particular to a battery connection state detection method, device and equipment. BACKGROUND

[0002] With the continuous expansion of the scale of electrochemical energy storage power stations, the connection reliability of the battery system has become increasingly prominent and has become a key factor affecting the safe operation of the power station. Therefore, the connection reliability of the battery system needs to be detected to ensure the safe operation of the power station.

[0003] The traditional detection method has many technical limitations when dealing with the complex working conditions of large-scale energy storage systems. For example, the traditional detection method relies too much on monitoring a single parameter. Although a single parameter such as voltage can reflect some electrical faults, it has serious shortcomings in early identification of progressive defects such as mechanical connection loosening, resulting in poor accuracy and serious detection lag. For example, in actual operation, the loosening of the connection part often needs to develop to a significant reduction in contact area and a significant increase in contact resistance before it causes observable voltage changes. Therefore, the defects of the traditional detection method may cause the battery system to miss the best maintenance opportunity, leading to the continuous accumulation of fault hazards.

[0004] In view of the above-mentioned defects, the traditional detection method cannot accurately detect the battery connection state and cannot comprehensively diagnose the connection reliability of the battery. Moreover, even if the problem of the battery connection state is detected, the traditional technology often directly prompts. Since there are a large number of batteries, once multiple prompts occur, the relevant personnel are not clear about which one to handle first, which may lead to serious problems that are delayed in being solved, causing more serious problems, and is not conducive to the sustainable development of the system. SUMMARY

[0005] The embodiments of the present disclosure provide a battery connection state detection method, device and equipment to solve the problem that the battery connection state cannot be accurately detected in the related art and the lack of processing priority for the problem of the battery connection state leads to more serious problems.

[0006] In a first aspect, the embodiments of the present disclosure provide a battery connection state detection method, comprising:

[0007] obtaining a plurality of measurement data at the battery connection; wherein the plurality of measurement data comprises voltage data, temperature data and connection impedance data;

[0008] inputting the voltage data, the temperature data, and the connection impedance data into a trained connection state classification model to obtain a connection state classification result of the battery connection; the connection state classification model is trained based on historical voltage data, historical temperature data, historical connection impedance data, and corresponding historical connection state classification results of a plurality of battery connections;

[0009] based on the connection state classification result, evaluating the connection state of the battery connection to obtain a connection state evaluation result;

[0010] based on the connection state evaluation result, performing a hierarchical early warning to perform corresponding operation and maintenance.

[0011] In some possible embodiments, the evaluating the connection state of the battery connection based on the connection state classification result to obtain a connection state evaluation result comprises:

[0012] based on the connection state classification result, determining a value of a weight coefficient in a preset connection state evaluation relationship; the connection state evaluation relationship is constructed based on a plurality of weight coefficients, a plurality of measurement data of the battery connection, and corresponding normal data, and the weight coefficient represents a proportion of a difference between the measurement data and the normal data in each connection state classification result;

[0013] based on the value of the weight coefficient and the connection state evaluation relationship, obtaining the connection state evaluation result.

[0014] In some possible embodiments, the determining the value of the weight coefficient in the preset connection state evaluation relationship based on the connection state classification result comprises:

[0015] based on the connection state classification result and a prestored corresponding relationship between a connection state classification result and an adjustment weight coefficient, determining a weight coefficient to be adjusted;

[0016] determining an adjustment size of the value of the weight coefficient to be adjusted according to a data feature; the data feature at least includes a trend and / or a degree of deviation of the voltage data, the temperature data, and the connection impedance data from corresponding normal data;

[0017] based on the weight coefficient to be adjusted and the adjustment size, determining the value of the weight coefficient in the connection state evaluation relationship.

[0018] In some possible embodiments, the obtaining the connection impedance data of the battery connection comprises:

[0019] obtaining an M-sequence pseudo-random signal;

[0020] The M-sequence pseudo-random signal is used to sweep the battery connection in a preset frequency range, and different frequency points and corresponding impedance data are obtained;

[0021] The impedance spectrum in the preset frequency range is reconstructed based on the different frequency points and corresponding impedance data, and the connection impedance data is determined based on the impedance spectrum.

[0022] In some possible embodiments, the connection state classification model includes a short-term memory network model and a random forest model.

[0023] The connection state classification result of the battery connection is obtained by inputting the voltage data, temperature data, and connection impedance data into the pre-trained connection state classification model, including:

[0024] The data features of the voltage data, temperature data, and connection impedance data are extracted by inputting the voltage data, temperature data, and connection impedance data into the short-term memory network model; the data features at least include the trend and / or degree of deviation of the voltage data, temperature data, and connection impedance data from corresponding normal data.

[0025] The connection state classification result of the battery connection is obtained by inputting the data features into the random forest model; the connection state classification result includes a normal state and an abnormal state, and the abnormal state includes multiple abnormal types.

[0026] In some possible embodiments, the connection state evaluation result is obtained based on the value of the weight coefficient and the connection state evaluation relationship, including:

[0027] The corresponding standard deviation is obtained based on the voltage data, temperature data, and connection impedance data and corresponding normal data.

[0028] The connection state evaluation result is obtained according to the value of the weight coefficient, the standard deviation, and the connection state evaluation relationship.

[0029] In some possible embodiments, the connection state evaluation relationship satisfies:

[0030]

[0031] wherein S represents the connection state evaluation result, v1, v2, and v3 respectively represent the weight coefficients corresponding to the voltage data, the temperature data, and the connection impedance data; V ref , T ref , and Z ref(ω) respectively represent normal data corresponding to the voltage data, the temperature data and the connection impedance data, ω represents an angular frequency; ΔV, ΔT, ΔZ(ω) respectively represent standard deviations of the voltage data, the temperature data and the connection impedance data and corresponding normal data.

[0032] In a second aspect, the embodiments of the present disclosure provide a battery connection state detection apparatus, comprising:

[0033] a obtaining module configured to obtain a plurality of measurement data of a battery connection; wherein the plurality of measurement data comprises voltage data, temperature data and connection impedance data;

[0034] a classification module configured to input the voltage data, the temperature data and the connection impedance data into a trained connection state classification model to obtain a connection state classification result of the battery connection; wherein the connection state classification model is trained based on historical voltage data, historical temperature data, historical connection impedance data and corresponding historical connection state classification results of a plurality of battery connections;

[0035] an evaluation module configured to evaluate a connection state of the battery connection based on the connection state classification result to obtain a connection state evaluation result;

[0036] a pre-warning module configured to perform hierarchical pre-warning according to the connection state evaluation result to perform corresponding operation and maintenance.

[0037] In a third aspect, the embodiments of the present disclosure provide an electronic device, comprising a memory and a processor, the memory stores a computer program, and the processor implements the method in the first aspect or any possible implementation manner of the first aspect when executing the computer program.

[0038] The battery connection status detection method, apparatus, and device provided in this disclosure acquire various measurement data at the battery connection points, including voltage data, temperature data, and connection impedance data. These data are then input into a trained connection status classification model to obtain a connection status classification result for the battery connection points. The connection status classification model is trained based on historical voltage data, historical temperature data, historical connection impedance data, and corresponding historical connection status classification results from multiple battery connection points. This means that the connection status classification model trained on historical data can learn the relationship between multiple parameters and connection status, reducing the misjudgments or omissions caused by traditional methods that detect battery connection status using only a single parameter. Furthermore, this embodiment further evaluates the connection status of the battery connection points based on the connection status classification results, obtaining a connection status evaluation result. Based on the connection status evaluation result, a tiered warning system is implemented for corresponding maintenance and repair, thus solving the problem of inconsistent priority in multi-battery alerts in traditional technologies, which leads to chaotic maintenance. Precise warnings allow for intervention at the optimal maintenance time, reducing maintenance costs, minimizing unnecessary downtime, and improving system operating efficiency. Attached Figure Description

[0039] Figure 1 This is a schematic flowchart of a battery connection status detection method provided in an embodiment of this disclosure;

[0040] Figure 2 This is a flowchart illustrating another battery connection status detection method provided in this embodiment of the disclosure;

[0041] Figure 3 This is a schematic diagram of the structure of a battery connection status detection device provided in an embodiment of this disclosure;

[0042] Figure 4 This is a schematic diagram of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0043] In response to the technical problems described in the background section, the applicant has discovered through research that traditional testing methods have the following shortcomings:

[0044] First, traditional detection methods typically rely on single parameters such as voltage and temperature, making it difficult to effectively identify loose mechanical connections in their early stages. Furthermore, they suffer from response delays, often triggering alarms only when the fault has progressed to a more severe stage. Even when other parameters are added during battery connection status identification, these parameters are often interpreted in isolation, and this fragmented monitoring approach easily leads to false positives and false negatives.

[0045] For example, for monitoring voltage only, voltage abnormalities can reflect some electrical faults, but have little ability to identify early-stage progressive defects such as loose mechanical connections. The loosening of the connection site often needs to develop to a significant reduction in contact area and a significant increase in contact resistance before it causes observable voltage changes. This monitoring lag can cause the system to miss the best maintenance opportunity, leading to the accumulation of fault hazards. For monitoring temperature only, there is also a significant response lag problem, that is, when the connection point temperature is abnormally high, it usually indicates that the site has experienced a long period of overheating, which can have caused irreversible damage such as material degradation and insulation aging. More seriously, in a closed battery compartment environment, local overheating can have triggered a thermal runaway chain reaction, creating a major safety hazard. This lag makes temperature monitoring more suitable for post-alarm rather than preventive diagnostic means.

[0046] Second, the traditional detection method lacks a graded evaluation of the battery connection state, making it difficult to guide workers to perform precise operation and maintenance.

[0047] To this end, the embodiments of the present disclosure obtain multiple measurement data of the battery connection, classify the connection state, obtain the connection state classification result of the battery connection, and further fuse and analyze the connection state classification result to obtain the connection state evaluation result. Thus, multi-dimensional measurement data is analyzed in coordination to perform early and accurate diagnosis of the battery connection state, improve detection sensitivity and real-time performance, reduce false positives and false negatives, and ensure safe operation of the system. Moreover, the embodiments of the present disclosure establish a graded early warning mechanism based on the connection state evaluation result and give corresponding processing suggestions to provide a scientific basis for operation and maintenance decisions, thereby helping to extend the life of the battery system, reduce unnecessary downtime, and promote the sustainable development of large-scale energy storage systems.

[0048] The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0049] Figure 1 is a flowchart of a battery connection state detection method provided by the embodiments of the present disclosure. As shown in Figure 1 the method includes the following steps:

[0050] S110, obtaining multiple measurement data of the battery connection; wherein the multiple measurement data includes voltage data, temperature data, and connection impedance data.

[0051] It can be seen that in actual detection, the deterioration process of the battery connection state will be reflected in multiple dimensions such as electricity, thermodynamics, and mechanics, for example, battery connection loosening, contact corrosion, etc. will cause changes in connection impedance, local overheating at the battery connection will cause temperature changes, and poor contact at the battery connection will cause voltage drop or increased ripple. Based on this, the embodiment can more comprehensively reflect the connection reliability of the battery connection by monitoring multiple measurement data at the battery connection.

[0052] The connection impedance data includes charge transfer resistance data (denoted by R ct ), double-layer capacitance data (denoted by C dl ), and series resistance data (denoted by R s ) for reflecting the connection resistance, which together constitute an electrochemical equivalent impedance (denoted by Z(ω)) related to the angular frequency. The electrochemical equivalent impedance Z(ω) satisfies the following formula:

[0053]

[0054] Where ω represents the angular frequency, and j represents the imaginary unit.

[0055] As can be seen from the above impedance formula, the impedance value at different frequencies is mainly composed of three parts, namely the charge transfer resistance data, the double-layer capacitance data, and the series resistance data. The impedance value at different frequencies can be extracted through Electrochemical Impedance Spectroscopy (EIS) analysis, in which R ct mainly appears in the medium frequency region of EIS, R s is the ohmic resistance component in EIS, and C dl is characterized by a semicircle in the medium-high frequency region of EIS. In practical applications, the three can reflect different fault characteristics, i.e., different abnormal types.

[0056] Specifically, abnormal changes in R ct can indirectly reflect the occurrence of loosening, corrosion, or poor contact of the connection state; abnormal changes in R s directly reflect the pure resistance characteristics of the wires, electrodes, and connecting parts of the battery system, which are highly sensitive to faults such as loosening, corrosion, and poor contact of the battery connection state, and are more direct diagnostic indicators than R ct ; C dl is the equivalent capacitance formed by the separation of interface charges of the electrode or electrolyte, which is extremely sensitive to changes in the microscopic morphology of the interface, such as corrosion, deposition, or shedding of active materials.

[0057] Exemplarily, when actually acquiring the plurality of measurement data, the embodiment can set a plurality of measurement nodes spaced apart from each other at the battery connection to be detected, such as setting an impedance measurement node at the crimping position of the battery monomer and the pole column, setting a temperature measurement node at the crimping position of the module connecting sheet and the pole column, and setting a voltage measurement node at the contact surface of the connecting head and the bus bar row, for measuring the voltage data, temperature data and connection impedance data of the battery connection, respectively. The battery connection can be specifically the position of electrical connection between two battery monomers, or the position of connection between battery modules or between battery clusters, and the specific number of the impedance measurement node, the temperature measurement node and the voltage measurement node and the type of the battery connection are not limited here.

[0058] S120, input the voltage data, the temperature data and the connection impedance data into the trained connection state classification model to obtain a connection state classification result of the battery connection.

[0059] The connection state classification result includes a normal state and an abnormal state, and the abnormal state includes a plurality of abnormal types. For example, when the voltage data, the temperature data and the connection impedance data are all normal, the connection state classification result is the normal state, and when the voltage data, the temperature data and the connection impedance data are abnormal respectively, different abnormal types can be represented, that is, different types of measurement data abnormality correspond to different abnormal types.

[0060] Exemplarily, Table 1 is a connection state classification table provided by the embodiment of the present disclosure, as shown in Table 1, different connection state classification results and the corresponding measurement data performance (referred to as data performance) are shown in the table.

[0061] Table 1 Connection state classification table

[0062] Connection state classification result Data performance Normal Voltage data, temperature data and connection impedance data are normal Loose connection [R s abnormalities (such as a significant rise)] Contact corrosion [R ct abnormalities (such as a significant rise)] Active material shedding [C dl abnormalities (such as a significant rise)] Overload heating Abnormal temperature data (e.g. significant rise) Abnormal charging and discharging Abnormal voltage fluctuation in voltage data (e.g. significant increase) Connection disconnection Zero voltage in connection bar in voltage data

[0063] In this way, based on whether the above-mentioned plurality of measurement data is abnormal or normal, the corresponding connection state classification result can be obtained.

[0064] The connection state classification model is trained based on historical voltage data, historical temperature data, historical connection impedance data of the plurality of battery connection positions, and corresponding historical connection state classification results. For example, for model training, a large amount of historical voltage data, a large amount of historical temperature data, a large amount of historical connection impedance data of the plurality of battery connection positions, and known connection state classification results (i.e., classification labels) corresponding to the above data are input into an initial connection state classification model. The connection state classification model optimizes and adjusts model parameters through a series of steps to reduce the difference between the predicted connection state classification results and the classification labels, improve the classification accuracy, and finally complete the model training and obtain the trained connection state classification model.

[0065] Specifically, in actual application, the voltage data, the temperature data, and the connection impedance data are input into the trained connection state classification model. The connection state classification model processes the above data in a series of steps and outputs the predicted connection state classification results. In this way, different abnormal types such as connection loosening, contact corrosion, and overload heating can be automatically distinguished by using a machine learning algorithm, and the related faults of the battery connection position can be accurately identified.

[0066] S130, based on the connection state classification result, evaluating the connection state of the battery connection position to obtain a connection state evaluation result.

[0067] The connection state evaluation result is used to evaluate the overall connection state of the battery connection position. For example, when at least one of the voltage data, the temperature data, and the connection impedance data is abnormal, the connection state classification result can correspond to one abnormal type or multiple abnormal types. In this case, the overall connection state of the battery connection position is evaluated based on the connection state classification result, for example, the health degree of the overall connection state of the current battery connection position is determined, and the corresponding connection state evaluation result is obtained. In short, the health degree of the overall connection state is different, and the connection state evaluation result is also different.

[0068] In this way, the disclosure embodiments obtain a plurality of measurement data of the battery connection position, and obtain the connection state classification result of the battery connection position corresponding to the plurality of measurement data. The connection state evaluation result is obtained based on the connection state classification result. The health degree of the overall battery connection position is determined based on the fusion analysis and evaluation of the plurality of measurement data. The detection accuracy of the battery connection state is improved, the misjudgment and omission problem is avoided, and the battery connection reliability can be comprehensively diagnosed.

[0069] S140, according to the connection state evaluation result, performing a hierarchical early warning to perform corresponding operation and maintenance.

[0070] Specifically, on the basis of obtaining the connection state classification result, the health degree of the overall connection state of the battery connection is determined, and further, hierarchical early warning is implemented, that is, after determining the early warning level corresponding to the current connection state evaluation result, a prompt is given to take corresponding operation and maintenance measures in time. In this way, when the positions of the monitored battery connections are numerous, once multiple prompts appear, the staff can clearly process the priority through the hierarchical early warning prompt, and then take corresponding operation and maintenance measures in time to ensure that problems existing in each battery connection are processed according to their severity, preventing the situation from deteriorating and further promoting the sustainable development of the system.

[0071] For example, when the health degree of the overall connection state of the battery connection is poor, it indicates that the severity is high, so it is classified as a serious early warning level, and a red light is used for prompting to facilitate the staff to take corresponding operation and maintenance, such as immediate shutdown; when the health degree of the overall connection state of the battery connection is good, it indicates that the severity is low, so it is classified as a normal early warning level, and a green light is used for prompting to facilitate the staff to take corresponding operation and maintenance, such as continuous monitoring of the status of the battery connection. The specific classification of hierarchical early warning is described in detail below.

[0072] The battery connection state detection method provided by the embodiments of the present disclosure includes: obtaining a plurality of measurement data of a battery connection; wherein the plurality of measurement data includes voltage data, temperature data and connection impedance data; inputting the voltage data, the temperature data and the connection impedance data into a trained connection state classification model to obtain a connection state classification result of the battery connection; wherein the connection state classification model is trained based on historical voltage data, historical temperature data, historical connection impedance data and corresponding historical connection state classification results of a plurality of battery connections; based on the connection state classification result, the connection state of the battery connection is evaluated to obtain a connection state evaluation result; and according to the connection state evaluation result, hierarchical early warning is performed to take corresponding operation and maintenance. Thus, early and accurate diagnosis of the battery connection state is achieved, and hierarchical early warning is used to improve the timeliness of maintenance.

[0073] In addition, the connection state classification result is used to evaluate the connection state of the battery connection, and a preset connection state evaluation relationship is used to determine the connection state evaluation result. The connection state evaluation relationship is constructed based on a plurality of weight coefficients, a plurality of measurement data of the battery connection, and corresponding normal data. That is, the connection state evaluation relationship is used to quantitatively evaluate and accurately analyze the connection state of the battery. In addition, the connection state classification model can include a short-term memory network model and a random forest model. Therefore, the voltage data, the temperature data, the connection impedance data, the short-term memory network model, and the random forest model are used to accurately and timely detect the connection state classification result of the battery connection. Figure 2 is a flowchart of another battery connection state detection method provided by the present disclosure, as shown in Figure 2 The method includes the following steps:

[0074] S210, obtaining a plurality of measurement data of the battery connection; wherein the plurality of measurement data includes voltage data, temperature data, and connection impedance data.

[0075] Here, the implementation of step S210 is described in detail in the related description of the Figure 1 embodiment, which will not be repeated here.

[0076] S220, inputting the voltage data, the temperature data, and the connection impedance data into a short-term memory network model to extract data features of the voltage data, the temperature data, and the connection impedance data; the data features at least include a trend and / or degree of deviation of the voltage data, the temperature data, and the connection impedance data from corresponding normal data.

[0077] The voltage data, the temperature data, and the connection impedance data are all time series data. Specifically, after the voltage data, the temperature data, and the connection impedance data are input into the short-term memory network model, the short-term memory network model processes the time series data to extract the time series features, that is, the data features of the voltage data, the temperature data, and the connection impedance data. These data features can capture important information in the time series data.

[0078] In this way, the short-term memory network model is used to process the voltage data, the temperature data, and the connection impedance data to extract the corresponding data features, so as to capture important information of the voltage data, the temperature data, and the connection impedance data in the fluctuation and change.

[0079] The voltage, temperature and impedance data of the battery connection have strong time sequence, the traditional feature extraction method can only reflect the state at a certain moment, and cannot capture the gradual change trend of the parameters over time. The embodiment utilizes the natural adaptability of the short-term memory network model to the time sequence data, accurately captures the gradual change trend of the parameters over time, and improves the accuracy of the battery connection state detection.

[0080] In S230, the data features are input into a random forest model to generate a connection state classification result of the battery connection. The connection state classification result includes a normal state and an abnormal state, and the abnormal state includes multiple abnormal types.

[0081] Specifically, the data features of the voltage data, temperature data and connection impedance data extracted by the short-term memory network model are used as input features of the random forest model. Then, the random forest model classifies based on the data features of the above data to obtain the connection state classification result of the battery connection. For example, if the connection state classification result is an abnormal state, it can be classified into different abnormal types, such as loose connection, corrosion, overload heating, abnormal charging and discharging, etc. In other embodiments, statistical features (such as mean, variance, etc.) of the voltage data, temperature data and connection impedance data can also be used as additional input features.

[0082] In this way, the random forest model classifies the data features based on the relationship between the data features and the connection state classification result of the battery connection, which helps to realize early detection of various problems such as loose connection, corrosion and degradation.

[0083] The random forest model classifies the data features to realize accurate classification of the connection state with high stability of ensemble learning, breaks through the limitation of traditional methods that only judge whether it is abnormal or not, can identify multiple abnormal types in fine granularity, and the model has stronger adaptability to complex working conditions and unbalanced data, provides reliable decision basis for subsequent grading early warning and differentiated operation and maintenance, and finally improves the fine level of fault diagnosis and operation and maintenance efficiency of the energy storage system.

[0084] In S240, the value of the weight coefficient in the preset connection state evaluation relationship is determined based on the connection state classification result. The connection state evaluation relationship is constructed based on multiple weight coefficients, multiple measurement data of the battery connection and corresponding normal data. The weight coefficient represents the proportion of the difference between the measurement data and the normal data in each connection state classification result.

[0085] Exemplarily, the embodiment can determine the weight coefficient to be adjusted based on the connection state classification result and the pre-stored corresponding relationship between the connection state classification result and the adjustment weight coefficient, and then determine the adjustment size of the value of the weight coefficient to be adjusted according to the data features, wherein the data features at least include the trend and / or degree of deviation of the voltage data, the temperature data and the connection impedance data from the corresponding normal data, so that the value of the weight coefficient in the connection state evaluation relationship can be determined based on the weight coefficient to be adjusted and the adjustment size.

[0086] Specifically, based on the determination of the connection state classification result and the pre-stored corresponding relationship, the weight coefficient mapped thereto can be determined by substituting the connection state classification result into the pre-stored corresponding relationship, and the value of the weight coefficient can be determined based on the severity of the connection state classification result, for example: if the difference between the measurement data and the normal data is large, the severity of the corresponding connection state classification result is higher, and then the proportion of the difference between the measurement data and the normal data in the connection state classification result is high, that is, the value of the weight coefficient is larger; on the contrary, if the difference between the measurement data and the normal data is small, the severity of the corresponding connection state classification result is lower, and then the proportion of the difference between the measurement data and the normal data in the connection state classification result is low, that is, the value of the weight coefficient is smaller.

[0087] Exemplarily, the normal data of the temperature data can be set to 50℃, and the normal data of the connection impedance data can be set to 1 mΩ, which is not limited herein.

[0088] In this way, when the weight coefficient to be adjusted and the adjustment size thereof are determined, the size of the value of each weight coefficient in the connection state evaluation relationship is also determined, so that the output result of the connection state evaluation relationship can be further calculated.

[0089] In the traditional evaluation, the weight coefficient is usually a fixed value, but under different connection states, the characterization ability of each parameter to the state has essential differences, for example, the core feature of slight looseness is slow increase of contact resistance, at this time, the difference between the connection impedance data and the normal data has greater influence on the state evaluation, and the voltage data may not have obvious fluctuation. The embodiment determines the weight coefficient to be adjusted (such as adjusting the impedance weight when the looseness is slight) through the pre-stored corresponding relationship between the connection state classification result and the adjustment weight coefficient, so that the weight distribution is consistent with the physical evolution law of this type of state, and the mismatch of parameter importance caused by the fixed weight is avoided, and the pertinence of the evaluation is fundamentally improved.

[0090] Moreover, the embodiment further optimizes the weight size in combination with the trend and degree of voltage, temperature, and impedance deviating from normal data (data features), such as if a parameter (e.g., impedance) deviates from normal data significantly, it means that the parameter has a more prominent influence on the current state, and the weight of the parameter can be increased; if the deviation is slight, the weight is reduced to avoid small deviation parameters from excessively affecting the evaluation result, solve the limitation that the fixed weight cannot respond to the dynamic change of the parameter, and make the evaluation result more accurately reflect the real risk of the current connection state.

[0091] S250, based on the value of the weight coefficient and the connection state evaluation relationship, a connection state evaluation result is obtained.

[0092] Specifically, since the connection state evaluation relationship is constructed based on multiple weight coefficients, multiple measurement data at the battery connection, and corresponding normal data, and the multiple measurement data at the battery connection and the corresponding normal data in the connection state evaluation relationship are known, on this basis, the value of the determined weight coefficient is substituted into the aforementioned connection state evaluation relationship, and the output result of the connection state evaluation relationship, i.e., the connection state evaluation result, can be calculated.

[0093] In this way, the connection state evaluation relationship realizes the fusion processing of the connection state classification results, comprehensively considers the relevant data corresponding to each connection state classification result, such as the value of the weight coefficient and the corresponding standard deviation, obtains the connection state evaluation result of the overall connection state of the battery connection, and accurately determines the health degree of the connection.

[0094] Optionally, the connection state evaluation relationship of the embodiment satisfies:

[0095]

[0096] wherein S represents the connection state evaluation result, v1, v2, and v3 represent the weight coefficients corresponding to the voltage data, the temperature data, and the connection impedance data, respectively; V ref , T ref , Z ref (ω) represent the normal data corresponding to the voltage data, the temperature data, and the connection impedance data, respectively, and ω represents the angular frequency; ΔV, ΔT, and ΔZ(ω) represent the standard deviations of the voltage data, the temperature data, and the connection impedance data from the corresponding normal data, respectively.

[0097] The connection state evaluation relationship formula is constructed based on the plurality of weight coefficients, the plurality of measurement data of the battery connection and the standard deviations of the corresponding normal data. The standard deviation is used to measure the difference between each measurement data and the corresponding normal data. If the standard deviation is small, it means that the corresponding measurement data is close to the corresponding normal data, and thus the corresponding measurement data is normal and good. Conversely, if the standard deviation is large, it means that there is a large difference between the corresponding measurement data and the corresponding normal data, and thus the corresponding measurement data is abnormal and bad.

[0098] Specifically, the normal data corresponding to the voltage data, the temperature data and the connection impedance data are known data. By substituting the values of the weight coefficients and the standard deviations determined in the foregoing into the connection state evaluation relationship formula, the connection state evaluation result can be obtained by calculating according to the operation relationship in the connection state evaluation relationship formula.

[0099] It should be noted that the connection state evaluation result S is a comprehensive score for the connection state of the battery connection. Since the smaller the standard deviation of the voltage data, the temperature data and the connection impedance data and the corresponding normal data is, the smaller the connection state evaluation result S is, which means that the connection state of the battery connection is healthier.

[0100] For example, the connection state evaluation result S can be 0.5, 0.7, 0.8 or other values, which can be used to evaluate the health degree of the connection state of the battery connection by different values. For example, if the connection state evaluation result S is greater than a score threshold, the health degree of the connection state of the battery connection is worse; if the connection state evaluation result S is less than the score threshold, the health degree of the connection state of the battery connection is better.

[0101] Optionally, the S250 in the embodiment specifically includes:

[0102] Step one, obtaining the corresponding standard deviation based on the voltage data, the temperature data and the connection impedance data and the corresponding normal data.

[0103] The connection state evaluation relationship formula is constructed based on the plurality of weight coefficients, the plurality of measurement data of the battery connection and the standard deviations of the corresponding normal data. The standard deviation is used to measure the difference between each measurement data and the corresponding normal data. If the standard deviation is small, it means that the corresponding measurement data is close to the corresponding normal data, and thus the corresponding measurement data is normal and good. Conversely, if the standard deviation is large, it means that there is a large difference between the corresponding measurement data and the corresponding normal data, and thus the corresponding measurement data is abnormal and bad.

[0104] Specifically, by obtaining the standard deviation between each kind of measurement data and its corresponding normal data based on the voltage data, temperature data and connection impedance data and the corresponding normal data, the connection state evaluation result can be determined further according to the known standard deviation subsequently.

[0105] Step two, substituting the value of the weight coefficient and the standard deviation into the connection state evaluation relationship to obtain the connection state evaluation result.

[0106] Specifically, by substituting the obtained value of the weight coefficient and the standard deviation into the connection state evaluation relationship, the output result of the connection state evaluation relationship can be calculated, i.e. the connection state evaluation result, to determine the health degree of the connection state of the current battery connection through the connection state evaluation result.

[0107] S260, determine the evaluation result interval where the connection state evaluation result is located based on the preset different evaluation result intervals.

[0108] Among them, the preset different evaluation result intervals are divided according to the score threshold and the set interval length. Based on the preset different evaluation result intervals, it is determined which evaluation result interval the connection state evaluation result falls into, so as to determine its warning level according to the result falling into the interval subsequently.

[0109] Exemplarily, the preset different evaluation result intervals can include two evaluation result intervals, three evaluation result intervals or other number of evaluation result intervals divided according to the score threshold and the set interval length, and the number of intervals is not limited here.

[0110] Specifically, based on the determination of the connection state evaluation result and the preset evaluation result intervals, the value of the connection state evaluation result and the preset evaluation result intervals are matched and queried to determine the evaluation result interval to which the value of the connection state evaluation result belongs.

[0111] S270, determine the warning degree based on the evaluation result interval where the connection state evaluation result is located.

[0112] Among them, there is a one-to-one mapping relationship between the evaluation result interval and the warning degree, and by determining the result of the connection state evaluation result falling into the interval, the warning degree can be determined accordingly.

[0113] Exemplarily, Table 2 is a mapping table of an evaluation result interval, a warning level and a processing suggestion provided by the embodiment of the present disclosure, as shown in Table 2, the first column of the table shows different evaluation result intervals, the score threshold of which is 0.8, when S < 0.8, it is normal, the second column shows different warning levels, and along the direction from top to bottom, the warning levels include: normal, attention, warning and serious, and the warning degree increases in turn, the third column shows different processing suggestions, on this basis, the evaluation result interval, the warning level and the processing suggestion of each row are one-to-one corresponding.

[0114] Table 2 Mapping table of evaluation result interval, warning level and processing suggestion

[0115] Evaluation result interval S Warning level Processing suggestion S<0.8 Normal Continue monitoring 0.8≤S<1.2 Pay attention Strengthen inspection 1.2≤S<1.5 Warning Plan maintenance S≥1.5 Serious Immediately stop

[0116] Specifically, in combination with Table 2, after determining the evaluation result interval in which the connection state evaluation result is located, the warning degree and the processing suggestion are further determined, so that the staff can clearly determine the processing priority through the corresponding warning level, and then take corresponding operation and maintenance measures in time. Therefore, the implementation of the hierarchical warning optimizes the operation and maintenance strategy, which can reduce unnecessary shutdown inspection, and greatly reduce the operation and maintenance cost.

[0117] The technical scheme provided by the embodiment of the present disclosure can be widely applied to scenarios such as energy storage power stations, electric vehicles, data centers, and station DC power supply systems. Through accurate detection of the connection state of the battery and maintenance after hierarchical warning, the safety and reliability of the battery system are effectively improved.

[0118] Specifically, after determining the evaluation result interval in which the connection state evaluation result is located, the warning degree mapped thereto can be determined based on the preset correlation between each evaluation result interval and the warning degree, so that the staff can perform corresponding operation and maintenance based on the warning degree subsequently.

[0119] S280, according to the above warning degree, a corresponding level of warning is performed; wherein the level of warning is positively correlated with the warning degree.

[0120] Wherein, the more serious the warning degree is, the higher the level of warning is, and correspondingly, the lower the warning degree is, the lower the level of warning is. In this way, by determining the warning level according to the warning degree, hierarchical warning is realized to give corresponding processing suggestions, provide scientific basis for operation and maintenance decision, and improve the warning accuracy and processing timeliness.

[0121] The embodiments of the present disclosure obtain various measurement data at the battery connection, classify the connection state thereof, obtain the connection state classification result of the battery connection, and further fuse and analyze the connection state classification result to obtain the connection state evaluation result. In this way, multi-dimensional measurement data is cooperatively analyzed to early and accurately diagnose the connection state of the battery, improve the detection sensitivity and real-time performance, reduce false positives and false negatives, and ensure the safe operation of the system. Moreover, based on the connection state evaluation result, the embodiments of the present disclosure establish a hierarchical early warning mechanism and give corresponding processing suggestions to provide a scientific basis for operation and maintenance decisions, thereby helping to extend the service life of the battery system, reduce unnecessary downtime, and promote the sustainable development of large-scale energy storage systems.

[0122] In some embodiments, based on the connection state classification result, the value of the weight coefficient in the preset connection state evaluation relationship is determined in S240, including the following steps: Figure 2

[0123] Step one, based on the connection state classification result and the corresponding relationship between the pre-stored connection state classification result and the adjustment weight coefficient, the weight coefficient to be adjusted is determined.

[0124] For example, the plurality of weight coefficients can include v1, v2 and v3, which respectively represent the weight coefficients corresponding to the voltage data, the temperature data and the connection impedance data. The corresponding relationship between the pre-stored connection state classification result and the adjustment weight coefficient can be in the form of a table. Table 3 is a mapping table of the connection state classification result and the adjustment weight coefficient provided by the embodiments of the present disclosure. As shown in Table 3, different connection state classification results and the adjustment weight coefficients corresponding thereto are shown in the table.

[0125] Table 3: Mapping table of connection state classification result and adjustment weight coefficient

[0126] Connection state classification result Adjust weight coefficient Normal [v1, v2 and v3] Loose connection <![CDATA[v3]]> Contact corrosion <![CDATA[v3]]> Overload heating <![CDATA[v2]]> Abnormal charging and discharging <![CDATA[v2]]> Connection disconnection Connection disconnection <![CDATA[v2]]>

[0127] Specifically, in combination with Table 3, the determined connection state classification result is substituted into the pre-stored mapping table of the connection state classification result and the adjustment weight coefficient to query the adjustment weight coefficient corresponding thereto.

[0128] Step two, the adjustment size of the value of the weight coefficient to be adjusted is determined according to the data characteristics.

[0129] The data characteristics at least include the trend and / or degree of the voltage data, the temperature data and the connection impedance data deviating from the corresponding normal data.

[0130] ​Specifically, in combination with Table 1 and Table 3, when the connection state classification result is an abnormal state, it is embodied as a certain abnormal type, which indicates that the measurement data related thereto is abnormal. In the case of the voltage data and the connection impedance data being normal and the temperature data being abnormal, when the temperature data shows an upward trend relative to the corresponding normal data and deviates from the corresponding normal data to a large extent, it indicates that the temperature data shows a significant upward trend, and thus the temperature data is abnormal to a large extent. At this time, the value of the weight coefficient v2 can be adjusted to be larger and the values of the remaining v1 and v3 can be adjusted to be smaller according to the data characteristics of the temperature data.

[0131] It should be noted that each weight coefficient has an initial value, and the sum of the plurality of weight coefficients is a constant value. Therefore, the adjustment of the size of each weight coefficient is based on the range defined by the initial value and the constant value. For example, v1+v2+v3=1, the initial values of v1, v2 and v3 are 0.3, 0.3 and 0.4 respectively. If the value of the weight coefficient v2 needs to be adjusted to be larger and the values of the remaining v1 and v3 need to be adjusted to be smaller, the value of v2 can be adjusted to 0.5, the value of v1 can be adjusted to 0.2, and the value of v3 can be adjusted to 0.3. In other embodiments, the values can also be adjusted to other sizes according to the detection requirements of the battery connection state, which are not limited herein.

[0132] In addition, if multiple measurement data are abnormal, the sizes of the degrees of abnormality of the multiple measurement data are compared. The larger the degree of abnormality, the larger the corresponding weight coefficient. The smaller the degree of abnormality, the smaller the corresponding weight coefficient. For example, v1+v2+v3=1, the initial values of v1, v2 and v3 are 0.3, 0.3 and 0.4 respectively. If the voltage data, the temperature data and the connection impedance data are all abnormal, the degree of abnormality of the voltage data is the largest, and the degrees of abnormality of the connection impedance data and the temperature data are relatively small. Therefore, the value of v1 can be adjusted to 0.5, the value of v2 can be adjusted to 0.2, and the value of v3 can be adjusted to 0.3. In this way, the size of the weight coefficient is dynamically adjusted.

[0133] Step three, determining the values of the weight coefficients in the connection state evaluation relationship based on the weight coefficient to be adjusted and the adjustment size.

[0134] Specifically, as known from the foregoing steps, since the connection state evaluation relationship is constructed based on the plurality of weight coefficients, the plurality of measurement data of the battery connection and the corresponding normal data, and the plurality of weight coefficients include v1, v2 and v3, when the weight coefficient to be adjusted and the adjustment size are determined, the sizes of the values of the weight coefficients in the connection state evaluation relationship are also determined, thereby facilitating the further calculation of the output result of the connection state evaluation relationship.

[0135] In some embodiments, in Figure 1 and Figure 2On the basis of the above, the connection impedance data of the battery connection is obtained, including the following steps:

[0136] Step one, obtain the M sequence pseudo-random signal.

[0137] Wherein, the M sequence pseudo-random signal is the excitation signal, by using the M sequence pseudo-random signal as the excitation signal, the environmental noise can be effectively suppressed. Illustratively, the M sequence pseudo-random signal can be generated by using a signal generating device, and the length of the M sequence pseudo-random signal can be selected according to the frequency range and resolution required by the sweep, for example: a longer M sequence pseudo-random signal is used to provide higher frequency resolution, in addition, the amplitude of the M sequence pseudo-random signal is usually small, and the specific amplitude can be set to not more than 4.5V, and the length and amplitude of the M sequence pseudo-random signal can be set based on the sweep requirement, which is not limited.

[0138] Step two, in the preset frequency range, the M sequence pseudo-random signal is used to sweep the battery connection, and the different frequency points and corresponding impedance data are obtained.

[0139] Specifically, the M sequence pseudo-random signal covering the preset frequency range is applied to the battery connection, and then the frequency domain response data generated by the battery connection in response to the excitation signal is obtained, and further the frequency domain response data is processed by a related algorithm to obtain the different frequency points and corresponding impedance data, wherein the different frequency points and corresponding impedance data are all discrete data points in the preset frequency range. Illustratively, the preset frequency range can be 0.1kHz-100kHz, and in other embodiments, it can also be set to other frequency ranges according to the actual frequency coverage requirement, which is not limited.

[0140] Illustratively, when collecting the different frequency points and corresponding impedance data, a 24-bit analog-to-digital converter (ADC) can be used to record the applied signal and the frequency domain response data synchronously, and the sampling frequency in this process can be set to 500kS / s, and multi-channel phase synchronization is supported.

[0141] Step three, reconstruct the impedance spectrum in the preset frequency range for the different frequency points and corresponding impedance data, and determine the connection impedance data based on the impedance spectrum.

[0142] Specifically, on the basis of obtaining the different frequency points and corresponding impedance data, they are arranged in frequency order to form a preliminary impedance spectrum, and then the orthogonal matching pursuit (OMP) algorithm is applied to fit the impedance spectrum in the preset frequency range (i.e. the initial impedance spectrum), which is reconstructed into a continuous impedance spectrum under the preset frequency range, so that each frequency point has corresponding impedance data, thereby determining the connection impedance data of the battery connection.

[0143] In the traditional sweep mode, a sine signal sweep is often used, which leads to three defects in impedance spectrum detection. The first defect is high time cost of sweep, for example, in the case of sweep in the same frequency range as the present disclosure, the traditional sweep mode requires 30-60 seconds for complete sweep, which is difficult to meet the demand of online monitoring of impedance spectrum. The second defect is that the impedance data obtained is redundant and difficult to compress. The third defect is that the detection of connection resistance requires disconnecting the system load, which not only affects the normal operation and scheduling benefit of the power station, but also makes it difficult to achieve high-frequency periodic detection. For a large energy storage system containing thousands of connection points, comprehensive detection requires a lot of manpower and time, and the operation and maintenance cost is high.

[0144] Compared with the traditional sweep mode, the battery connection state detection method provided by the embodiments of the present disclosure uses impedance spectrum compressed sensing sweep technology, that is, by using compressed sensing theory, the complete impedance spectrum is reconstructed through non-uniform sparse sweep, the data volume can be compressed to more than ten times of the traditional sweep data volume, ultra-fast detection is achieved, and the sweep time can be shortened from 30-60 seconds of the traditional sweep mode to 0.5-2 seconds, achieving millisecond-level response. In addition, the embodiments of the present disclosure also improve the sensitivity of impedance spectrum detection, can identify a small change of 10% of the connection resistance, and realize online monitoring of impedance spectrum without shutdown detection, thereby facilitating real-time dynamic evaluation of the battery connection state.

[0145] The present disclosure detects the change of electrical characteristics at the battery connection through impedance spectrum detection, thereby achieving early detection of problems such as connection loosening, corrosion, and degradation, and avoiding further deterioration of the fault.

[0146] In some embodiments, the construction process of the short-term memory network model includes:

[0147] An input layer, a short-term memory network layer, and an output layer of the short-term memory network model are established.

[0148] In the input layer of the short-term memory network model, the dimension of the input data is [number of samples, time step, sample type], for example, if each sample has 100 time steps and each time step has 3 types of samples (such as voltage data, temperature data, and connection impedance data), the input dimension is [number of samples, 100, 3].

[0149] The short-term memory network layer is used to capture long-term dependencies in time series data. For example, the number of short-term memory network layers can be set to multiple, so that multiple short-term memory network layers can be used to enhance the expression ability of the short-term memory network model.

[0150] In addition, the output layer can be a fully connected layer, which usually outputs a feature representation of each time step, in the embodiments of the present disclosure, for outputting data features of voltage data, temperature data and connection impedance data.

[0151] In some embodiments, the training process of the short-term memory network model includes the following steps:

[0152] Step one, perform data preprocessing operations.

[0153] Specifically, a large amount of various historical measurement data is collected, and then the historical measurement data is sequentially subjected to data cleaning and data segmentation. The data cleaning process includes filtering the historical measurement data to remove possible noise points, and then filling in missing values of the historical measurement data, such as: if there are missing values in the historical measurement data, the missing values can be filled in by interpolation or using the mean value method, and then the various historical measurement data is standardized or normalized to make its range between [0, 1] or [-1, 1], so as to improve the efficiency and stability of model training. The data segmentation process includes dividing the historical measurement data into a training set, a validation set and a test set, for example, the division can be performed according to the proportions of 70% training set, 15% validation set and 15% test set.

[0154] Step two, perform model training operations.

[0155] Specifically, before model training, a suitable loss function and an optimizer are selected, wherein the loss function is used to measure the gap between the model prediction and the true value, and the optimizer is used to update the model parameters, for example: the loss function can adopt the mean square error (MSE), and the optimizer can adopt the adaptive moment estimation optimization algorithm (Adam). In the model training process, the short-term memory network model is trained using the previously divided training set, and the hyperparameters of the model, such as the learning rate, the number of short-term memory network units, etc., are adjusted through the validation set. When the model training is completed, the short-term memory network model is used to extract data features of the training set, the validation set and the test set, and convert the time series data into output feature representations of the short-term memory network model.

[0156] In some embodiments, the construction process of the random forest model includes constructing a random forest classifier and selecting a suitable number of decision trees.

[0157] The random forest classifier is used to classify the input features into normal state and abnormal state, and to classify the abnormal state into different abnormal types.

[0158] The random forest classifier is composed of multiple decision trees to form an ensemble learning model, each decision tree can independently predict the input domain features, and the final prediction result of the random forest classifier is the mode of the prediction results of all decision trees. For example, the number of decision trees can be set to 100 or more, which can be set according to the classification requirements of the random forest model, and is not limited here.

[0159] In some embodiments, the training process of the random forest model includes training the data features and classification labels extracted by the short-term memory network model, and adjusting the hyperparameters of the random forest, such as the number of decision trees and the maximum depth, through the validation set. The training process of the random forest model can be understood in combination with the related content discussed above, and will not be repeated here.

[0160] In some embodiments, after the short-term memory network model and the random forest model are trained, the following steps are further included:

[0161] Step one, evaluate the short-term memory network model and the random forest model based on the test set, and calculate the classification accuracy and recall rate and other indicators.

[0162] Step two, analyze the performance of the short-term memory network model and the random forest model in classifying various measurement data, find out the weaknesses of the short-term memory network model and the random forest model, and evaluate the performance of the above models to obtain the model performance evaluation result.

[0163] Step three, adjust the hyperparameters of the short-term memory network model and the random forest model according to the model performance evaluation result.

[0164] For example, if the short-term memory network model and the random forest model perform poorly in classifying various measurement data, the number of samples corresponding to the connection state classification result can be increased, or further feature engineering can be performed on the various measurement data.

[0165] Step four, deploy the trained short-term memory network model and random forest model to practical applications for real-time classification of various measurement data.

[0166] It should be noted that in subsequent applications, the trained short-term memory network model and random forest model are usually updated and maintained to adapt to changes in various measurement data, thereby improving the accuracy of battery connection state classification.

[0167] The connection state classification model provided by the embodiments of the present disclosure is composed of a short-term memory network model and a random forest model, forming a hybrid model combining deep learning and traditional machine learning methods. The short-term memory network model effectively processes time series data and captures complex dynamic relationships in the time series data to extract data features. On this basis, the random forest model efficiently classifies according to the data features extracted by the short-term memory network model to obtain the connection state classification result.

[0168] In some embodiments, on the basis of Figure 1 and Figure 2 , after the relevant steps are classified and pre-warned, the connection state evaluation result determined by the evaluation result interval and the pre-warning level can be manually checked and reviewed, and then the relevant information that has been reviewed without errors is fed back into the machine learning library for optimization learning and improvement, forming a machine learning closed loop. In this way, according to the machine learning optimization, the detection accuracy of the battery connection state can be continuously optimized, and the pre-warning accuracy can be improved.

[0169] It should be noted that as the single scale of the energy storage power station expands, the number of connection points grows exponentially. Taking a 100MWh energy storage power station as an example, its battery system may contain more than 100,000 electrical connection points, so the traditional manual inspection and regular maintenance mode has been difficult to meet the reliability management needs of such a large scale. In view of this, the embodiments of the present disclosure develop advanced detection technologies for online monitoring and intelligent diagnosis to realize real-time evaluation and early warning of the connection state, which is an inevitable choice to improve the safe operation level of the energy storage power station and the efficiency of battery connection state detection.

[0170] In order to better understand the application of the battery connection state detection method provided by the embodiments of the present disclosure to the energy storage power station, the following will be described in conjunction with specific examples.

[0171] Exemplarily, taking the detection of the battery connection state of a 2MVh energy storage container as an example, the system uses lithium iron phosphate (LFP) batteries, the capacity of the battery monomer can be 280Ah, and the batteries are connected by bolts. When connected, a standard torque of 15N·m is used. In actual application, the system can set multiple monitoring points, such as covering 576 battery monomers, 24 battery modules and 4 battery clusters. On this basis, the relevant data of the battery connection in the entire energy storage system are monitored and a data comparison table is formed.

[0172] For example, Table 4 is a data comparison table for various measurement indexes provided by an embodiment of the present disclosure. As shown in Table 4, the first column shows various measurement indexes from top to bottom, including series resistance, connection impedance, temperature difference of connection point, voltage standard deviation at connection point, etc. The second column shows the normal range corresponding to the various measurement indexes. The third column shows the abnormal measurement value corresponding to the various measurement indexes. The fourth column shows the change rate of the abnormal measurement value relative to the normal range. The measurement indexes, normal range, abnormal measurement value and change rate in each row are mapped.

[0173] Table 4 is a data comparison table for various measurement indexes

[0174] Measurement index Normal range Abnormal measurement value Change rate Series resistance 3.2±0.5mΩ 7.8mΩ +144％ Connection impedance <0.2 mΩ 1.5mΩ +650％ Connection point temperature difference <3℃ 9.5℃ +217％ Voltage standard deviation <0.5％ 1.2％ +140％

[0175] Exemplarily, in the process of detecting the battery connection state, the operations performed can include, in sequence, pre-warning of the corresponding level after the comprehensive score is greater than the score threshold, automatic triggering of manual or infrared camera review, on-site confirmation of the loosening of the battery connection, and optimization of the related model (such as the connection state classification model). The time interval of the above-mentioned execution operations can be 15 minutes or other time length, which can be set according to the actual detection situation, and is not limited herein.

[0176] Compared with the traditional detection method, the battery connection state detection method provided by the embodiment of the present disclosure can discover the abnormal state of the battery connection in advance, so as to carry out corresponding operation and maintenance, thereby avoiding the occurrence of fire accidents caused by the abnormality of the battery connection, ensuring the safe and efficient operation of the energy storage power station, reducing the operation and maintenance cost and loss cost, and achieving remarkable economic benefits.

[0177] It should be understood that the size of the serial number of each step in the above-mentioned embodiment does not mean the order of execution. The execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiment of the present disclosure.

[0178] The following is a device embodiment of the present disclosure. For details not described in detail, reference can be made to the corresponding method embodiments described above.

[0179] Figure 3 is a structural schematic diagram of a battery connection state detection device provided by an embodiment of the present disclosure. For the convenience of description, only the parts related to the embodiment of the present disclosure are shown, and the details are as follows:

[0180] As Figure 3As shown, the battery connection state detection apparatus 03 comprises: an acquisition module 31 configured to acquire a plurality of measurement data of the battery connection; wherein the plurality of measurement data comprises voltage data, temperature data and connection impedance data; a classification module 32 configured to input the voltage data, the temperature data and the connection impedance data into a trained connection state classification model to obtain a connection state classification result of the battery connection; wherein the connection state classification model is trained based on historical voltage data, historical temperature data, historical connection impedance data and corresponding historical connection state classification results of a plurality of battery connections; an evaluation module 33 configured to evaluate the connection state of the battery connection based on the connection state classification result to obtain a connection state evaluation result; and an early warning module 34 configured to perform hierarchical early warning according to the connection state evaluation result to perform corresponding operation and maintenance.

[0181] In some embodiments, the connection state classification model comprises a short-term memory network model and a random forest model; the classification module 32 is configured to input the voltage data, the temperature data and the connection impedance data into the short-term memory network model to extract data features of the voltage data, the temperature data and the connection impedance data; the data features at least comprise trends and / or degrees of deviation of the voltage data, the temperature data and the connection impedance data from corresponding normal data; and input the data features into the random forest model to generate the connection state classification result of the battery connection; the connection state classification result comprises a normal state and an abnormal state, and the abnormal state comprises a plurality of abnormal types.

[0182] In some embodiments, the evaluation module 33 is configured to obtain corresponding standard deviations based on the voltage data, the temperature data and the connection impedance data and corresponding normal data; and then obtain the connection state evaluation result according to a weight coefficient, the standard deviations and a connection state evaluation relationship.

[0183] In some embodiments, the early warning module 34 is configured to determine an evaluation result interval in which the connection state evaluation result is located based on preset different evaluation result intervals; then determine a warning degree based on the evaluation result interval in which the connection state evaluation result is located; and finally perform early warning of a corresponding level according to the warning degree; wherein the level of early warning is positively correlated with the warning degree.

[0184] Figure 4 is a schematic diagram of an electronic device provided by an embodiment of the present disclosure. As shown in Figure 4 The electronic device 04 of this embodiment comprises a processor 41 and a memory 42. The memory 42 stores a computer program 43. The processor 41 implements the steps in each of the method embodiments described above when executing the computer program 43. Alternatively, the processor 41 implements the functions of each module / unit in each of the device embodiments described above when executing the computer program 43.

[0185] For example, the computer program 43 can be divided into one or more modules / units, which are stored in the memory 42 and executed by the processor 41 to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program 43 in the electronic device 04.

[0186] The electronic device 04 can include, but is not limited to, the processor 41, the memory 42. Those skilled in the art can understand that, Figure 4 The electronic device 04 is only an example and does not constitute a limitation on the electronic device 04, and can include more or fewer components than the illustration, or combine certain components, or different components, for example, the electronic device 04 can also include an input / output device, a network access device, a bus, etc.

[0187] For the convenience and brevity of description, only the above-mentioned division of each functional module / unit is exemplified, and in actual application, the above-mentioned functions can be completed by different functional modules / units according to needs. The above-mentioned modules / units can be realized in the form of hardware, software or a combination of hardware and software.

[0188] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described or recorded in a certain embodiment can be referred to the related description of other embodiments. If there is no special description and logical conflict, the terms and / or descriptions of different embodiments are consistent and can be mutually referenced, and the technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationship.

[0189] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A battery connection state detection method characterized by, The method comprises: acquiring a plurality of measurement data of a battery connection; wherein the plurality of measurement data comprises voltage data, temperature data and connection impedance data; inputting the voltage data, temperature data and connection impedance data into a trained connection state classification model to obtain a connection state classification result of the battery connection; wherein the connection state classification model is trained based on historical voltage data, historical temperature data, historical connection impedance data and corresponding historical connection state classification results of a plurality of battery connections; based on the connection state classification result, evaluating the connection state of the battery connection to obtain a connection state evaluation result; based on the connection state evaluation result, performing a hierarchical early warning to perform corresponding operation and maintenance.

2. The battery connection state detection method according to claim 1, characterized by, The connection state evaluation result based on the connection state classification result comprises: determining the value of the weight coefficient in the preset connection state evaluation relationship based on the connection state classification result; the connection state evaluation relationship is constructed based on a plurality of weight coefficients, a plurality of measurement data of the battery connection and corresponding normal data, and the weight coefficient represents the proportion of the difference between the measurement data and the normal data in each connection state classification result; based on the value of the weight coefficient and the connection state evaluation relationship, the connection state evaluation result is obtained.

3. The battery connection state detection method according to claim 2, characterized by, The value of the weight coefficient in the preset connection state evaluation relationship based on the connection state classification result comprises: based on the connection state classification result and the corresponding relationship between the pre-stored connection state classification result and the adjustment weight coefficient, the weight coefficient to be adjusted is determined; based on the data characteristics, the adjustment size of the value of the weight coefficient to be adjusted is determined; wherein the data characteristics at least include the trend and / or degree of deviation of the voltage data, temperature data and connection impedance data from the corresponding normal data; based on the weight coefficient to be adjusted and the adjustment size, the value of the weight coefficient in the connection state evaluation relationship is determined.

4. The battery connection state detection method according to claim 1, characterized by, The connection impedance data of the battery connection comprises: acquiring an M sequence pseudo-random signal; in a preset frequency range, the M sequence pseudo-random signal is used to sweep the battery connection to obtain different frequency points and corresponding impedance data; based on the impedance spectrum, the connection impedance data is determined.

5. The battery connection state detection method according to claim 1, characterized by, The connection state classification model comprises a short-term memory network model and a random forest model; The voltage data, temperature data and connection impedance data are input into the pre-trained connection state classification model to obtain the connection state classification result of the battery connection, comprising: inputting the voltage data, temperature data and connection impedance data into the short-term memory network model to extract the data characteristics of the voltage data, temperature data and connection impedance data; the data characteristics at least include the trend and / or degree of deviation of the voltage data, temperature data and connection impedance data from the corresponding normal data; The data features are input into a random forest model to generate a connection state classification result of the battery connection; the connection state classification result includes a normal state and an abnormal state, and the abnormal state includes multiple abnormal types.

6. The battery connection state detection method according to claim 2 or 3, characterized by, The connection state evaluation result is obtained based on the value of the weight coefficient and the connection state evaluation relationship. Based on the voltage data, temperature data, and connection impedance data and corresponding normal data, corresponding standard deviations are obtained. The connection state evaluation result is obtained according to the value of the weight coefficient, the standard deviation, and the connection state evaluation relationship.

7. The battery connection state detection method according to claim 6, characterized by, The connection state evaluation relationship satisfies: Wherein, S represents the connection state evaluation result, v1, v2, v3 respectively represent the weight coefficients corresponding to the voltage data, the temperature data and the connection impedance data; V ref , T ref , Z ref (ω) respectively represent the normal data corresponding to the voltage data, the temperature data and the connection impedance data, ω represents the angular frequency; ΔV, ΔT, ΔZ(ω) respectively represent the standard deviation of the voltage data, the temperature data and the connection impedance data and the corresponding normal data.

8. The battery connection state detection method according to any one of claims 1 to 7, characterized by, The classification and warning are performed according to the connection state evaluation result, including: Based on different preset evaluation result intervals, the evaluation result interval in which the connection state evaluation result is located is determined; Based on the evaluation result interval in which the connection state evaluation result is located, the warning degree is determined; According to the warning degree, a corresponding level of warning is performed; wherein the level of warning is positively correlated with the warning degree.

9. A battery connection state detection device, comprising: An acquisition module for acquiring multiple measurement data of a battery connection; wherein the multiple measurement data include voltage data, temperature data, and connection impedance data; A classification module for inputting the voltage data, temperature data, and connection impedance data into a trained connection state classification model to obtain a connection state classification result of the battery connection; wherein the connection state classification model is trained based on historical voltage data, historical temperature data, historical connection impedance data, and corresponding historical connection state classification results of multiple battery connections; An evaluation module for evaluating the connection state of the battery connection based on the connection state classification result to obtain a connection state evaluation result; A warning module for performing a classification and warning according to the connection state evaluation result to perform corresponding operation and maintenance.

10. An electronic device, comprising: A memory and a processor, the memory stores a computer program, and the processor implements the method of any one of claims 1 to 8 when executing the computer program.

Images