A heat management and power coordination control method and system of a liquid-cooled energy storage system

Abstract

This invention relates to the field of liquid-cooled energy storage thermoelectric control technology, specifically disclosing a method and system for thermal management and power coordinated control of a liquid-cooled energy storage system. The method is applied to the control device of the liquid-cooled energy storage system, which includes a first liquid-cooled unit, a second liquid-cooled unit connected in parallel, and a valve assembly containing a first valve and a second valve. This invention, through feature fusion and conflict level determination of multi-dimensional data, enables the system to identify and predict the contradiction between power demand and thermal safety. It utilizes a dynamic weight allocation strategy to find the optimal balance between power tracking, temperature control, and energy consumption. Combined with a heterogeneous collaborative mechanism of dual liquid-cooled units, the high-power unit ensures a global reference temperature, while the low-power unit enables rapid targeted adjustment for local hotspots or power surges. This eliminates thermal hysteresis and avoids energy waste caused by over-powering the system.

Description

Technical Field

[0001] This invention relates to the field of liquid-cooled energy storage thermoelectric control technology, specifically to a method and system for thermal management and power coordinated control of a liquid-cooled energy storage system. Background Technology

[0002] With the increasing penetration of renewable energy, the demand for frequency regulation and peak shaving of large-scale energy storage systems in the power grid is becoming increasingly urgent, and liquid-cooled energy storage systems have become the mainstream choice due to their high energy density and heat exchange efficiency. However, the existing energy storage system control architecture usually treats battery thermal management and power dispatch as two independent control islands: the thermal management system passively adjusts the cooling power based only on the real-time battery temperature, resulting in significant thermal lag; while the power dispatch system only responds to grid commands and lacks proactive perception of the battery thermal state.

[0003] This storage-computing separation control mode makes it impossible for the system to predict the heat accumulation trend under high-frequency power throughput, making it difficult to intervene in the heat in advance under complex operating conditions, which seriously restricts the dynamic response capability of the energy storage system.

[0004] More importantly, when faced with rapid frequency regulation or power surge commands from the power grid, existing technologies face a severe triangular conflict problem, namely the irreconcilable differences between grid response speed, battery thermal safety, and system cooling energy consumption.

[0005] Specifically, in order to meet the millisecond-level grid power response, the battery will experience a sharp temperature rise in a short period of time. If the cooling system maintains high power operation for a long time for absolute safety, it will lead to huge parasitic energy consumption and may cause condensation short circuit risk. If the cooling system is used for energy saving or response lag, it may trigger high temperature protection, causing the system to shut down, which in turn will result in grid performance failure.

[0006] Furthermore, under extreme high temperature or high load conditions, traditional logic based on hard thresholds often forcibly cuts off power because it cannot find a feasible solution that satisfies all constraints, thus failing to achieve flexible support and continuous operation under extreme conditions. Summary of the Invention

[0007] This invention aims to at least partially address one of the technical problems in related technologies. Therefore, the objective of this invention is to propose a method and system for thermal management and power coordination control of a liquid-cooled energy storage system, thereby improving the availability and safety of the liquid-cooled energy storage system.

[0008] To achieve the above objectives, a first aspect of the present invention provides a thermal management and power coordination control method for a liquid-cooled energy storage system, which is applied in a control device for a liquid-cooled energy storage system. The liquid-cooled energy storage system includes a first liquid-cooled unit, a second liquid-cooled unit, and a valve assembly including a first valve and a second valve connected in parallel.

[0009] The method includes: responding to the collected multi-dimensional operational data, performing feature fusion on power grid dispatching requirements and battery thermal status to determine the conflict level and power grid dispatching scenario at the current moment;

[0010] Based on the conflict level and the power grid dispatch scenario, a three-dimensional prediction model of the thermal power grid is invoked for multi-objective optimization to calculate the dynamic allocation weights of four dimensions: power point tracking, temperature safety, cooling energy consumption, and power grid response.

[0011] Based on the dynamically allocated weights, coordinated operation instructions and battery power adjustment instructions are generated for the first liquid-cooled unit and the second liquid-cooled unit to perform phased control of the liquid-cooled energy storage system.

[0012] To achieve the above objectives, a second aspect of the present invention provides a thermal management and power coordination control system for a liquid-cooled energy storage system, the system comprising:

[0013] The data processing module is used to collect multi-dimensional operational data and perform feature fusion on grid dispatching requirements and battery thermal status to determine the conflict level and grid dispatching scenario.

[0014] The decision module is used to run a three-dimensional prediction model of the thermal power grid, perform multi-objective optimization based on the conflict level and the grid dispatch scenario, and output dynamic weights for four dimensions: power tracking, temperature safety, cooling energy consumption and grid response.

[0015] The execution control module is used to generate control signals according to the dynamically allocated weights, and adjust the operating power of the first liquid cooler unit, the operating power of the second liquid cooler unit, the opening degree of the valve assembly, and the output power of the battery cluster, respectively.

[0016] To achieve the above objectives, a third aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory. When the computer program is executed by the processor, it implements the above-described method for thermal management and power coordinated control of a liquid-cooled energy storage system.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0018] This invention provides a thermal management and power coordination control method and system for a liquid-cooled energy storage system, which effectively solves the multi-objective conflict problem of energy storage systems in complex scheduling scenarios. Its core advantages are:

[0019] By fusing features from multidimensional data and determining conflict levels, the system can identify and predict the contradiction between power demand and thermal safety. It uses a dynamic weight allocation strategy to find the optimal balance between power tracking, temperature control, and energy consumption. Combined with the heterogeneous collaboration mechanism of dual liquid-cooled units, the system uses high-power units to ensure the global reference temperature and low-power units to achieve rapid targeted adjustment for local hot spots or power surges. This eliminates thermal hysteresis and avoids energy waste caused by using a large power unit for a small load.

[0020] In addition, the introduced speculative precooling and soft constraint relaxation mechanisms enable the system to prevent condensation risks by dew point clamping when facing false alarms or extreme operating conditions, or to sacrifice short-term heat capacity to ensure continuous execution of grid commands. This achieves a leap from passive defense to proactive optimization, significantly improving the all-weather availability and operational safety of the liquid-cooled energy storage system. Attached Figure Description

[0021] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0022] Figure 1 This is a flowchart illustrating the thermal management and power coordinated control method for the liquid-cooled energy storage system provided by the present invention.

[0023] Figure 2 This is a schematic diagram of the three-dimensional conflict level determination space based on multi-dimensional feature fusion in the thermal management and power coordinated control method of the liquid-cooled energy storage system provided by the present invention.

[0024] Figure 3 This is a multi-objective Pareto optimization front distribution diagram of the thermoelectric coordinated control in the thermal management and power coordinated control method of the liquid-cooled energy storage system provided by the present invention;

[0025] Figure 4 This is a schematic diagram of the power mutation prediction curve and confidence interval based on LSTM in the thermal management and power coordinated control method of the liquid-cooled energy storage system provided by the present invention.

[0026] Figure 5 This invention provides a comparison of battery temperature rise under power surge conditions in the thermal management and power coordinated control method of the liquid-cooled energy storage system provided by the present invention, between traditional control and the coordinated control of the present invention under power surge conditions.

[0027] Figure 6 This is a schematic diagram of the evolution of the soft constraint penalty function in the constraint relaxation mode of the thermal management and power coordinated control method of the liquid-cooled energy storage system provided by the present invention.

[0028] Figure 7This is a safety protection trajectory diagram of fluid temperature and dew point during the asynchronous speculative precooling process in the thermal management and power coordination control method of the liquid-cooled energy storage system provided by the present invention.

[0029] Figure 8 This is a schematic diagram illustrating the implementation of the thermal management and power coordination control system of the liquid-cooled energy storage system provided by the present invention;

[0030] Figure 9 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0031] This method is applicable to various liquid-cooled energy storage systems, including but not limited to energy storage power stations using lithium iron phosphate and ternary lithium battery systems, and performs particularly well in dynamic scenarios such as grid frequency regulation and peak shaving. The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0032] The following description, with reference to the accompanying drawings, outlines a method, system, and electronic device for thermal management and power coordination control of a liquid-cooled energy storage system according to embodiments of the present invention.

[0033] Example 1:

[0034] This embodiment provides a thermal management and power coordination control method for a liquid-cooled energy storage system. The method is configured to be executed in a control device of a liquid-cooled energy storage system, such as a BMS battery management system, an EMS energy management system, or an independent coordination controller.

[0035] Specifically, before introducing the specific control process, the hardware architecture upon which this embodiment is based will be described in detail. The liquid-cooled energy storage system adopts a heterogeneous dual-cycle topology in its physical architecture, comprising a first liquid-cooled unit and a second liquid-cooled unit connected in parallel. The first liquid-cooled unit is configured as a high-power main refrigeration unit, possessing high cooling capacity and high flow throughput, primarily used to handle the system-level baseline heat load. The second liquid-cooled unit is configured as a low-power auxiliary or precision regulation unit, possessing a wider frequency conversion regulation range and higher response accuracy, primarily used to eliminate local hot spots, maintain temperature uniformity, and cope with low-load conditions.

[0036] The two units are connected in parallel to the main pipe via pipelines and then to a valve assembly containing a first valve and a second valve. The first valve and the second valve can be proportional regulating valves or electronic expansion valves, respectively, used to finely control the flow of coolant to different battery clusters or modules. This hardware design provides the physical basis for subsequent coordinated control.

[0037] like Figure 1 As shown, the method described in this embodiment mainly includes the following steps:

[0038] Step S1: In response to the collected multi-dimensional operational data, which includes at least one of battery temperature, state of charge, grid dispatch instructions, and ambient temperature and humidity, feature fusion is performed on grid dispatch requirements and battery thermal state to determine the conflict level and grid dispatch scenario at the current moment.

[0039] Specifically, to break down the information silos between thermal management and power scheduling in traditional energy storage systems, this step first establishes a multi-dimensional state awareness space. The collected multi-dimensional operational data includes, but is not limited to: real-time cell temperature on the battery side (including maximum, minimum, average, and temperature difference distribution), battery state of charge (SOC), and battery state of health (SOH); real-time scheduling commands on the grid side (including AGC and AVC commands), peak and off-peak electricity price signals, and scheduling priority indicators; and environmental data such as ambient temperature and humidity.

[0040] It is also important to note that feature fusion is not simply data aggregation, but rather a calculation of key performance indicators of the system's operation using specific algorithmic logic. To accurately quantify the current conflict state faced by the system, this embodiment determines the conflict level through the following specific calculation logic:

[0041] First, calculate the power demand difference. Temperature safety margin Cooling energy consumption ratio and grid response deviation Regarding the power demand difference It is defined as the dispatch target power issued by the power grid. With the current maximum available charge and discharge power of the battery system The absolute value of the difference between them is expressed by the formula:

[0042] ;

[0043] in, It is based on the power boundary under the current SOC and SOH constraints.

[0044] For temperature safety margin It reflects the current temperature of the hottest battery cell. Distance from system high temperature alarm threshold The degree of closeness is expressed by the formula:

[0045] ;

[0046] in, The reference operating temperature set for the system is usually the battery's optimal operating temperature or initial design temperature, such as 25°C.

[0047] Regarding the proportion of cooling energy consumption It characterizes the impact of parasitic power consumption of the thermal management system on energy storage efficiency, and is calculated as the total power consumption of the two liquid-cooled units and auxiliary equipment at the current moment. Total output power of energy storage system The ratio:

[0048] ;

[0049] in, To prevent tiny positive numbers with a denominator of zero.

[0050] For power grid response deviation Its calculation is the actual grid-connected power. With scheduling target power Deviation rate between:

[0051] ;

[0052] in, This is the rated power of the energy storage system.

[0053] Optionally, after obtaining the above four characteristic parameters, the system determines the conflict level to one of the following three levels based on the preset numerical range of the power demand difference, the temperature safety margin, the cooling energy consumption ratio, and the grid response deviation:

[0054] 1. Level 1 conflict (power priority conflict): When Greater than the first preset threshold, and If the power consumption is within a safe range, such as greater than 20%, the main challenge for the system is how to meet the huge power throughput, while thermal risks are temporarily secondary.

[0055] 2. Secondary conflict (equilibrium state type conflict): When all parameters are in the middle range, that is, while the system meets the power demand, heat accumulation begins to appear, and it is necessary to take into account the balance between power and temperature.

[0056] 3. Level 3 conflict (temperature safety priority conflict): When If the value is less than the second preset threshold, such as less than 5%, the battery is about to trigger thermal runaway or high temperature protection. At this point, regardless of the power demand, the system must prioritize survival safety.

[0057] like Figure 2As shown, this three-dimensional decision space diagram intuitively illustrates the decision logic space constructed by the control system of this invention based on multi-dimensional operating data. The X-axis represents the power demand difference, with a value range from zero to one hundred kilowatts; the Y-axis represents the temperature safety margin, with a value range from zero percent to one hundred percent; and the Z-axis represents the cooling energy consumption percentage, with a value range from zero percent to twenty percent.

[0058] pass Figure 2 The different colored dots clearly delineate three key control zones. The red dots at the bottom are concentrated in the area where the temperature safety margin is less than 15%, which corresponds to a level three conflict, namely a temperature safety priority conflict. In this case, regardless of the power demand and energy consumption ratio, the system will force the battery thermal safety to be the top priority.

[0059] The blue scattered area in the upper right corner of the diagram is located in a space where the power demand difference is greater than 60 kilowatts and the temperature safety margin is relatively high. This corresponds to the first-level conflict, namely the power priority conflict, indicating that the system mainly responds to the high-power dispatching command of the power grid in this range.

[0060] The yellow scattered area in the middle represents the second-level conflict, namely the equilibrium state conflict. At this time, the system satisfies the power output while also taking into account the optimization of thermal management energy consumption.

[0061] The three-dimensional spatial diagram proves that the present invention does not use a simple linear threshold judgment, but achieves accurate clustering and prediction of the system's operating state through the three-dimensional fusion of multi-dimensional features, thereby ensuring the scientificity and effectiveness of the subsequent dynamic weight allocation strategy.

[0062] Specifically, in addition to the conflict level, this step also requires the simultaneous determination of the power grid dispatch scenario. The logic for determining the power grid dispatch scenario strictly adheres to the economic and functional indicators of the power grid, specifically including:

[0063] Scenario A (Power Priority Scheduling Scenario): When the peak-valley electricity price signal is detected as peak, i.e., a high-price discharge period, and the scheduling priority is high, the system determines that it is currently in a power priority scheduling scenario. In this scenario, in order to maximize economic benefits and grid support capacity, the system logic will automatically relax the upper limit of the temperature safety range (e.g., allow short-term temperature rise to 35°C instead of the standard 25°C) and the cooling energy consumption ratio threshold in exchange for releasing power output capacity.

[0064] Scenario B (Cooling Maintenance Scenario): When the signal during the peak-valley electricity price period is detected as being in the off-peak period, i.e., during low-price charging or idle periods, and the scheduling priority is low, the system determines it to be a cooling maintenance scenario. At this time, the system focuses on battery life and energy efficiency, and therefore tightens the temperature safety margin (aiming for the optimal temperature control point of 25℃) and the cooling energy consumption ratio threshold (limiting high-energy-consumption cooling modes), utilizing low-load gaps for refined thermal conditioning.

[0065] Scenario C (Fast Response Scenario): When receiving a command containing a frequency modulation response, such as a primary frequency modulation command with a required power adjustment rate. When the rate is less than or equal to a preset threshold, it is considered a fast response scenario. This scenario requires the system to have extremely low latency, and the thermal management system must adjust agilely to accommodate rapid power fluctuations.

[0066] Step S2: Based on the conflict level and the power grid dispatch scenario, call the three-dimensional prediction model of the thermal power grid to perform multi-objective optimization and calculate the dynamic allocation weights of four dimensions: power tracking, temperature safety, cooling energy consumption and power grid response.

[0067] Specifically, in order to achieve coordinated control, this embodiment introduces a three-dimensional prediction model of a thermoelectric power grid that combines physical and data-driven approaches.

[0068] Optionally, the step of calling the three-dimensional prediction model of the thermal power grid for multi-objective optimization specifically includes the following sub-steps:

[0069] First, using the thermoelectric coupling model in the aforementioned three-dimensional prediction model of the thermoelectric power grid, the battery temperature field change is predicted based on thermal diffusion characteristics and cooling efficiency characteristics. This model describes the dynamic process of battery heat generation and dissipation through a discretized thermal resistance network method or a simplified finite element model. Its state equation can be expressed as:

[0070] ;

[0071] in, For battery heat capacity, For battery temperature, For the predicted current sequence, For internal resistance, To consider the overall heat transfer coefficient, This refers to the coolant flow rate. Let be the fluid temperature. Using this model, the system can predict the temperature trajectory that a given power curve will lead to over a future period.

[0072] Secondly, using the grid response prediction sub-model in the aforementioned three-dimensional prediction model of the thermal power grid, the deviation between the power adjustment rate constraint and the grid response is predicted based on the dispatch power demand input. This sub-model considers the ramp-up rate limit of the PCS (Power Conversion System) and communication delay, and outputs the predicted actual power curve. .

[0073] Next, multi-objective Pareto optimization is performed. The system constructs a multi-objective optimization problem with four objective functions:

[0074] 1. Minimize power point tracking error: ;

[0075] 2. Maximize temperature safety (i.e., minimize the risk of high temperatures): ;

[0076] 3. Minimize cooling energy consumption: ;

[0077] 4. Minimize grid response delay: .

[0078] The conflict level is matched with the power grid dispatching scenario, and the optimal control strategy is selected based on the matching results in the multi-objective Pareto optimization front. This means that the system does not seek an absolutely optimal solution, but rather a compromise solution that fits the current scenario among multiple conflicting objectives.

[0079] like Figure 3 As shown, the multi-objective Pareto optimization front distribution diagram intuitively demonstrates the optimization results and decision-making logic of the control system of the present invention in the multi-dimensional objective function space. Figure 3 A three-dimensional coordinate system was established, which includes the target values ​​of power tracking error, cooling energy consumption, and temperature risk. The smaller the value of the coordinate axis, the better the performance of that dimension.

[0080] Figure 3 The set of colored scattered points distributed in a curved shape constitutes the Pareto optimal front of the system. Each point on this surface represents a non-dominated solution that cannot improve a certain objective without compromising other objectives. The dynamic weight allocation strategy of the system under different scheduling scenarios is clearly revealed by the three special geometric points marked in the figure.

[0081] Among them, the red pentagram markers are located in the region with extremely small power tracking error but relatively high temperature risk and cooling energy consumption. This corresponds to the power priority scheduling scenario, in which the system sacrifices some thermal safety margin and energy consumption to achieve accurate execution of grid commands. The green square markers are located in the region with low temperature risk and low cooling energy consumption but large power tracking error. This corresponds to the cooling maintenance scenario, in which the system prioritizes battery life and energy saving during off-peak hours and relaxes power response requirements. The blue circular markers are located in the central region of the curved surface, corresponding to the equilibrium state scenario, where all indicators are at a moderate level.

[0082] The figure demonstrates that the present invention does not employ a single fixed control parameter, but rather can calculate and select the compromise solution that best meets the current working conditions among conflicting multidimensional objectives in real time, thereby maximizing the overall system efficiency.

[0083] It is also important to note that after selecting the optimal solution, it needs to be translated into specific weight allocations for the execution layer to understand. The dynamic weight allocation for the four dimensions of power point tracking, temperature safety, cooling energy consumption, and grid response is calculated according to the following logic:

[0084] Logic 1: If the scenario is determined to be a power-priority scheduling scenario (corresponding to peak discharge), then the maximum weight is assigned. Assign the second largest weight to the power tracking dimension. To the aforementioned temperature safety dimension. For example, setting... This ensures that the system prioritizes power generation while maintaining thermal safety, allowing for temporary tolerance of higher energy consumption.

[0085] Logic 2: If the scenario is determined to be the aforementioned cooling maintenance scenario (corresponding to a low-temperature recovery period), then the maximum weight is assigned. Assign the second largest weight to the aforementioned temperature safety dimension. To the aforementioned cooling energy consumption dimension. For example, set... At this point, the system focuses on regulating the battery temperature to its optimal state with minimal energy consumption, thereby extending battery life.

[0086] Logic 3: If the scenario is determined to be a rapid response scenario (corresponding to frequency regulation), then increase the weight of the power grid response dimension. This makes it higher than the weight of the cooling energy consumption dimension. For example, setting This ensures the system responds to grid commands within milliseconds, preventing response timeouts caused by thermal management lag.

[0087] Step S3: Based on the dynamically allocated weights, generate coordinated operation instructions and battery power adjustment instructions for the first liquid cooling unit and the second liquid cooling unit to perform phased control of the liquid-cooled energy storage system.

[0088] Specifically, the allocation of weights ultimately translates into the actions of the physical equipment. This embodiment achieves a synergistic effect of large units providing a safety net and small units providing adjustment by precisely controlling the opening degrees of the first liquid cooling unit, the second liquid cooling unit, the first valve, and the second valve.

[0089] For example, the generation of cooperative operation instructions based on the dynamically allocated weights includes the following typical control modes:

[0090] Mode 1 (Powerful Cooling Mode): In a scenario of primary conflict, i.e., high power demand and power-priority scheduling, the system determines that the heat load is extremely high. At this time, the first liquid chiller is controlled to operate at rated power (Full-Load) to provide the maximum basic flow and cooling capacity; at the same time, the second liquid chiller is controlled to operate at auxiliary power, such as 50%-80% load, as a dynamic supplement; and the first valve and the second valve are opened to the fully open or large-open state to reduce flow resistance and ensure that all battery clusters receive sufficient cooling.

[0091] Mode 2 (Energy-Saving Equalizing Temperature Mode): In scenarios involving Level 3 conflict (extremely high thermal risk) and cooling maintenance (although it's Level 3 conflict, in a maintenance scenario it means there might be a localized thermal runaway risk that needs to be eliminated, rather than a system-wide overheating), or during routine maintenance, the first and second liquid-cooled units are controlled to enter self-circulation mode. Specifically, the water pump is started to maintain fluid circulation to equalize the temperature difference between the battery cells, but the compressor is stopped, relying solely on natural heat dissipation or fluid thermal inertia for equalizing temperature, greatly reducing energy consumption.

[0092] Mode 3 (Precise Response Mode): In rapid response scenarios and under secondary conflict (equilibrium state), the system needs to cope with frequent power fluctuations. In this case, the first liquid cooler unit is controlled to maintain basic power operation, for example, at a constant low frequency, providing a stable cooling pool; the second liquid cooler unit is controlled to perform variable frequency response, quickly adjusting its output to follow power fluctuations; simultaneously, the opening of the second valve for high-power battery clusters is adjusted to cool only the hottest battery clusters, avoiding indiscriminate overcooling.

[0093] Step S4: Pre-regulation step based on power mutation prediction.

[0094] It should also be noted that traditional feedback control has an inherent lag, meaning that cooling only begins after the temperature has risen. To overcome this drawback, this embodiment, in addition to the aforementioned control loop, also independently runs a feedforward control loop based on time series prediction, namely a pre-adjustment step based on power surge prediction.

[0095] Specifically, this step includes: First, predicting the power change rate over a predetermined time period, such as the next 5-15 minutes, based on the power demand curve of the power grid dispatch and a Long Short-Term Memory (LSTM) network. The LSTM network input layer receives historical power data, weather forecast data, and the current dispatch plan, while the output layer provides the power prediction value for the future time. and its rate of change .

[0096] like Figure 4 As shown, the power mutation prediction curve and confidence interval diagram based on the Long Short-Term Memory network clearly reveal the core timing logic of the feedforward control strategy in this invention. The horizontal axis in the figure represents the time process, covering the typical scheduling period from 2 PM to 3 PM; the vertical axis represents the charging and discharging power of the system, in kilowatts.

[0097] The solid blue line represents the actual load dispatching instructions issued by the power grid. It can be seen that before 2:30 PM, the system was operating at low power, but at that moment it suddenly experienced a step increase, entering a full-power discharge state. The dashed red line represents the predicted power curve generated based on the Long Short-Term Memory network.

[0098] It is worth noting that the prediction curve begins to rise at 14:15, which is 15 minutes before the actual load change, forming a clear lead time window. It is by utilizing this time difference that the system can start the liquid chiller unit in advance for pre-cooling operation, thereby physically eliminating the thermal lag of the thermal management system.

[0099] Furthermore, the light red shaded area surrounding the prediction curve represents the confidence interval of the prediction model, and the width of this interval reflects the degree of uncertainty in the prediction. A narrower interval during the power plateau period indicates high prediction reliability; while a significantly wider interval during the power ramp-up period suggests that the system should perform secondary verification using a load confirmation time window mechanism at this time.

[0100] Secondly, set a mutation threshold. Used to identify normal fluctuations and abrupt changes in operating conditions:

[0101] Operating Condition 1: If the predicted power change rate exceeds the abrupt change threshold ( Furthermore, the discharge is a sudden increase in power, meaning the power jumps dramatically from low to high, and the system determines that a large amount of Joule heat is about to be generated. At this point, before the heat is generated, the flow rate of the first liquid cooler is increased to the target value (pre-cooling flow rate), and the second liquid cooler is controlled to pre-cool the battery clusters (i.e., high-power battery clusters) that will soon operate at high power. This cooling reserve strategy utilizes the specific heat capacity of the battery and coolant, essentially creating a thermal buffer zone that effectively reduces the temperature peak at the moment of power change.

[0102] Operating Condition 2: If the predicted power change rate exceeds the abrupt change threshold and there is a sudden drop in power, i.e., a sudden disconnection from high load or a switch to low load, a large amount of residual heat, i.e., thermal inertia, still exists inside the battery. To prevent energy waste due to overcooling or stress damage caused by a sudden drop in temperature, the system reduces the pump speed of the first liquid cooling unit in advance and gradually reduces the flow rate; and controls the second liquid cooling unit to maintain low power operation, so as to eliminate residual hot spots in a gentle way and achieve a smooth transition of thermal state.

[0103] By organically combining the above steps S1 to S4, this embodiment constructs a closed-loop, forward-looking thermoelectric synergistic control system.

[0104] For example, to further illustrate the technical effects of this embodiment, we applied the above method to a liquid-cooled energy storage power station configured with 2MWh for verification.

[0105] During a typical summer afternoon peak shaving and valley filling operation, the ambient temperature reached 38℃. At 14:00, the system was in a valley power period, identified as a cooling maintenance scenario, with a conflict level of two. The system allocated high weights to temperature safety and energy consumption, controlling the two generators to enter a low-power self-circulation mode, keeping the maximum temperature difference between the batteries within 2℃. At 14:15, the LSTM network predicted that a large scheduling command would be issued at 14:30, corresponding to a power surge.

[0106] The system immediately triggered a pre-adjustment step, with the first liquid cooling unit increasing the flow rate to 80% 15 minutes in advance, pre-cooling the average battery temperature from 28°C to 24°C. At 14:30, the power grid issued a full-power discharge command, which the system determined to be a power-priority scheduling scenario, resulting in a level one conflict.

[0107] Due to pre-cooling, the battery temperature rise slope is significantly gentler. At this point, the dynamic weighting favors power tracking, with the first unit operating at full load and the second unit assisting, and all valves fully open. At 15:00, the discharge ends, and the power drops sharply. The system recognizes this situation and quickly reduces the pump speed of the first unit, utilizing the low-power operation of the second unit to remove residual heat, thus avoiding the overcooling phenomenon commonly seen in traditional control systems.

[0108] like Figure 5As shown in the simulation diagram, the comparison of battery temperature rise under the conventional control and the collaborative control of the present invention under the power sudden change condition intuitively demonstrates the superiority of the present invention in dealing with extreme heat loads.

[0109] Figure 5 The horizontal axis represents time, covering the typical operating period from 2 PM to 3 PM; the vertical axis represents the highest monitored temperature within the battery cluster.

[0110] The red dashed line represents the temperature change trajectory using a traditional hysteresis-based temperature control strategy. As can be seen, after the power grid issued a full-power discharge command at 14:30, due to the lack of proactive intervention, the red curve showed a steep upward trend, quickly approaching or even exceeding the system's high-temperature alarm boundary, revealing the safety hazards caused by thermal hysteresis.

[0111] In contrast, the solid blue line represents the temperature trajectory after applying the collaborative control strategy of this invention. The most significant feature is that, starting from 14:15, the curve shows a clear active decline, which corresponds to the system described in the embodiment starting the liquid cooling unit in advance to pre-cool the battery based on the prediction results of the long short-term memory network, thereby reducing the average battery temperature from 28 degrees Celsius to 24 degrees Celsius.

[0112] Thanks to this 4-degree Celsius cooling buffer, when the power surge occurred at 14:30, the upward slope of the blue curve was significantly smoother, and the final peak temperature was firmly locked within a safe range. This fully demonstrates that the present invention successfully mitigates the threat of transient power surges to battery thermal safety through timing-based peak shaving and valley filling.

[0113] In summary, this embodiment successfully resolved the contradiction between heat and electricity in liquid-cooled energy storage systems under complex operating conditions by introducing multi-dimensional feature fusion, conflict level determination, three-dimensional model prediction, and dynamic weight allocation mechanisms. In particular, through the heterogeneous collaboration of the first and second liquid-cooled units and the precise coordination of valves, it not only ensured power support capability under extreme operating conditions but also significantly reduced cooling energy consumption throughout the entire life cycle, achieving an optimal balance between safety, economy, and functionality.

[0114] Example 2:

[0115] This embodiment, based on the multi-objective optimization framework described in Embodiment 1, further introduces constraint relaxation and motion smoothing steps for extreme operating conditions. This mechanism is mainly used to address the dilemma when the system faces extremely harsh power grid commands or extremely severe environmental conditions, making it impossible to find a feasible control solution under conventional hard constraints, and to provide protection logic designed to prevent high-frequency control commands from causing physical damage to mechanical components.

[0116] Specifically, in the process of selecting the optimal control strategy in the multi-objective Pareto optimization front based on the matching results, the method in this embodiment does not directly output the calculation results. Instead, it embeds a complete closed-loop processing flow that includes feasibility determination, soft constraint transformation, re-optimization search, and action smoothing. This flow ensures that the system can maintain continuous operation in a costly but acceptable manner when faced with tasks that are almost impossible to complete, rather than directly reporting errors and shutting down.

[0117] Step 1: Calculate the volume of the feasible solution space at the current moment and determine the mode.

[0118] First, the system needs to perform a mathematical quantitative assessment of the difficulty of solving the multi-objective optimization problem at the current moment. In the multi-dimensional space of control variables, all physical constraints together enclose a polyhedral region, which is the feasible solution space.

[0119] Specifically, calculating the feasible solution space volume at the current moment refers to the measure of the solution set that simultaneously satisfies all hard numerical boundaries within a multi-dimensional state space comprised of the power of the first liquid-cooled unit, the power of the second liquid-cooled unit, the valve opening degree, and the allowable power of the battery. For ease of real-time calculation, the controller can estimate this volume using the Monte Carlo sampling method or the hyperplane cutting method. We define this feasible solution space volume as... .

[0120] It should also be noted that under normal operating conditions, the constraints in various dimensions are relatively loose. The values ​​are usually large, indicating that the system has ample adjustment mechanisms. However, under extreme conditions, such as when the grid requires full power output while the battery temperature is nearing the alarm threshold, the power demand necessitates increased output, while temperature safety requirements limit output. The constraints formed by these two factors rapidly approach or even overlap, leading to... The contraction is rapid and approaches zero.

[0121] Therefore, the system sets a preset convergence threshold. If the calculated volume of the feasible solution space... Less than the preset convergence threshold This indicates that the current physical constraints are extremely tight, and conventional optimization algorithms are highly likely to get stuck in local deadlock or directly return to an empty set. At this point, the system determines to enter a constraint relaxation mode. This decision logic gives the control system the ability to sense the crisis, which is a prerequisite for achieving flexible control.

[0122] Step 2: Constructing the soft constraint function under the constraint relaxation mode

[0123] Once the constraint relaxation mode is entered, the control logic undergoes a fundamental shift. At this point, the system no longer adheres to absolute safety boundaries but seeks a compromise with manageable risks.

[0124] Specifically, in the constraint relaxation mode, the temperature safety range and the power adjustment rate constraints are transformed from hard numerical boundaries into soft constraint functions with penalty weights.

[0125] In normal mode, the temperature safety range is a hard constraint, such as the battery's maximum temperature. It must be less than or equal to the cutoff temperature Mathematically, this is a step function, and exceeding its limits results in an illegal solution. In this step, we transform this hard constraint into a soft constraint function. The soft constraint function allows the predicted value to exceed the hard numerical boundary within a preset short-term thermal capacity window, but the penalty evaluation value increases with the magnitude of the exceedance.

[0126] The preset short-term heat capacity time window mentioned here is based on the physical characteristic that the battery itself has a large specific heat capacity. A battery is a large thermally inertial body; even if the coolant temperature is momentarily insufficient or the power is momentarily overloaded, the internal temperature of the battery will not experience a transient jump, but rather exhibit a slow, integral-like increase. Therefore, allowing short-term parameter deviations is physically safe.

[0127] For example, to accurately describe this process, we define a temperature penalty function under soft constraints. Assume the current predicted temperature is... Hard boundary is Then the penalty function can be constructed as a non-linear exponential growth form:

[0128] ;

[0129] in, Basic penalty coefficient, This is the sensitivity factor. The formula indicates that when predicting temperature... Less than or equal to At that time, the penalty value is extremely small or zero; but once Exceed Punishment evaluation value It will rise exponentially.

[0130] like Figure 6 As shown in the diagram, the evolution of the soft constraint penalty function under the constraint relaxation mode vividly reveals the mathematical mechanism by which the present invention transforms rigid physical boundaries into elastic control targets.

[0131] Figure 6The horizontal axis represents the system's predicted maximum battery temperature, covering the critical range of 40 to 50 degrees Celsius; the vertical axis represents the penalty evaluation value in the optimization algorithm, reflecting the system's degree of dissatisfaction with the current operating state. The vertical black solid line in the figure is marked at 45 degrees Celsius, representing the system's set hard safety boundary.

[0132] As can be seen, within the safe zone to the left of 45 degrees Celsius, the penalty value remains zero for both high-sensitivity and low-sensitivity curves, meaning that the system operates in full compliance within this range without the need for additional control costs.

[0133] Once the predicted temperature exceeds 45 degrees Celsius and enters the over-limit zone on the right, both curves exhibit a significant exponential growth characteristic. The solid red line represents the penalty trajectory under the high sensitivity setting, with an extremely steep upward slope, indicating that the system has a very low tolerance for over-temperature behavior. Even a small temperature deviation will result in a huge penalty, forcing the controller to quickly adjust power. The dashed blue line represents the penalty trajectory under the low sensitivity setting, with a relatively gentle rise, meaning that under extreme conditions, the system allows the battery temperature to exceed the safety boundary significantly in a short period to ensure the execution of critical grid commands. This nonlinear function evolution characteristic explains how this invention, in the face of deadlock risk, achieves an intelligent switch from rigid protection to flexible support by dynamically adjusting the sensitivity factor.

[0134] Similarly, we can define a similar soft constraint for the power regulation rate constraint. Assume the required power change rate of the grid is... The maximum physical ramp rate allowed by the system hardware is... ,when Greater than In normal operation, hard enforcement can overload the device. However, in relaxed mode, we allow short-term overclocking at the expense of device lifespan, defined by a rate penalty function. for:

[0135] ;

[0136] in, This is the weighting coefficient for rate violations.

[0137] Through this transformation, the original yes / no question (i.e. whether it is feasible) becomes a calculation question (i.e. how much cost), thus ensuring that the optimization problem always has a solution.

[0138] Step 3: Rerun the multi-objective optimization to search for a compromise strategy.

[0139] After reconstructing the constraints, the system needs to find the optimal path in this expanded solution space.

[0140] Specifically, this step involves rerunning the multi-objective optimization to search for the compromise control strategy with the lowest total penalty evaluation value in the solution space containing the soft constraint function. The optimization objective function at this point is no longer simply power, temperature, energy consumption, and response as in Example 1, but includes the aforementioned penalty term. The new overall objective function... This can be expressed as the sum of the original multi-objective weighted sum and the penalty term:

[0141] ;

[0142] in, and These represent the weights of each dimension and the objective function value mentioned in Example 1, respectively.

[0143] Optimization algorithms, such as particle swarm optimization or genetic algorithms, will iteratively search within the new solution space. Due to the introduced penalty mechanism, the algorithm automatically avoids regions with severe overheating or overspeeding, as the penalty values ​​there are too high; simultaneously, the algorithm will avoid completely ignoring power grid commands, as that would worsen the objective function of the power grid response dimension. Ultimately, the algorithm will converge to a point with the lowest total penalty evaluation value.

[0144] For example, this compromise control strategy might manifest as allowing the battery temperature to temporarily reach 46 degrees Celsius (exceeding the conventional 45-degree Celsius limit) for the next two minutes in exchange for the full execution of a primary frequency regulation command to the grid, but then forcing full-power cooling to bring the temperature back to a safe level after two minutes. This strategy is impossible to generate under strict hard constraints, but it is crucial for ensuring grid stability.

[0145] Step 4: Smooth the movement of the actuator.

[0146] While the control strategy obtained through the above calculations is mathematically optimal, it may present numerical abrupt changes in its engineering and physical implementation. For example, the algorithm might require the compressor speed of the liquid-cooled unit to be 100% in the first second, drop to 20% in the second second, and then rise to 90% in the third second. Such drastic fluctuations not only cause severe mechanical shocks to actuators such as compressors, water pumps, and electric valves, shortening equipment lifespan, but also cause oscillations and cavitation in fluid pipelines.

[0147] Therefore, when selecting the optimal control strategy, a motion smoothing step must be included.

[0148] Specifically, the system calculates in real time the difference between the actuator action quantity corresponding to the candidate control strategy and the actual action quantity at the previous moment. Let the candidate valve opening given by the optimization at the current moment be... The valve opening degree actually executed in the previous control cycle was... The difference between the two is for:

[0149] ;

[0150] Then, the difference is compared with a preset mechanical response threshold. The mechanical response threshold is set based on the physical characteristics of the actuator; for example, the maximum operating rate of an electric control valve may be limited to changing the opening by 10% per second.

[0151] If the difference Exceeding the mechanical response threshold If the current candidate strategy is deemed too aggressive, a damping correction needs to be applied to the candidate control strategy.

[0152] It's also important to note that damping correction is not simply truncating commands, but rather generating new commands based on first-order inertial elements or rate-limiting logic. The final control strategy outputs a smoothed result. The calculation logic is as follows: If Greater than And if the difference exceeds the limit, then:

[0153] ;

[0154] if Less than And if the difference exceeds the limit, then:

[0155] ;

[0156] If the difference does not exceed the limit, then it will be adopted directly:

[0157] ;

[0158] In this way, the control commands are smoothed in the time domain, making the power regulation curve of the liquid chiller and the valve opening change curve exhibit continuous and gentle characteristics, avoiding abrupt changes.

[0159] In summary, this embodiment successfully solves the control challenge of liquid-cooled energy storage systems facing both high safety risks and high task requirements by introducing a solution space volume monitoring mechanism, intelligently switching to constraint relaxation mode under extreme conditions, replacing rigid hard boundaries with soft constraint functions with penalty weights, and combining motion smoothing processing based on mechanical response characteristics. This solution not only improves the system's survivability and online rate under extreme conditions but also effectively extends the service life of key mechanical components, demonstrating an efficient solution approach that combines thermal management and power coordinated control.

[0160] Example 3:

[0161] Building upon Examples 1 and 2, this embodiment further elaborates on the deep execution logic and safety protection mechanism for the pre-regulation steps based on power mutation prediction. In the actual operation of liquid-cooled energy storage systems, while AI-based power prediction can provide a forward-looking scheduling perspective, it also faces the risks of energy waste due to false alarms and condensation short-circuit risks caused by changes in ambient temperature and humidity. To address these two challenges, this embodiment proposes a composite control process that includes environmental physical constraint perception and a two-level hierarchical confirmation mechanism: asynchronous speculative control for false alarms and dew point safety protection.

[0162] Specifically, this process treats environmental safety boundaries as an inviolable red line, uses the heterogeneous characteristics of the dual liquid-cooled units as a tactical execution tool, and achieves effective management of future uncertainties through precise timing control. The following will elaborate on each stage of this process in logical execution order.

[0163] Phase 1: Dew point safety protection based on environmental perception.

[0164] It's also important to note that in liquid-cooled thermal management systems, lower coolant temperatures are not always better. When the fluid temperature is below the ambient dew point, condensation can easily form on the cold plate surfaces and pipe connections inside the battery pack. For high-voltage DC systems, condensation can lead to a decrease in insulation resistance and even short-circuit sparking. Therefore, before implementing any pre-conditioning actions aimed at reducing battery temperature, the system must first establish a safety firewall based on the laws of physics.

[0165] Specifically, before performing the pre-adjustment, the system activates a real-time environmental monitoring mechanism. Temperature and humidity sensors distributed around the air inlet of the energy storage container and the battery clusters collect ambient temperature and humidity data in real time and calculate the dew point temperature. To ensure that the calculation results can adapt to a wide range of climate changes and possess extremely high accuracy, this embodiment uses a modified Magnus formula for real-time calculation.

[0166] For example, to avoid confusion in symbol definition with the scene labels or optimization coefficients appearing in the foregoing embodiments, we redefine the empirical constants in the Magnus formula as... and Set constants The value is 17.27, and a constant is set. The value is 237.7 degrees Celsius. The system reads the current dry-bulb temperature. and relative humidity The calculation logic for dew point temperature is as follows:

[0167] First, calculate the dimensionless intermediate variable. This variable characterizes the natural logarithmic property of saturated water vapor pressure, and its calculation formula is as follows:

[0168] ;

[0169] Furthermore, this intermediate variable is utilized. Inverse calculation of dew point temperature :

[0170] ;

[0171] Optionally, considering the inherent measurement drift of the sensor, the potential for localized humidity buildup within the battery pack's internal microenvironment, and the temperature non-uniformity of the coolant during pipeline transport, directly using the calculated dew point temperature as the control boundary still poses certain safety risks. Therefore, this embodiment introduces an additional safety buffer, setting the minimum permissible inlet temperature of the coolant as the control boundary. Set as the dew point temperature With the preset safety margin of anti-gelling lotion The sum of. Its mathematical expression is:

[0172] ;

[0173] Under normal circumstances, the preset anti-condensation safety margin It can be set to 2 to 3 degrees Celsius, which is sufficient to cover the vast majority of measurement errors and environmental fluctuations.

[0174] Specifically, when the system's power prediction module determines that a high-power discharge mutation is about to occur, and calculates the pre-cooling target fluid temperature required to suppress the temperature rise... At that time, the control algorithm will immediately perform a safety check. If the calculated target fluid temperature... Below the minimum permissible inlet temperature If so, it is determined that the current pre-cooling requirement conflicts with the safety of preventing condensation.

[0175] In this situation, the system triggers a forced clamping mechanism, ignoring the cooling requirements at the thermal management level, and forcibly clamps the temperature control setpoints of the first liquid chiller and the second liquid chiller to the minimum allowable inlet temperature. This logic ensures that the system always operates above the dew point, completely eliminating the risk of condensation.

[0176] like Figure 7 As shown in the figure, this diagram intuitively illustrates the dynamic response process and security protection mechanism of the present invention in dealing with the risk of predicted false alarms.

[0177] Figure 7The horizontal axis represents the time process, and the vertical axis represents the temperature value. Among them, the gray dashed line represents the ambient dew point temperature that changes slightly with the environment; the red solid line represents the minimum allowable inlet temperature boundary calculated by the system based on the dew point temperature and the preset anti-condensation safety margin, which constitutes the safety red line of the liquid cooling system; the blue solid line represents the actual inlet temperature change trajectory of the coolant.

[0178] As can be seen, at the tenth second, the system initiated an asynchronous speculative pre-cooling strategy based on power mutation prediction. The coolant temperature began to drop rapidly, and when it approached the red safety threshold, it automatically triggered the forced clamping logic, ensuring that the system always operated above the minimum allowable inlet temperature boundary, thereby physically eliminating the risk of condensation.

[0179] Subsequently, at the 25th second, since the load confirmation time window expired and no actual power surge was detected, the system determined it to be a false alarm. At this time, the blue curve did not rebound sharply, but instead showed a smooth upward trend, which corresponds to the damping back-off process, that is, controlling the second liquid cooling unit to gently return to steady state at a preset damping rate, thereby effectively avoiding hydraulic shock and extending equipment life.

[0180] Phase Two: Activation of the Asynchronous Speculative Pre-cooling Strategy.

[0181] After confirming the safe boundary of the fluid temperature lower limit, the next challenge the system faces is the uncertainty of prediction. If the prediction model indicates a high-power surge in 5 minutes, but the power grid does not actually issue a command, resulting in a false alarm, blindly starting both liquid-cooled units simultaneously for full-power precooling at this time will lead to huge wasted energy consumption and increased compressor start-up and shutdown wear. To address this, this embodiment designs a startup strategy that combines speculative and asynchronous approaches.

[0182] Specifically, asynchronous speculative precooling means that at the initial prediction moment when a power surge is determined, the system does not immediately mobilize all cooling resources, but only controls the second liquid chiller to increase the flow rate and adjusts the opening of the second valve to a preset value, while keeping the operating state of the first liquid chiller unchanged.

[0183] For example, the engineering logic behind this strategy lies in fully utilizing the heterogeneous characteristics of the two units. The first liquid-cooled unit, as the high-power main refrigeration unit, has a large compressor capacity, high starting current surge, and high thermal inertia, making it unsuitable for frequent start-stop operations or as an exploratory load. In contrast, the second liquid-cooled unit, as a low-power auxiliary unit, is typically equipped with a more sensitive inverter driver, offering faster response and lower energy consumption. Therefore, the system selects the second liquid-cooled unit as the pioneer unit.

[0184] During this phase, the control unit instructs the second liquid cooler unit to rapidly increase its pump speed and widen the opening of the second valve on the branch corresponding to the battery clusters that will soon bear high power loads. Through this operation, the system utilizes the existing coolant in the pipeline for accelerated circulation, providing initial, minor cooling to the batteries. Simultaneously, the first liquid cooler unit remains in standby or low-load maintenance mode, remaining inactive. This speculative behavior has extremely low costs: if the prediction is accurate, the second unit's advance operation has pre-set a high-flow-rate state for the system, paving the way for the subsequent intervention of the first unit; if the prediction fails, the system only loses a small amount of electrical energy from the short-term operation of the second unit, avoiding unnecessary startup of the main compressor.

[0185] Phase 3: Actual verification based on load confirmation time window.

[0186] While the second liquid-cooled unit was started up for pre-cooling, the system entered a highly alert testing and verification phase. At this point, the control system no longer relied solely on predictive models trained from historical data, but instead turned to millisecond-level real-time current sampling data to capture substantial evidence of power surges.

[0187] Specifically, the system begins monitoring the rate of change of the real-time current of the battery cluster over time and sets a load confirmation time window. This load confirmation time window... This is a short time interval, for example, set to 10 seconds, starting from the predicted onset of the mutation. During this period, the control device acquires current data from the battery side at high frequency. And by combining the differential algorithm and the moving average filtering technique, the current rate of change is calculated. .

[0188] Optionally, the system presets a measured trigger threshold. This threshold represents the current rise slope that the system must meet for a power surge to be recognized. Only when the measured rate of current change exceeds this threshold does the system consider the grid dispatch command to have been truly issued. At this stage, the control logic is divided into two distinct execution branches based on the measured results:

[0189] Branch 1: Predictive Confidence and Full Power Response

[0190] If in the load confirmation time window Within, the system detected the rate of change of the current. Exceeding the measured trigger threshold If the predicted power surge has materialized, it indicates that the predicted power surge has become a reality. At this point, the system determines that the prediction is certain.

[0191] Specifically, once the determination is confirmed, the system immediately ends the speculative testing state and switches to full-power cooling mode. The control device immediately controls the first liquid cooling unit to execute the action of increasing the flow rate to the target value. The first liquid cooling unit quickly unlocks its standby lock, and its compressor frequency and water pump speed are synchronously increased to the pre-cooling setpoint determined by thermodynamic calculations. At this time, since the second liquid cooling unit has already established a circulating flow field, the cooling energy injected by the first liquid cooling unit can be delivered to the battery terminal more quickly. The two units work together to clamp the upcoming current thermal shock with maximum cooling flux, ensuring that the battery temperature is always maintained within a safe range.

[0192] Branch 2: Predicting False Alarms and Damped Backoff

[0193] If in the load confirmation time window After exhaustion, the system still did not detect the rate of change of current. Exceeding the measured trigger threshold This indicates that the predicted power surge did not occur as expected. This could be due to temporary changes in the grid dispatch center's plans, communication link delays, or occasional misjudgments in the prediction model.

[0194] Specifically, the system determines this to be a false alarm. To eliminate the system state altered by speculative operations and to avoid hydraulic shock to the pipeline network due to sudden changes in action, the control device executes a damping retraction strategy. The system controls the second liquid chiller and the second valve to retract to their pre-adjustment state at a preset damping rate.

[0195] For example, the damping rate refers to the smoothness limit of the actuator's resetting action. If the pump of the second liquid cooler unit is shut off directly or the valve is suddenly closed, the kinetic energy of the high-speed fluid flow will be instantly converted into pressure potential energy, causing a severe water hammer effect, which may lead to loosening of pipe joints or damage to valve cores. Therefore, the control system smoothly restores the operating condition of the second liquid cooler unit to the steady-state value before pre-adjustment by following a preset linear descent rate, such as a 5% decrease in pump speed per second and a 2-degree decrease in valve opening per second. This soft-landing mechanism ensures that even under fluctuating operating conditions with frequent false alarms, the mechanical components of the system are protected from shock, thereby significantly extending the service life of the equipment.

[0196] In summary, this embodiment introduces... , The Magnus formula with constant correction establishes a precise dew point sensing system, solving the safety problem of condensation during low-temperature precooling of liquid cooling systems. At the same time, by creatively utilizing the heterogeneous capabilities of dual liquid cooling units, an asynchronous speculative control strategy is implemented, and combined with the load confirmation time window and damping backoff mechanism, the contradiction between the uncertainty of artificial intelligence prediction and the reliability of industrial control is perfectly resolved.

[0197] This entire process constitutes the core defense system of liquid-cooled energy storage systems in response to complex power grid dispatching environments, ensuring thermal safety under extreme operating conditions and achieving optimal energy efficiency and equipment protection throughout the entire life cycle.

[0198] Example 4:

[0199] This embodiment provides a thermal management and power coordination control system for a liquid-cooled energy storage system, configured to execute the method described in any one of embodiments one to three above. Addressing the technical challenge in existing technologies where the battery thermal management system and power dispatch system are disconnected, leading to an inability to simultaneously consider the triangular conflict between grid response speed, battery thermal safety, and system cooling energy consumption, the system proposed in this embodiment achieves a leap from passive defense to active collaborative optimization through a highly integrated modular design.

[0200] Specifically, such as Figure 8 As shown, this thermal management and power co-control system relies on a high-performance industrial control computer or embedded edge computing gateway at the physical level, and mainly comprises three core functional units at the logical level: a data processing module, a decision-making module, and an execution control module. These three modules interact with each other through a high-speed internal bus or shared memory mechanism, together forming a closed-loop intelligent control entity.

[0201] First, this system includes a data processing module. This module serves as the sensing hub of the entire control system, establishing real-time communication connections with the battery management system (BMS), the power conversion system (PCS), and the grid dispatch terminal. The data processing module is configured to collect multi-dimensional operational data, encompassing everything from microscopic data such as cell temperature, voltage, and state of charge, to macroscopic data such as ambient temperature and humidity, grid frequency deviation, and real-time electricity price information.

[0202] Unlike the background technology which only monitors a single temperature indicator, the data processing module in this embodiment incorporates a multi-source heterogeneous data fusion engine to perform feature fusion of grid dispatching requirements and battery thermal state. This module is not merely a data transporter, but also an information extractor. Based on the calculation logic described in Embodiment 1, it calculates key characteristic parameters in real time, such as power demand difference, temperature safety margin, cooling energy consumption ratio, and grid response deviation. Based on these quantitative indicators, the data processing module can intelligently identify the main contradictions currently facing the system, thereby determining the conflict level and grid dispatching scenario. For example, when the module identifies a peak electricity price and a high-priority dispatching signal, it marks the scenario as a power-priority dispatching scenario; when it identifies a frequency regulation command and severe heat accumulation, it marks the conflict level as a level three conflict requiring emergency intervention. This processing provides structured state input for subsequent accurate decision-making, solving the problems of vague perception and slow response in traditional systems.

[0203] Secondly, this system includes a decision module. This module is the core of the system and is typically deployed in a digital signal processor (DSP) or graphics processing unit (GPU) with floating-point computing capabilities. The decision module is configured to run a three-dimensional predictive model of the thermoelectric power grid, which deeply couples the electrochemical heat generation mechanism, the hydrodynamic heat dissipation characteristics, and the power grid dispatch response model.

[0204] Based on the conflict level and power grid dispatch scenario provided by the data processing module, the decision module performs multi-objective optimization calculations within the multi-dimensional solution space. This module integrates a multi-objective Pareto optimization algorithm, capable of finding the optimal balance among four mutually constraining dimensions—power point tracking accuracy, temperature safety boundary, cooling energy consumption cost, and power grid response delay—within a millisecond-level time window. Furthermore, this module also incorporates the constraint relaxation logic described in Embodiment 2. When it is determined that the current operating condition is extreme, causing the feasible solution space to shrink, it can automatically transform hard safety constraints into soft constraints with penalty functions, preventing the system from crashing due to a lack of solutions. Finally, the decision module outputs dynamically allocated weights for the four dimensions of power point tracking, temperature safety, cooling energy consumption, and power grid response. These weight values ​​essentially constitute the system's tactical guidelines at the current moment, determining whether the system aggressively pursues power output, conservatively implements thermal protection, or economically operates in a low-power mode.

[0205] Finally, this system includes an execution control module. This module acts as the system's execution mechanism, responsible for translating abstract decision weights into concrete physical drive signals. The outputs of the execution control module are connected to the frequency converter of the first liquid-cooled unit, the drive controller of the second liquid-cooled unit, the actuator of the valve assembly, and the power controller of the battery cluster, respectively.

[0206] The execution control module is configured to generate control signals based on the dynamically allocated weights, adjusting the operating power of the first liquid chiller unit, the operating power of the second liquid chiller unit, the opening degree of the valve assembly, and the output power of the battery cluster, respectively. In specific implementation, this module executes the complex timing control strategy detailed in Embodiment 3. For example, when the decision module issues a command with a high power response and thermal risk, the execution control module generates a coordinated operation command, driving the first liquid chiller unit to run at full speed to provide a baseline cooling capacity, while finely adjusting the opening degree of the second liquid chiller unit and specific valves to eliminate local hot spots. In addition, the module also incorporates a motion smoother and an anti-condensation safety lock. Upon receiving a feedforward signal based on power mutation prediction, it can execute asynchronous speculative control, first starting the second liquid chiller unit for pipeline pre-cooling, and then activating the main cooling capacity after confirming the load. This completely eliminates thermal hysteresis at the physical execution level and avoids energy waste caused by over-powering.

[0207] In summary, the thermal management and power coordination control system for the liquid-cooled energy storage system provided in this embodiment constructs a complete electromechanical-thermal integrated solution through global perception of the data processing module, intelligent optimization of the decision-making module, and precise coordination of the execution control module. This system not only supports efficient collaboration between heterogeneous dual liquid-cooled units in its hardware architecture but also achieves flexible adaptation to complex grid dispatch commands in its software logic, significantly improving the survivability and economic benefits of energy storage power stations in a source-grid-load-storage interactive environment.

[0208] Example 5:

[0209] Corresponding to the above embodiments, the present invention also proposes an electronic device.

[0210] like Figure 9 The diagram shows a structural schematic of an electronic device according to the present invention. The electronic device 100 includes a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, via a bus 102. Optionally, the electronic device 100 may further include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one unit, and the structure of this electronic device 100 does not constitute a limitation on the embodiments of the present invention.

[0211] Processor 101 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0212] Bus 102 may include a pathway for transmitting information between the aforementioned components. Bus 102 may be a PCI bus or an EISA bus, etc. Bus 102 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0213] The memory 103 stores a computer program corresponding to the thermal management and power coordination control method of the liquid-cooled energy storage system in the above embodiments of the present invention. This computer program is controlled and executed by the processor 101. The processor 101 executes the computer program stored in the memory 103 to implement the content shown in the aforementioned method embodiments.

[0214] Among them, electronic devices 100 include, but are not limited to: mobile terminals such as laptops and PADs (tablet computers) and fixed terminals such as desktop computers. Figure 9 The electronic device 100 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0215] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A thermal management and power coordination control method for a liquid-cooled energy storage system, characterized in that, In a control device used in a liquid-cooled energy storage system, the liquid-cooled energy storage system includes a first liquid-cooled unit, a second liquid-cooled unit connected in parallel, and a valve assembly including a first valve and a second valve; The method includes: responding to the collected multi-dimensional operational data, performing feature fusion on power grid dispatching requirements and battery thermal status to determine the conflict level and power grid dispatching scenario at the current moment; Based on the conflict level and the power grid dispatch scenario, a three-dimensional prediction model of the thermal power grid is invoked for multi-objective optimization to calculate the dynamic allocation weights of four dimensions: power point tracking, temperature safety, cooling energy consumption, and power grid response. Based on the dynamically allocated weights, coordinated operation instructions and battery power adjustment instructions are generated for the first liquid-cooled unit and the second liquid-cooled unit to perform phased control of the liquid-cooled energy storage system.

2. The method of claim 1, wherein, The system responds to the collected multi-dimensional operational data by performing feature fusion on grid dispatching requirements and battery thermal status to determine the current conflict level and grid dispatching scenario, including: Calculate the power demand difference, temperature safety margin, cooling energy consumption ratio, and grid response deviation; Based on the preset value range of the power demand difference, the temperature safety margin, the cooling energy consumption ratio, and the grid response deviation, the conflict level is determined as a first-level conflict with power priority, a second-level conflict in a balanced state, or a third-level conflict with temperature safety priority.

3. The method of claim 1, wherein, In determining the conflict level and power grid dispatch scenario at the current moment, the judgment logic for the power grid dispatch scenario includes: When the peak-valley electricity price signal is detected as peak and the scheduling priority is high, it is determined to be a power priority scheduling scenario, and the upper limit of the temperature safety range and the threshold of cooling energy consumption ratio are relaxed. When the peak-valley electricity price period signal is detected as being in the valley segment and the dispatch priority is identified as low, it is determined to be a cooling maintenance scenario, and the temperature safety margin and cooling energy consumption ratio thresholds are tightened. When a frequency modulation response command is received and the required power adjustment rate is less than or equal to a preset rate threshold, it is determined to be a fast response scenario.

4. The method of claim 1, wherein, The multi-objective optimization of the three-dimensional prediction model of the thermal power grid includes: Using the thermoelectric coupling model in the three-dimensional prediction model of the thermoelectric grid, the battery temperature field change is predicted based on the thermal diffusion characteristics and cooling efficiency characteristics; Using the grid response prediction sub-model in the three-dimensional prediction model of the thermal power grid, the deviation between the power adjustment rate constraint and the grid response is predicted based on the dispatch power demand input; The conflict level is matched with the power grid dispatch scenario, and the optimal control strategy is selected in the multi-objective Pareto optimization front based on the matching result.

5. The method according to claim 1, characterized in that, The calculated dynamic weight allocation for four dimensions—power point tracking, temperature safety, cooling energy consumption, and grid response—includes: If the scenario is determined to be the power-priority scheduling scenario, the maximum weight is assigned to the power tracking dimension, and the second largest weight is assigned to the temperature safety dimension. If the scenario is determined to be the cooling maintenance scenario, the maximum weight is assigned to the temperature safety dimension, and the second largest weight is assigned to the cooling energy consumption dimension. If the scenario is determined to be a rapid response scenario, the weight of the power grid response dimension is increased to be higher than the weight of the cooling energy consumption dimension.

6. The method according to claim 1, characterized in that, The step of generating coordinated operation instructions for the first liquid-cooled unit and the second liquid-cooled unit based on the dynamically allocated weights includes: In a scenario of primary conflict and power priority scheduling, the first liquid chiller unit is controlled to operate at rated power, the second liquid chiller unit is controlled to operate at auxiliary power, and the first valve and the second valve are opened. In a Level 3 conflict and cooling maintenance scenario, control the first liquid cooling unit and the second liquid cooling unit to enter self-circulation mode, start the water pump and stop the compressor; In a fast-response scenario and under secondary conflict, the first liquid cooling unit is controlled to maintain basic power operation, the second liquid cooling unit is controlled to perform frequency conversion response, and the opening degree of the second valve for the high-power battery cluster is adjusted.

7. The method according to claim 4, characterized in that, The process of selecting the optimal control strategy in the multi-objective Pareto optimization front based on the matching results also includes constraint relaxation and motion smoothing steps for extreme conditions, specifically including: Calculate the volume of the feasible solution space at the current moment. If the volume of the feasible solution space is less than a preset convergence threshold, then determine to enter the constraint relaxation mode. In the constraint relaxation mode, the temperature safety range and the power adjustment rate constraints are transformed from hard numerical boundaries into soft constraint functions with penalty weights. The soft constraint function allows the predicted value to exceed the hard numerical boundary within a preset short-time heat capacity time window, but the penalty evaluation value increases as the exceedance increases. Rerun the multi-objective optimization to search for the compromise control strategy with the lowest total penalty evaluation value in the solution space containing the soft constraint function; When selecting the optimal control strategy, the difference between the actuator action amount corresponding to the candidate control strategy and the actual action amount at the previous moment is calculated. If the difference exceeds the mechanical response threshold, a damping correction is applied to the candidate control strategy, and the final control strategy after smoothing is output.

8. The method according to claim 1, characterized in that, The method further includes a pre-regulation step based on power mutation prediction, specifically including: Predict the rate of power change within a preset time period based on the power demand curve of power grid dispatch and long short-term memory network. If the predicted power change rate exceeds the mutation threshold and is a high-power discharge mutation, the flow rate of the first liquid cooler unit is increased to the target value in advance, and the second liquid cooler unit is controlled to pre-cool the high-power battery cluster. If the predicted power change rate exceeds the mutation threshold and is a sudden power drop, the pump speed of the first liquid chiller unit is reduced in advance, and the second liquid chiller unit is controlled to maintain low power operation to eliminate residual hot spots.

9. The method according to claim 8, characterized in that, The pre-conditioning step based on power mutation prediction also includes asynchronous speculative control and dew point safety protection procedures for predicted false alarms, specifically including: Before performing the pre-adjustment, ambient temperature and humidity data are collected in real time and the dew point temperature is calculated. The minimum allowable inlet temperature of the coolant is set as the sum of the dew point temperature and the preset anti-condensation safety margin. If a high-power discharge mutation is predicted, the target fluid temperature required for precooling is calculated. If the target fluid temperature is lower than the minimum allowable inlet temperature, the temperature control setpoints of the first liquid cooler and the second liquid cooler are forcibly clamped to the minimum allowable inlet temperature. Asynchronous speculative precooling is performed. At the initial prediction moment when a power surge is determined, only the flow rate of the second liquid chiller is increased and the opening of the second valve is adjusted to a preset value, while the operating state of the first liquid chiller remains unchanged. Monitor the rate of change of the real-time current of the battery cluster over time and set a load confirmation time window; If the rate of change is detected to exceed the measured trigger threshold within the load confirmation time window, the prediction is determined to be certain, and the first liquid cooling unit is immediately controlled to perform the action of increasing the flow rate to the target value. If the rate of change is not detected to exceed the measured trigger threshold after the load confirmation time window has expired, it is determined to be a false alarm, and the second liquid cooling unit and the second valve are controlled to return to the state before pre-adjustment at a preset damping rate.

10. A thermal management and power coordination control system for a liquid-cooled energy storage system, characterized in that, The system for performing the method according to any one of claims 1 to 9 comprises: The data processing module is used to collect multi-dimensional operational data and perform feature fusion on grid dispatching requirements and battery thermal status to determine the conflict level and grid dispatching scenario. The decision module is used to run a three-dimensional prediction model of the thermal power grid, perform multi-objective optimization based on the conflict level and the grid dispatch scenario, and output dynamic weights for four dimensions: power tracking, temperature safety, cooling energy consumption and grid response. The execution control module is used to generate control signals according to the dynamically allocated weights, and adjust the operating power of the first liquid cooler unit, the operating power of the second liquid cooler unit, the opening degree of the valve assembly, and the output power of the battery cluster, respectively.

Images