A method of thermal management of an energy storage unit

Abstract

The present application relates to the technical field of energy storage battery thermal management, and particularly relates to a thermal management method of an energy storage unit, which comprises the following steps: collecting key parameters in a current control period and inputting the key parameters into a pre-trained temperature difference prediction model to output temperature difference prediction values and trend directions of each evaporation branch in a next control period; judging current attributes of each evaporation branch and adjusting an opening degree of a corresponding expansion valve in advance according to the current attributes; feeding forward to calculate a preliminary expansion valve opening degree, calculating a feedback correction amount; calculating a corrected valve position control amount; combining a maximum allowable opening degree of the expansion valve, combining a redistribution coordination mechanism to obtain a final valve opening degree instruction of each expansion valve, and issuing, driving and controlling a final opening degree of a corresponding expansion valve in each evaporation branch to realize dynamic optimization management of each evaporation branch. Through the thermal management method, the fundamental problem that the prior art cannot effectively monitor and adjust a single evaporation branch is solved, and sensors and valves need not be additionally arranged for each battery direct cooling plate.

Description

Technical Field

[0001] This invention relates to the field of thermal management technology for energy storage units, and in particular to a thermal management method for energy storage units. Background Technology

[0002] An energy storage unit is a device or system capable of storing energy, allowing excess energy to be used when needed. Energy storage units are commonly used to improve energy efficiency, reduce energy waste, and enhance grid stability. Energy storage units can include devices or technologies such as batteries, supercapacitors, flywheels, and hydraulic systems, with batteries being one of the most commonly used energy storage methods.

[0003] Existing energy storage units typically employ air cooling, liquid cooling, or direct cooling / heating methods with refrigerant to dissipate heat or provide heat to the battery cells. In direct cooling / heating systems, a single direct cooling / heating loop completes the cooling / heating of the refrigerant, and the heat exchange occurs directly between the refrigerant and the battery cells via a cold plate; a solenoid valve switches between the direct cooling and direct heating paths.

[0004] Existing technologies also propose refrigerant distribution and control systems, where one expansion valve controls a parallel branch consisting of two or more battery direct cooling plates. For example, Chinese invention patent document CN118943568A, published on November 12, 2024. However, existing refrigerant distribution methods, represented by this patent, cannot guarantee sufficient refrigerant flow on a single battery direct cooling plate when there are a large number of plates. Specifically, because the refrigerant is in an unstable two-phase flow state before entering the distributor, it is easily affected by factors such as the gas-liquid phase ratio, flow pattern, and branch structure, leading to uneven refrigerant distribution in each branch. Uneven cooling capacity distribution will result in insufficient cooling capacity of some battery direct cooling plates, causing temperature increases or increased temperature differences, further exacerbating the uneven temperature distribution between batteries. Furthermore, the outlet overheating of a single battery direct cooling plate cannot be effectively monitored, affecting the consistency and service life of the battery direct cooling plates, and even leading to safety hazards such as thermal runaway.

[0005] To address the aforementioned technical problems, a Chinese invention patent document with publication number CN119468527A and publication date of February 18, 2025, has been proposed in the prior art. This patent document achieves closed-loop monitoring and regulation of the overheating of a single battery's direct cooling plate outlet by adding an independent NTC and electronic expansion valve to each parallel sub-branch. However, this control method sacrifices shunt accuracy for increased hardware, significantly increasing the overall system cost and failure rate, and is not conducive to large-scale deployment and maintenance in energy storage power station scenarios. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a thermal management method for energy storage units. Using battery temperature uniformity as the control input, the opening of the corresponding expansion valve is independently adjusted, solving the fundamental problem that existing technologies cannot effectively monitor and adjust individual evaporation branches, and eliminating the need to add sensors and valves to each battery direct cooling plate.

[0007] This invention is achieved by adopting the following technical solution: A thermal management method for an energy storage unit, the energy storage unit including a refrigeration circuit, the refrigeration circuit including a compressor and several parallel evaporation branches, each evaporation branch including an expansion valve, a distributor and several parallel battery direct cooling plates; the thermal management method includes the following steps: Step S1. Collect key parameters within the current control cycle; Step S2. Input the collected key parameters into the pre-trained temperature difference prediction model, and output the predicted temperature difference values ​​for each evaporation branch in the next control cycle. and trend direction; Step S3. Based on the trend direction, determine the current attributes of each evaporation branch, and adjust the opening of the corresponding expansion valve in advance accordingly; the current attributes include hot branch, cold branch, or intermediate branch. Step S4. Based on the predicted temperature difference In addition to the current attributes of each evaporation branch, the feedforward calculation is used to determine the initial expansion valve opening. Simultaneously, the actual temperature difference measured in the current control cycle is compared with the preset target temperature difference, and the feedback correction amount is calculated based on the difference. ; Step S5. Based on the initial expansion valve opening and feedback correction, calculate the corrected valve position control quantity: ; Step S6. Combine the maximum allowable opening of the expansion valve to determine whether the redistribution coordination mechanism is triggered; if not, use the corrected valve position control quantity as the final valve opening command; if so, coordinate and redistribute the corrected valve position control quantity to obtain the final valve opening command for each expansion valve. Step S7. Issue the final valve opening command to drive and control the final opening of the corresponding expansion valve in each evaporation branch, thereby achieving dynamic optimization management of each evaporation branch.

[0008] The key parameters include: the highest temperature, lowest temperature, and real-time temperature difference of each battery direct cooling plate. Evaporation pressure, condensation pressure, suction temperature, condensation outlet temperature, ambient temperature, load current, compressor operating frequency, and the current opening degree of the expansion valves in each evaporation branch.

[0009] The temperature difference prediction model is based on a recurrent neural network constructed by a gated recurrent unit to establish a nonlinear mapping between the state of the battery direct cooling plate and the refrigeration circuit and future temperature difference changes. The temperature difference prediction model introduces a physical state gating mechanism in the hidden layer and explicitly embeds the physical boundary conditions of the refrigeration system in the network structure and training objectives.

[0010] During the training process of the temperature difference prediction model, the time backpropagation algorithm is used in conjunction with the physical constraint loss function; wherein, the loss function consists of three terms: the mean square value of the temperature difference prediction error, the thermal constraint penalty term, and the smoothing term.

[0011] If the prediction error deviates over a long period, the bias parameters of the temperature difference prediction model are locally updated.

[0012] Step S3 specifically refers to: when the trend direction of the evaporation branch is predicted to be an upward trend in temperature difference, the evaporation branch is determined to be a hot branch, and the opening of the corresponding expansion valve is increased; when the trend direction of the evaporation branch is predicted to be a downward trend in temperature difference, the evaporation branch is determined to be a cold branch, and the opening of the corresponding expansion valve is decreased; when the trend direction of the evaporation branch is predicted to be a stable trend in temperature difference, the evaporation branch is determined to be an intermediate branch, and the opening of the corresponding expansion valve is maintained.

[0013] The feedback correction amount The calculation method is as follows: the feedback correction amount is obtained by using a table lookup or proportional-integral algorithm. .

[0014] In step S6, determining whether the redistribution coordination mechanism is triggered specifically means: if the expansion valve of a certain hot branch is already at the maximum allowable opening, and the temperature difference of the corresponding battery direct cooling plate is still too high, then the redistribution coordination mechanism is triggered; the corrected valve position control quantity is coordinated and redistributed, specifically means: identifying the cold branch whose temperature difference is lower than the preset target temperature difference and whose deviation exceeds the preset threshold, defining it as a flow-allowable branch, and gradually reducing the opening of the expansion valve of the flow-allowable branch according to a set ratio.

[0015] The redistribution coordination mechanism also includes checking the compressor suction and discharge superheat and condensing pressure before adjusting the opening of the expansion valve in the refrigerant branch, in order to ensure that the refrigerant flow adjustment will not cause liquid slugging or high pressure over-limit.

[0016] The step S7 of issuing the final valve opening command specifically refers to issuing the final valve opening command to each actuator in a grouped or staggered manner.

[0017] After issuing the final valve opening command, the system also records the execution status and key parameters of each evaporation branch to form the operating data for the next control cycle, which is used for the next status acquisition and trend prediction.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. In this invention, instead of relying on the total outlet superheat, which is prone to failure in multi-parallel branches, as the control target, the battery temperature uniformity is used as the control input. This is achieved by real-time monitoring of the maximum temperature difference of the battery direct cooling plate within the evaporation branch controlled by each expansion valve. and the real-time temperature difference As a basis for control, the opening degree of the corresponding expansion valve is adjusted independently. This method solves the fundamental problem that existing technologies cannot effectively monitor and adjust individual evaporation branches, eliminating the need to add sensors and valves to each battery direct cooling plate.

[0019] Specifically, during the expansion valve opening adjustment process, the valve opening is first intervened in advance based on the trend of temperature difference changes, without relying on real-time temperature difference, thereby reducing overshoot and hysteresis. Then, using a temperature difference prediction model and the actual measured temperature difference, a corrected valve position control quantity is obtained to offset prediction errors and environmental disturbances. Finally, combined with the maximum allowable opening of the expansion valve, it is determined whether to trigger the redistribution coordination mechanism to obtain the final valve opening command. This can further improve the overall temperature consistency of the system under extreme loads or uneven refrigerant distribution.

[0020] This invention controls the refrigerant flow rate of each parallel evaporation branch on a per-branch basis. Through the coordination of the aforementioned components, precise and dynamic regulation of the refrigerant flow rate in each branch is achieved. Compared to traditional static flow distribution strategies, this method adaptively optimizes the refrigerant flow distribution within the energy storage unit, significantly improving the temperature uniformity of the battery's direct cooling plate while effectively ensuring control accuracy under critical operating conditions. Furthermore, the overall cost, structural complexity, and failure rate of this energy storage unit are significantly reduced, making it more suitable for the high reliability, high security, and scalability requirements of large-scale energy storage scenarios.

[0021] 2. In this invention, the temperature difference prediction model introduces a physical state gating mechanism in the hidden layer and explicitly embeds the physical boundary conditions of the refrigeration circuit in the network structure and training target, so that the propagation of temperature-related features is limited to the safe range of suction superheat and condensation pressure, making the prediction results of the temperature difference prediction model more reasonable.

[0022] 3. In this invention, the setting of the loss function enables the recurrent neural network to maintain thermodynamic rationality while learning data features, thereby improving the transferability of predictions under different loads and environments.

[0023] 4. If the prediction error deviates over a long period, the bias parameters of the temperature difference prediction model can be locally updated to achieve long-term adaptive correction.

[0024] 5. In this invention, for channels predicted to be hot branches, the opening of their expansion valves is increased in advance to mitigate impending insufficient cooling; for channels predicted to be cold branches, the valve opening is appropriately reduced to allow some refrigerant flow; and for intermediate branches, the current state is maintained. This invention intervenes in advance based on the trend of temperature difference changes, without relying on instantaneous temperature difference deviations, thereby reducing overshoot and hysteresis.

[0025] 6. Through the redistribution coordination mechanism, more refrigerant flow is redistributed to the hot branch, thereby forming a dynamic equilibrium flow redistribution process among multiple evaporation branches.

[0026] 7. The redistribution coordination mechanism also checks the compressor's suction and discharge superheat and condensing pressure to ensure that refrigerant flow adjustments do not cause compressor liquid slugging or high-pressure over-limits. This ensures that the redistribution coordination process is constrained by the system's thermodynamic safety boundary conditions, guaranteeing that the energy storage unit operates within a safe range during redistribution. This strategy, based on traditional static flow control, introduces temperature difference trend prediction and real-time redistribution collaborative logic, realizing adaptive optimization of system cooling capacity distribution, which can improve the temperature uniformity and operational safety of energy storage units.

[0027] 8. Issue the final valve opening command to each actuator in a grouped or staggered manner to prevent pressure disturbance caused by simultaneous operation.

[0028] 9. Record the execution status and key parameters of each evaporation branch to form the operating data for the next control cycle. This data will be used for the next status acquisition and trend prediction, enabling the temperature difference prediction model to continuously optimize its judgment parameters and allowing the control performance to adaptively improve under different environmental and load conditions. The entire process is executed continuously in a cyclical manner, forming a closed-loop operation mechanism of prediction, regulation, and protection. Attached Figure Description

[0029] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments, wherein: Figure 1 This is a schematic diagram of the process of the present invention; Figure 2 This is a schematic diagram of the energy storage unit of the present invention. Detailed Implementation

[0030] Example 1 As a basic embodiment of the present invention, the present invention includes a thermal management method for an energy storage unit. The energy storage unit includes a refrigeration circuit, which includes a compressor and several parallel-arranged evaporation branches. Each evaporation branch includes an expansion valve, a distributor, and several parallel-arranged battery direct cooling plates. The thermal management method includes the following steps: Step S1. Collect key parameters within the current control cycle.

[0031] Step S2. Input the collected key parameters into the pre-trained temperature difference prediction model, and output the predicted temperature difference values ​​for each evaporation branch in the next control cycle. And the direction of trends.

[0032] Step S3. Based on the trend direction, determine the current attributes of each evaporation branch and adjust the opening of the corresponding expansion valve accordingly. The current attributes include hot branch, cold branch, or intermediate branch.

[0033] Step S4. Based on the predicted temperature difference In addition to the current attributes of each evaporation branch, the feedforward calculation is used to determine the initial expansion valve opening. Simultaneously, the actual temperature difference measured in the current control cycle is compared with the preset target temperature difference, and the feedback correction amount is calculated based on the difference. .

[0034] Step S5. Based on the initial expansion valve opening and feedback correction, calculate the corrected valve position control quantity: .

[0035] Step S6. Combine the maximum allowable opening of the expansion valve to determine whether the redistribution coordination mechanism is triggered; if not, use the corrected valve position control quantity as the final valve opening command; if so, coordinate and redistribute the corrected valve position control quantity to obtain the final valve opening command for each expansion valve.

[0036] Step S7. Issue the final valve opening command to drive and control the final opening of the corresponding expansion valve in each evaporation branch, thereby achieving dynamic optimization management of each evaporation branch.

[0037] Example 2 In a preferred embodiment of the present invention, the present invention includes a thermal management method for an energy storage unit. The energy storage unit includes a refrigeration circuit, which includes a compressor and several parallel-arranged evaporation branches. Each evaporation branch includes an expansion valve, a distributor, and several parallel-arranged battery direct cooling plates. The thermal management method includes the following steps: Step S1. Collect key parameters within the current control cycle. These key parameters include: the highest temperature, lowest temperature, and real-time temperature difference of each battery direct cooling plate. Evaporation pressure, condensation pressure, suction temperature, condensation outlet temperature, ambient temperature, load current, compressor operating frequency, and the current opening degree of the expansion valves in each evaporation branch.

[0038] Step S2. Input the collected key parameters into the pre-trained temperature difference prediction model, and output the predicted temperature difference values ​​for each evaporation branch in the next control cycle. The temperature difference prediction model is constructed using a recurrent neural network based on gated recurrent units to establish a nonlinear mapping between the state of the battery direct cooling plate and the refrigeration circuit and future temperature difference changes. Specifically, the temperature difference prediction model introduces a physical state gating mechanism in the hidden layer and explicitly embeds the physical boundary conditions of the refrigeration circuit in the network structure and training objective. More specifically, during the training process of the temperature difference prediction model, a time backpropagation algorithm combined with a physical constraint loss function is used. The loss function consists of three terms: the mean square value of the temperature difference prediction error, a thermal constraint penalty term, and a smoothing term.

[0039] Step S3. Based on the trend direction, determine the current attributes of each evaporation branch and adjust the opening of the corresponding expansion valve accordingly. The current attributes include hot branch, cold branch, or intermediate branch.

[0040] Step S4. Based on the predicted temperature difference In addition to the current attributes of each evaporation branch, the feedforward calculation is used to determine the initial expansion valve opening. Simultaneously, the actual temperature difference measured in the current control cycle is compared with the preset target temperature difference, and the feedback correction amount is calculated based on the difference. .

[0041] Step S5. Based on the initial expansion valve opening and feedback correction, calculate the corrected valve position control quantity: .

[0042] Step S6. Combine the maximum allowable opening of the expansion valve to determine whether the redistribution coordination mechanism is triggered; if not, use the corrected valve position control quantity as the final valve opening command; if so, coordinate and redistribute the corrected valve position control quantity to obtain the final valve opening command for each expansion valve.

[0043] Step S7. Issue the final valve opening command to drive and control the final opening of the corresponding expansion valve in each evaporation branch, thereby achieving dynamic optimization management of each evaporation branch.

[0044] Example 3 In another preferred embodiment of the present invention, the present invention includes a thermal management method for an energy storage unit. The energy storage unit includes a refrigeration circuit, which includes a compressor and several parallel-arranged evaporation branches. Each evaporation branch includes an expansion valve, a distributor, and several parallel-arranged battery direct cooling plates. The thermal management method includes the following steps: Step S1. Collect key parameters within the current control cycle.

[0045] Step S2. Input the collected key parameters into the pre-trained temperature difference prediction model, and output the predicted temperature difference values ​​for each evaporation branch in the next control cycle. And the direction of trends.

[0046] Step S3. Based on the trend direction, determine the current attribute of each evaporation branch and adjust the opening of the corresponding expansion valve accordingly. The current attribute includes hot branch, cold branch, or intermediate branch. Specifically, when the trend direction of the evaporation branch is predicted to be an upward trend in temperature difference, the evaporation branch is determined to be a hot branch, and the opening of the corresponding expansion valve is increased; when the trend direction of the evaporation branch is predicted to be a downward trend in temperature difference, the evaporation branch is determined to be a cold branch, and the opening of the corresponding expansion valve is decreased; when the trend direction of the evaporation branch is predicted to be a stable trend in temperature difference, the evaporation branch is determined to be an intermediate branch, and the opening of the corresponding expansion valve is maintained.

[0047] Step S4. Based on the predicted temperature difference In addition to the current attributes of each evaporation branch, the feedforward calculation is used to determine the initial expansion valve opening. Simultaneously, the actual temperature difference measured in the current control cycle is compared with the preset target temperature difference, and the feedback correction amount is calculated based on the difference. .

[0048] Step S5. Based on the initial expansion valve opening and feedback correction, calculate the corrected valve position control quantity: .

[0049] Step S6. Based on the maximum allowable opening of the expansion valve, determine whether the redistribution coordination mechanism is triggered. If not, use the corrected valve position control quantity as the final valve opening command. If so, redistribute the corrected valve position control quantity to obtain the final valve opening command for each expansion valve. Specifically, determining whether the redistribution coordination mechanism is triggered means that if the expansion valve of a hot branch is already at its maximum allowable opening, and the temperature difference of the corresponding battery direct cooling plate remains high, then the redistribution coordination mechanism is triggered. Redistributing the corrected valve position control quantity specifically means identifying cold branches where the temperature difference is lower than the preset target temperature difference and the deviation exceeds a preset threshold, defining them as allowable flow branches, and gradually reducing the opening of the expansion valves in the allowable flow branches according to a set ratio.

[0050] Step S7. Issue the final valve opening command to drive and control the final opening of the corresponding expansion valve in each evaporation branch, thereby achieving dynamic optimization management of each evaporation branch.

[0051] Example 4 As another preferred embodiment of the present invention, the present invention includes a thermal management method for an energy storage unit. (Refer to the appendix of the specification.) Figure 2The energy storage unit includes a refrigeration circuit, which comprises a compressor, a regenerator, a condenser, a liquid receiver, and six parallel evaporation branches. Each evaporation branch includes an expansion valve, a distributor, and eight parallel battery direct cooling plates. To prevent low superheat of the compressor suction gas during regulation, a regenerator is added. One side of the regenerator receives the high-temperature refrigerant from the condenser outlet, and the other side receives the low-temperature refrigerant from the evaporator outlet; heat exchange between the two refrigerants increases the compressor suction gas temperature.

[0052] Refer to the instruction manual appendix Figure 1 The thermal management method includes the following steps: Step S1. Collect key parameters within the current control cycle. These key parameters include: the highest temperature, lowest temperature, and real-time temperature difference of each battery direct cooling plate. The data includes evaporation pressure, condensation pressure, suction temperature, condensation outlet temperature, ambient temperature, load current, compressor operating frequency, and the current opening degree of the expansion valves in each evaporation branch. Real-time temperature difference is also included. This parameter represents the difference between the real-time maximum and minimum temperatures of the direct cooling plate for each battery. Based on this key parameter, derivative quantities such as the rate of change of temperature difference, intake superheat, and condensation subcooling can be calculated, providing an input basis for subsequent trend prediction and control decisions.

[0053] Step S2. Input the collected key parameters into the pre-trained temperature difference prediction model, and output the predicted temperature difference values ​​for each evaporation branch in the next control cycle. And the direction of trends.

[0054] The temperature difference prediction model utilizes the time-series modeling capabilities of recurrent neural networks. By combining thermal constraints with data-driven learning, it achieves a nonlinear mapping between the state of the battery direct cooling plate and the refrigeration circuit and future temperature difference changes. In this temperature difference prediction model, the neural network not only learns the statistical patterns of historical data but also explicitly embeds the physical boundary conditions of the refrigeration system into the network structure and training objectives, making the prediction results of the temperature difference prediction model more reasonable.

[0055] The temperature difference prediction model uses the highest temperature, lowest temperature, and real-time temperature difference of the battery direct cooling plate. The system takes as input parameters the following: evaporation pressure, condensation pressure, suction temperature, condenser outlet temperature, ambient temperature, electronic expansion valve opening, compressor frequency, and battery system charge / discharge load. A time-series sample is generated by continuously sampling multiple control cycles, and the input window length is adaptively adjusted based on the system's thermal inertia. The neural network (RNN) employs a gated recurrent unit structure to avoid the gradient vanishing problem during long-sequence training, and introduces physical state gating in the hidden layers to limit the propagation of temperature-related features to a safe range of suction superheat and condensation pressure.

[0056] During the training phase, the temperature difference prediction model employs a time backpropagation algorithm combined with a physical constraint loss function. The loss terms include: (1) the mean square value of the temperature difference prediction error; (2) a thermodynamic constraint penalty term, used to limit the system thermal state (such as suction superheat and condensing pressure) caused by the prediction results from exceeding the safe range; and (3) a smoothing term, used to penalize unreasonable abrupt changes in predictions at adjacent time steps. This loss function enables the neural network to maintain thermodynamic rationality while learning data features, improving the transferability of predictions under different loads and environments.

[0057] During the online application phase, the temperature difference prediction model takes into account data from the most recent several cycles in each control cycle and outputs the predicted temperature difference for each evaporation branch in the next cycle. The prediction results, including the trend direction (rising, stable, or falling), do not directly drive the actuator. Instead, they are input as a trend signal to the feedforward control module to adjust the expansion valve opening in advance. If the feedback module detects a long-term deviation in the prediction error, the controller will activate the online fine-tuning function to locally update the bias parameters of the temperature difference prediction model, achieving long-term adaptive correction.

[0058] Step S3. Based on the trend direction, determine the current attributes of each evaporation branch and adjust the opening of the corresponding expansion valve accordingly. The current attributes include hot branches, cold branches, or intermediate branches. Specifically, when the trend direction of the evaporation branch is predicted to be an upward trend in temperature difference, the evaporation branch is determined to be a hot branch, and the opening of the corresponding expansion valve is increased in advance to improve the impending insufficient cooling. When the trend direction of the evaporation branch is predicted to be a downward trend in temperature difference, the evaporation branch is determined to be a cold branch, and the opening of the corresponding expansion valve is decreased in advance to allow some refrigerant flow. When the trend direction of the evaporation branch is predicted to be a stable trend in temperature difference, the evaporation branch is determined to be an intermediate branch, and the opening of the corresponding expansion valve is maintained. This step intervenes in advance based on the temperature difference change trend, without relying on the instantaneous temperature difference deviation, thereby reducing overshoot and hysteresis.

[0059] Step S4. Based on the predicted temperature difference In addition to the current attributes of each evaporation branch, the feedforward calculation is used to determine the initial expansion valve opening. Simultaneously, the actual temperature difference measured in the current control cycle is compared with the preset target temperature difference, and closed-loop correction is performed based on the difference. Specifically, a lookup table or proportional-integral algorithm is used to calculate the feedback correction amount. It is used to offset prediction errors, environmental disturbances, etc.

[0060] Step S5. Based on the initial expansion valve opening and feedback correction, calculate the corrected valve position control quantity: .

[0061] Step S6. Based on the maximum allowable opening of the expansion valve, determine whether the redistribution coordination mechanism is triggered. If not, use the corrected valve position control quantity as the final valve opening command. If the expansion valve of a certain hot branch is already at its maximum allowable opening, and the temperature difference of the corresponding battery direct cooling plate remains high, the redistribution coordination mechanism is triggered to coordinate and redistribute the corrected valve position control quantity to obtain the final valve opening command for each expansion valve. Specifically, identify cold branches with a temperature difference lower than the preset target temperature difference and a deviation exceeding a preset threshold, i.e., identify cold branches with a temperature difference significantly lower than the target temperature difference, define them as allowable flow branches, and gradually reduce the opening of the expansion valve of the allowable flow branch according to a set ratio.

[0062] This step distributes more refrigerant flow to hotter branches with larger temperature differences, while keeping the total cooling capacity constant. This passive redistribution process can further improve the overall temperature consistency of the system under extreme loads or uneven refrigerant distribution.

[0063] Meanwhile, this coordinated redistribution process is constrained by thermal safety boundary conditions: before adjusting the opening of the expansion valve in the yield branch, the controller checks the compressor's suction and discharge superheat and condensing pressure to ensure that refrigerant flow adjustment will not cause liquid slugging or high-pressure over-limit. This strategy, based on traditional static flow control, introduces temperature difference trend prediction and real-time redistribution collaborative logic, achieving adaptive optimization of system cooling capacity distribution, which can improve the temperature uniformity and operational safety of the battery system.

[0064] Step S7. Issue the final valve opening command to drive and control the final opening of the corresponding expansion valve in each evaporation branch, achieving dynamic optimization management of each evaporation branch. Specifically, the final valve opening command is issued to each actuator in a grouped or staggered manner to prevent pressure disturbances caused by simultaneous actions. After issuing the final valve opening command, the execution status and key parameters of each evaporation branch are recorded to form the operating data for the new control cycle, which is used for the next status acquisition and trend prediction. As the operating data accumulates, the temperature difference prediction model can continuously optimize its judgment parameters, enabling the control performance to adaptively improve under different environmental and load conditions. The entire process is continuously executed in a cyclical manner, forming a closed-loop operation mechanism of prediction-regulation-protection.

[0065] In summary, any other corresponding modifications made by those skilled in the art after reading this invention document, without requiring creative mental effort, based on the technical solutions and concepts of this invention, are all within the scope of protection of this invention.

Claims

1. A thermal management method for an energy storage unit, the energy storage unit comprising a refrigeration circuit, the refrigeration circuit comprising a compressor and a plurality of parallel-arranged evaporation branches, each evaporation branch comprising an expansion valve, a distributor, and a plurality of parallel-arranged battery direct cooling plates; characterized in that: Thermal management methods include the following steps: Step S1. Collect key parameters within the current control cycle; Step S2. Input the collected key parameters into the pre-trained temperature difference prediction model, and output the predicted temperature difference values ​​for each evaporation branch in the next control cycle. and trend direction; Step S3. Based on the trend direction, determine the current attributes of each evaporation branch, and adjust the opening of the corresponding expansion valve in advance accordingly; the current attributes include hot branch, cold branch, or intermediate branch. Step S4. Based on the predicted temperature difference value In addition to the current attributes of each evaporation branch, the feedforward calculation is used to determine the initial expansion valve opening. Simultaneously, the actual temperature difference measured in the current control cycle is compared with the preset target temperature difference, and the feedback correction amount is calculated based on the difference. ; Step S5. Based on the initial expansion valve opening and feedback correction, calculate the corrected valve position control quantity: ； Step S6. Combine the maximum allowable opening of the expansion valve to determine whether the redistribution coordination mechanism is triggered; if not, use the corrected valve position control quantity as the final valve opening command; if so, coordinate and redistribute the corrected valve position control quantity to obtain the final valve opening command for each expansion valve. Step S7. Issue the final valve opening command to drive and control the final opening of the corresponding expansion valve in each evaporation branch, thereby achieving dynamic optimization management of each evaporation branch.

2. The thermal management method for an energy storage unit according to claim 1, characterized in that: The key parameters include: the highest temperature, lowest temperature, and real-time temperature difference of each battery direct cooling plate. Evaporation pressure, condensation pressure, suction temperature, condensation outlet temperature, ambient temperature, load current, compressor operating frequency, and the current opening degree of the expansion valves in each evaporation branch.

3. The thermal management method for an energy storage unit according to claim 1, characterized in that: The temperature difference prediction model is based on a recurrent neural network constructed by a gated recurrent unit to establish a nonlinear mapping between the state of the battery direct cooling plate and the refrigeration circuit and future temperature difference changes. The temperature difference prediction model introduces a physical state gating mechanism in the hidden layer and explicitly embeds the physical boundary conditions of the refrigeration system in the network structure and training objectives.

4. The thermal management method for an energy storage unit according to claim 3, characterized in that: During the training process of the temperature difference prediction model, the time backpropagation algorithm is used in conjunction with the physical constraint loss function; wherein, the loss function consists of three terms: the mean square value of the temperature difference prediction error, the thermal constraint penalty term, and the smoothing term.

5. The thermal management method for an energy storage unit according to claim 4, characterized in that: If the prediction error deviates over a long period, the bias parameters of the temperature difference prediction model are locally updated.

6. The thermal management method for an energy storage unit according to claim 1, characterized in that: Step S3 specifically refers to: when the trend direction of the evaporation branch is predicted to be an upward trend in temperature difference, the evaporation branch is determined to be a hot branch, and the opening of the corresponding expansion valve is increased; when the trend direction of the evaporation branch is predicted to be a downward trend in temperature difference, the evaporation branch is determined to be a cold branch, and the opening of the corresponding expansion valve is decreased; when the trend direction of the evaporation branch is predicted to be a stable trend in temperature difference, the evaporation branch is determined to be an intermediate branch, and the opening of the corresponding expansion valve is maintained.

7. The thermal management method for an energy storage unit according to claim 6, characterized in that: The feedback correction amount The calculation method is as follows: the feedback correction amount is obtained by using a table lookup or proportional-integral algorithm. .

8. A thermal management method for an energy storage unit according to claim 1 or 7, characterized in that: In step S6, determining whether the redistribution coordination mechanism is triggered specifically means: if the expansion valve of a certain hot branch is already at the maximum allowable opening, and the temperature difference of the corresponding battery direct cooling plate is still too high, then the redistribution coordination mechanism is triggered; the corrected valve position control quantity is coordinated and redistributed, specifically means: identifying the cold branch whose temperature difference is lower than the preset target temperature difference and whose deviation exceeds the preset threshold, defining it as a flow-allowable branch, and gradually reducing the opening of the expansion valve of the flow-allowable branch according to a set ratio.

9. A thermal management method for an energy storage unit according to claim 8, characterized in that: The redistribution coordination mechanism also includes checking the compressor suction and discharge superheat and condensing pressure before adjusting the opening of the expansion valve in the refrigerant branch, in order to ensure that the refrigerant flow adjustment will not cause liquid slugging or high pressure over-limit.

10. A thermal management method for an energy storage unit according to claim 8, characterized in that: The step S7 of issuing the final valve opening command specifically refers to issuing the final valve opening command to each actuator in a grouped or staggered manner.

11. A thermal management method for an energy storage unit according to claim 10, characterized in that: After issuing the final valve opening command, the system also records the execution status and key parameters of each evaporation branch to form the operating data for the next control cycle, which is used for the next status acquisition and trend prediction.

Images