Energy storage system thermal management control method and device based on deep learning, equipment and storage medium

Abstract

The application discloses a kind of based on deep learning energy storage system thermal management control method, device, equipment and storage medium, it is related to thermal management system control technical field, comprising: by collecting battery temperature, ambient temperature, charge-discharge power and liquid cooling system working condition data, data is analyzed using hybrid neural network model, output dynamic temperature control interval and liquid cooling system adjustment parameter, and generate target control instruction to regulate and control liquid cooling system.According to real-time data dynamic adjustment thermal management strategy controls the temperature of energy storage system battery cell, realizes intelligent thermal management and prolongs battery life, while reducing energy consumption, improve the safety and reliability of energy storage system.

Description

Technical Field

[0001] This invention relates to the field of thermal management system control technology, and in particular to a thermal management control method, device, equipment and storage medium for energy storage systems based on deep learning. Background Technology

[0002] With the increasing global demand for renewable energy and the transformation of the energy structure, energy storage systems have received widespread attention as an important component for improving grid flexibility, stability, and economy. Especially in distributed energy systems (such as industrial and commercial energy storage, large-scale energy storage power stations, microgrids, and virtual power plants), energy storage devices not only need to have efficient energy conversion capabilities, but also a reliable thermal management system to ensure that the cells operate within the optimal operating temperature range, thereby extending their service life and improving the overall system efficiency.

[0003] Currently, thermal management of energy storage systems mainly relies on traditional temperature control strategies and simple control algorithms. These methods typically include: passive cooling, which maintains the battery pack's operating temperature through natural air cooling or liquid cooling systems, but this method lacks real-time adjustment capabilities. Rule-based active control methods activate fans or adjust the power of the liquid cooling system when the temperature exceeds a preset temperature threshold. While this control method can handle some common temperature fluctuations, it is not flexible enough in complex and ever-changing real-world application environments. Simple feedback control mechanisms use traditional control algorithms such as PID controllers for temperature regulation, but their inability to effectively handle nonlinear and dynamically changing environmental conditions results in limited control effectiveness. Therefore, there is an urgent need for a deep learning-based thermal management control method for energy storage systems to improve their safety and reliability. Summary of the Invention

[0004] The main objective of this application is to provide a method, device, equipment, and storage medium for thermal management control of energy storage systems based on deep learning, aiming to solve the technical problem of how to improve the safety and reliability of energy storage systems.

[0005] To achieve the above objectives, this application proposes a deep learning-based thermal management control method for energy storage systems, comprising:

[0006] Collect data on cell temperature, ambient temperature, charging and discharging power, and liquid cooling system operating conditions;

[0007] The cell temperature, ambient temperature, charging / discharging power, and liquid cooling system operating data are input into a preset temperature control model to obtain a dynamic temperature control range and liquid cooling system adjustment parameters. The preset temperature control model is a hybrid neural network model, which includes a spatial feature extraction layer, a temporal feature extraction layer, a feature fusion layer, and a dynamic decision layer. The spatial feature extraction layer includes a first convolutional block and a second convolutional block, each containing two one-dimensional convolutional layers, a batch normalization layer, and an activation function layer. The temporal feature extraction layer includes a long short-term memory network, an attention layer, and a fully connected layer. The long short-term memory network includes a three-layer bidirectional long short-term memory sub-network. The dynamic decision layer includes a temperature control sub-network and a liquid cooling adjustment sub-network. The temperature control sub-network outputs the dynamic temperature control range through the three-layer fully connected network, and the liquid cooling adjustment sub-network outputs the liquid cooling system adjustment parameters through a reinforcement learning strategy.

[0008] Based on the dynamic temperature control range and the liquid cooling system adjustment parameters, a target control command is generated;

[0009] The target control command is sent to the liquid cooling system to enable the liquid cooling system to adjust and complete the thermal management control.

[0010] In one embodiment, the step of inputting the cell temperature, the ambient temperature, the charge / discharge power, and the liquid cooling system operating condition data into a preset temperature control model to obtain the dynamic temperature control range and the liquid cooling system adjustment parameters includes:

[0011] The cell temperature and the ambient temperature are input into the time-series feature extraction layer of the preset temperature control model for analysis to obtain the cell temperature change trend characteristics.

[0012] The charging and discharging power and the liquid cooling system operating data are input into the spatial feature extraction layer of the preset temperature control model for pattern recognition to determine the priority features of the current operating condition.

[0013] The cell temperature change trend feature and the current operating condition priority feature are spliced ​​together through the feature fusion layer in the preset temperature control model to obtain the fused feature;

[0014] The fused features are passed through the dynamic decision layer in the preset temperature control model to obtain the dynamic temperature control range and liquid cooling system adjustment parameters.

[0015] In one embodiment, the step of inputting the cell temperature and the ambient temperature into the time-series feature extraction layer of a preset temperature control model for analysis to obtain the cell temperature change trend characteristics includes:

[0016] A time series is obtained based on the cell temperature and the ambient temperature;

[0017] The time series data is input into a long short-term memory network to generate time series features;

[0018] The time-series features are weighted through an attention layer to obtain key temperature change features;

[0019] The key temperature change characteristics are input into the fully connected layer to obtain the cell temperature change trend characteristics for a preset time period.

[0020] In one embodiment, the step of inputting the charging / discharging power and the liquid cooling system operating condition data into the spatial feature extraction layer of a preset temperature control model for pattern recognition to determine the priority features of the current operating condition includes:

[0021] The charging and discharging power and the liquid cooling system operating data are input into the spatial feature extraction layer to extract spatial features.

[0022] The spatial features are classified to obtain the current operating conditions;

[0023] Based on the current operating conditions, a condition priority label is obtained through matching.

[0024] The current operating condition priority feature is obtained by converting the operating condition priority label.

[0025] In one embodiment, the step of obtaining the dynamic temperature control range and liquid cooling system adjustment parameters by passing the fused features through the dynamic decision layer in a preset temperature control model includes:

[0026] The fused features are input into the temperature control subnetwork of the dynamic decision layer to extract primary features related to the temperature control range. The primary features are obtained by the temperature control subnetwork through the first fully connected network to perform preliminary processing on the fused features.

[0027] The primary features are input into the second fully connected layer of the temperature control sub-network for further processing to extract intermediate features related to the temperature control range.

[0028] The intermediate features are input into the third fully connected layer of the temperature control subnetwork for processing, and a dynamic temperature control range is output. The dynamic temperature control range is used to guide the real-time adjustment range of the battery temperature.

[0029] The fused features are input into the liquid-cooled regulation subnetwork of the dynamic decision layer to generate an initial policy. The initial policy is obtained by processing the fused features through the policy network in the reinforcement learning policy via the liquid-cooled regulation subnetwork.

[0030] Based on the initial strategy and a preset reward function, an optimized adjustment strategy is obtained.

[0031] The liquid cooling system adjustment parameters are generated based on the optimized adjustment strategy.

[0032] In one embodiment, the step of generating a target control command based on the dynamic temperature control range and the liquid cooling system adjustment parameters includes:

[0033] Obtain the real-time cell temperature of the battery;

[0034] The highest and lowest cell temperatures are determined based on the dynamic temperature control range.

[0035] Based on the liquid cooling system adjustment parameters, the operating parameters of the liquid cooling system are determined, including the target cooling power and the target circulation speed.

[0036] A pre-cooling command is generated when the real-time cell temperature exceeds the maximum cell temperature or does not exceed the minimum cell temperature.

[0037] The pre-cooling command and the liquid cooling system operating parameters are integrated to generate the target control command.

[0038] In one embodiment, the step of generating a target control command based on the dynamic temperature control range and the liquid cooling system adjustment parameters includes:

[0039] Obtain the real-time cell temperature of the battery;

[0040] The highest and lowest cell temperatures are determined based on the dynamic temperature control range.

[0041] Based on the liquid cooling system adjustment parameters, the operating parameters of the liquid cooling system are determined, including the target cooling power and the target circulation speed.

[0042] A pre-cooling command is generated when the real-time cell temperature exceeds the maximum cell temperature or does not exceed the minimum cell temperature.

[0043] The pre-cooling command and the liquid cooling system operating parameters are integrated to generate the target control command.

[0044] Furthermore, to achieve the above objectives, this application also proposes a deep learning-based thermal management control device for energy storage systems, the deep learning-based thermal management control device comprising:

[0045] The acquisition module is used to collect data on cell temperature, ambient temperature, charging and discharging power, and liquid cooling system operating conditions.

[0046] The processing module is used to input the cell temperature, ambient temperature, charging and discharging power, and liquid cooling system operating condition data into a preset temperature control model to obtain a dynamic temperature control range and liquid cooling system adjustment parameters. The preset temperature control model is a hybrid neural network model, which includes a spatial feature extraction layer, a temporal feature extraction layer, a feature fusion layer, and a dynamic decision layer. The spatial feature extraction layer includes a first convolutional block and a second convolutional block, each of which contains two one-dimensional convolutional layers, a batch normalization layer, and an activation function layer. The temporal feature extraction layer includes a long short-term memory network, an attention layer, and a fully connected layer. The long short-term memory network includes a three-layer bidirectional long short-term memory sub-network. The dynamic decision layer includes a temperature control sub-network and a liquid cooling adjustment sub-network. The temperature control sub-network outputs the dynamic temperature control range through the three-layer fully connected network, and the liquid cooling adjustment sub-network outputs the liquid cooling system adjustment parameters through a reinforcement learning strategy.

[0047] The instruction generation module is used to generate target control instructions based on the dynamic temperature control range and the liquid cooling system adjustment parameters;

[0048] The sending module is used to send the target control command to the liquid cooling system so that the liquid cooling system can adjust and complete the thermal management control.

[0049] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the deep learning-based thermal management control method for energy storage systems described above.

[0050] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the deep learning-based thermal management control method for energy storage systems as described above.

[0051] This application collects data on cell temperature, ambient temperature, charging / discharging power, and liquid cooling system operating conditions. It then analyzes this data using a hybrid neural network model to output a dynamic temperature control range and liquid cooling system adjustment parameters, generating target control commands to regulate the liquid cooling system. Based on real-time data, the thermal management strategy is dynamically adjusted to control the temperature of the battery cells in the energy storage system, achieving intelligent thermal management to extend battery life, reduce energy consumption, and improve the safety and reliability of the energy storage system. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a flowchart illustrating the first embodiment of the deep learning-based thermal management control method for energy storage systems in this application.

[0054] Figure 2 This is a block diagram of the preset temperature control model structure of the first embodiment of the deep learning-based thermal management control method for energy storage systems in this application;

[0055] Figure 3 This is a flowchart illustrating the second embodiment of the deep learning-based thermal management control method for energy storage systems in this application.

[0056] Figure 4 This is a flowchart illustrating the third embodiment of the deep learning-based thermal management control method for energy storage systems in this application.

[0057] Figure 5 This is a schematic diagram of the module structure of the thermal management control device for an energy storage system based on deep learning, according to the first embodiment of the thermal management control method for an energy storage system based on deep learning in this application.

[0058] Figure 6 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the deep learning-based thermal management control method for energy storage systems in this application embodiment.

[0059] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0060] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0061] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0062] Currently, thermal management of energy storage systems mainly relies on traditional temperature control strategies and simple control algorithms. These methods typically include passive cooling, which maintains the battery pack's operating temperature through natural air cooling or liquid cooling systems, but this method lacks real-time adjustment capabilities. Rule-based active control methods activate fans or adjust the power of the liquid cooling system when the temperature exceeds a preset temperature threshold. While this control method can handle some common temperature fluctuations, it is not flexible enough in complex and ever-changing real-world application environments. Simple feedback control mechanisms use traditional control algorithms such as PID controllers for temperature regulation, but their inability to effectively handle nonlinear and dynamically changing environmental conditions results in limited control effectiveness. Therefore, this application proposes a deep learning-based thermal management control method for energy storage systems to address the aforementioned problems. The main solution of this application embodiment is as follows: collect data on cell temperature, ambient temperature, charge / discharge power, and liquid cooling system operating conditions; input the cell temperature, ambient temperature, charge / discharge power, and liquid cooling system operating conditions into a preset temperature control model to obtain a dynamic temperature control range and liquid cooling system adjustment parameters; generate a target control command based on the dynamic temperature control range and liquid cooling system adjustment parameters; and send the target control command to the liquid cooling system to enable the liquid cooling system to adjust and complete thermal management control.

[0063] Based on the above, this application also provides a deep learning-based thermal management control method for energy storage systems, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the deep learning-based thermal management control method for energy storage systems according to this application. In this embodiment, the deep learning-based thermal management control method for energy storage systems includes steps S10 to S40:

[0064] Step S10: Collect data on cell temperature, ambient temperature, charging and discharging power, and liquid cooling system operating conditions.

[0065] It's important to note that real-time monitoring of cell temperature is crucial, as optimal battery performance and lifespan are highly dependent on their operating temperature range (typically 25°C to 35°C). A high-precision sensor network can acquire temperature information for each cell, ensuring optimal operation even during rapid charge and discharge. Secondly, changes in ambient temperature directly impact the heat dissipation efficiency of the energy storage system, especially under extreme climatic conditions such as high or low temperatures. This factor must be considered and dynamically adjusted accordingly. Furthermore, charge and discharge power data provides vital information about current load demands and battery usage. Real-time monitoring of this data allows the system to predict upcoming temperature fluctuations and adjust cooling strategies in advance. Finally, liquid cooling system operating data, including parameters such as flow rate, pressure, and coolant temperature, reflects the current operating status of the cooling system. Based on this comprehensive data, a hybrid neural network model can learn and predict the optimal cooling scheme under different conditions, thereby dynamically adjusting the operating parameters of the liquid cooling system, such as cooling power and circulation speed, to achieve optimal thermal management.

[0066] Step S20: Input the cell temperature, ambient temperature, charging and discharging power and liquid cooling system operating condition data into the preset temperature control model to obtain the dynamic temperature control range and liquid cooling system adjustment parameters.

[0067] It should be noted that the preset temperature control model is a hybrid neural network model, such as... Figure 2 The diagram shows the structure of the preset temperature control model. The preset temperature control model includes a spatial feature extraction layer, a temporal feature extraction layer, a feature fusion layer, and a dynamic decision layer. The spatial feature extraction layer includes a first convolutional block and a second convolutional block. Both the first and second convolutional blocks contain two one-dimensional convolutional layers, a batch normalization layer, and an activation function layer. The temporal feature extraction layer includes a long short-term memory network, an attention layer, and a fully connected layer. The long short-term memory network includes a three-layer bidirectional long short-term memory sub-network. The dynamic decision layer includes a temperature control sub-network and a liquid cooling regulation sub-network. The temperature control sub-network outputs the dynamic temperature control range through the three-layer fully connected network, and the liquid cooling regulation sub-network outputs the liquid cooling system regulation parameters through a reinforcement learning strategy.

[0068] Specifically, firstly, cell temperature, ambient temperature, charge / discharge power, and liquid cooling system operating condition data are input into the spatial feature extraction layer of a pre-defined temperature control model. This spatial feature extraction layer consists of two convolutional blocks: a first convolutional block and a second convolutional block. Each convolutional block contains two one-dimensional convolutional layers to extract spatial features from the data. These convolutional layers can capture local patterns and structures in the data, such as the distribution of cell temperature at different locations. Following the convolutional layers is a batch normalization layer, which normalizes the output of the convolutional layers, helping to accelerate the model training process and improve the model's stability and generalization ability. Finally, an activation function layer is applied to the batch-normalized data, introducing nonlinear factors that enable the model to learn more complex feature representations. Next, the spatial feature extraction layer consists of a three-layer bidirectional long short-term memory network (BiLSTM). BiLSTM can simultaneously consider the forward and backward temporal relationships of the data, effectively capturing long-term dependencies in time-series data. This is crucial for understanding the changes in charge / discharge power over time and the dynamic changes in the liquid cooling system operating conditions. Furthermore, the spatial feature extraction layer includes an attention layer that automatically learns the importance of data at different time steps for thermal management control, thus giving more attention to key information. Finally, a fully connected layer integrates and transforms the extracted temporal features, providing a foundation for subsequent feature fusion and decision-making. In the feature fusion layer, spatial and temporal features are effectively combined. This process is achieved through specific fusion mechanisms, such as weighted summation, concatenation, or more complex interactive operations, enabling the model to utilize information from both spatial and temporal dimensions to form a comprehensive feature representation. This fusion method can more comprehensively reflect the battery's thermal state and operating conditions, providing a more accurate basis for subsequent decisions. The dynamic decision layer is the core of the model, making the final control decision based on the fused features. This decision layer contains two sub-networks: a temperature control sub-network and a liquid cooling regulation sub-network. The temperature control sub-network processes the fused features through a three-layer fully connected network, ultimately outputting a dynamic temperature control range. This control range provides a target range for battery temperature regulation, ensuring that the battery operates within a safe and efficient temperature range. The liquid cooling regulation subnetwork outputs the liquid cooling system regulation parameters through a reinforcement learning strategy. This strategy dynamically adjusts the control strategy based on environmental feedback to achieve optimal control performance. This approach allows the liquid cooling system to flexibly adjust parameters such as cooling power and cycle speed according to the actual battery temperature and operating conditions, thereby achieving precise control of the battery temperature.

[0069] By combining deep learning and reinforcement learning, intelligent optimization control of thermal management in energy storage systems is achieved. This not only improves battery lifespan and performance but also reduces system energy consumption.

[0070] Step S30: Generate target control commands based on the dynamic temperature control range and liquid cooling system adjustment parameters.

[0071] It should be noted that by continuously monitoring the temperature changes of each battery cell through a high-precision sensor network, the current operating status can be grasped in a timely manner. Based on the dynamic temperature control range (e.g., 25°C to 35°C), when the battery cell temperature is detected to be outside this range, immediate action is required. If the battery cell temperature exceeds the maximum threshold or falls below the minimum threshold, it means that the current cooling or heating measures are insufficient to maintain ideal temperature conditions and adjustments are needed, generating a pre-cooling command.

[0072] Next, the liquid cooling system is adjusted based on parameters such as target cooling power and circulation speed. These parameters are optimally set by a deep learning model based on historical data and current operating conditions, aiming to minimize energy consumption while ensuring efficient thermal management. For example, if the cell temperature is too high, the system may increase the cooling power of the liquid cooling system and accelerate the circulation speed; conversely, if the temperature is too low, it may reduce the cooling power or even activate the heating mode.

[0073] Then, the pre-cooling command and liquid cooling system adjustment parameters are integrated into a target control command set and sent to the liquid cooling system for execution via a communication interface. To ensure operational accuracy and safety, the system also takes into account other environmental factors (such as ambient temperature and charging / discharging power). For example, in high-temperature environments, in addition to adjusting the liquid cooling system parameters, the charging / discharging power can be reduced to decrease heat generation, thereby achieving better temperature control.

[0074] Step S40: Send the target control command to the liquid cooling system to enable the liquid cooling system to adjust and complete the thermal management control.

[0075] It should be noted that the target control commands are sent to the liquid cooling system controller via a preset communication protocol. The choice of communication protocol is crucial to ensure that the transmission of commands is both fast and accurate, avoiding any possible signal interference or data loss. For example, using the industry-standard Modbus TCP / IP protocol can guarantee stable transmission of commands in complex industrial environments. After receiving the target control commands, the liquid cooling system controller parses the specific parameters in the commands, such as cooling power and circulation speed. These parameters determine the operating efficiency and cooling effect of the liquid cooling system. Based on the parsed parameters, the controller quickly adjusts the refrigeration modules and circulation pumps of the liquid cooling system to ensure that the flow rate and temperature of the coolant meet the thermal management requirements of the battery.

[0076] In addition, to ensure the stability and reliability of the system, a closed-loop feedback mechanism is introduced. Following step S40, the process includes: receiving feedback signals from the liquid cooling system. Specifically, after receiving the target control command and completing the corresponding parameter adjustments, the liquid cooling system sends a feedback signal. This feedback signal contains the current actual operating parameters of the liquid cooling system, such as actual cooling power, circulation speed, and coolant temperature. These parameters directly reflect the operating status of the liquid cooling system. The feedback signal is then analyzed to obtain the actual operating parameters. Specifically, upon receiving the feedback signal, the signal is immediately analyzed to extract the actual operating parameters. The analysis process needs to ensure the accuracy and completeness of the data so that subsequent comparative analysis can be based on reliable data. The analyzed actual operating parameters are stored in the temporary storage area of ​​the control unit, ready for the next comparative analysis. Next, the actual operating parameters are compared and analyzed with the liquid cooling system operating parameters set in the target control command to obtain the comparison results. Specifically, the analyzed actual operating parameters are compared and analyzed with the liquid cooling system operating parameters set in the target control command. This comparative analysis process is accomplished through a series of preset algorithms and logical rules, aiming to detect whether the actual operating state of the liquid cooling system is consistent with the expected target state. The results of the comparative analysis will clearly indicate whether there is a deviation, and the specific nature of the deviation, such as whether the cooling power is too high or too low, or whether the circulation speed has reached the target value. Finally, when a deviation is found in the comparison results, the deviation is recorded and a new adjustment command is generated and sent to the liquid cooling system for adjustment. Specifically, if the comparison results show a deviation, the specific nature of the deviation will be further analyzed to obtain analysis results, including the magnitude, direction, and possible causes of the deviation. Based on the analysis results, a new adjustment command will be generated. The new adjustment command aims to correct the deviation and bring the actual operating parameters of the liquid cooling system back to the range set by the target control command. For example, if the actual cooling power is lower than the target value, the new adjustment command will instruct the liquid cooling system to increase the cooling power; if the circulation speed is too high, the new adjustment command will require a decrease in the circulation speed. After generating the new adjustment command, the command is sent to the liquid cooling system through the same communication protocol. Upon receiving the new adjustment command, the liquid cooling system will immediately execute the adjustment operation. Through this closed-loop feedback and adjustment mechanism, the liquid cooling system can continuously optimize its operating parameters, ensuring that the battery temperature remains within the optimal operating range. This dynamic adjustment not only improves battery performance and lifespan but also reduces system energy consumption, enhancing the overall operating efficiency and economy of the energy storage system.

[0077] This embodiment collects data on cell temperature, ambient temperature, charging / discharging power, and liquid cooling system operating conditions. It then analyzes this data using a hybrid neural network model to output a dynamic temperature control range and liquid cooling system adjustment parameters, generating target control commands to regulate the liquid cooling system. Based on real-time data, the thermal management strategy is dynamically adjusted to control the temperature of the battery cells in the energy storage system, achieving intelligent thermal management to extend battery life, reduce energy consumption, and improve the safety and reliability of the energy storage system.

[0078] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 3 The deep learning-based energy storage system thermal management control method further includes steps S201-S205 in step S50:

[0079] Step S201: Input the cell temperature and ambient temperature into the time-series feature extraction layer of the preset temperature control model for analysis to obtain the cell temperature change trend features.

[0080] It should be noted that this process, through the time-series analysis capabilities of deep learning models, can accurately capture the changing trend of cell temperature over time, providing a key basis for subsequent thermal management decisions.

[0081] Further, step S201 includes: obtaining a time series based on the cell temperature and ambient temperature; specifically, acquiring cell temperature and ambient temperature data in real time using high-precision sensors. These sensors are distributed at key locations within the battery module, enabling precise measurement of the real-time temperature of the cell and changes in the surrounding environment's temperature. The acquired data is recorded at fixed time intervals (e.g., once per second) to form a time series. This time series contains the values ​​of the cell temperature and ambient temperature at different points in time.

[0082] Next, the time series data is input into a Long Short-Term Memory (LSTM) network to generate time-series features. Specifically, the generated time series data is input into a LSTM network. LSTM is a neural network specifically designed for processing time-series data, effectively capturing long-term dependencies in the data. In this embodiment, a three-layer bidirectional LSTM network is used, with each layer capable of handling both forward and backward time-series relationships. This bidirectional structure allows the model to consider both past and future information simultaneously, thereby gaining a more comprehensive understanding of the cell temperature variation trend and obtaining time-series features.

[0083] Then, the temporal features are weighted through an attention layer to obtain key temperature change features. Specifically, the attention mechanism assigns different weights to features at different time points, identifying and amplifying the periods that have the greatest influence on the final prediction results, thus obtaining the key temperature change features of the main influencing periods.

[0084] Finally, the key temperature change features are input into the fully connected layer to obtain the cell temperature change trend features over a preset time period. Specifically, the weighted key temperature change features are then input into the fully connected layer. The role of the fully connected layer is to further integrate and transform these features to generate the final cell temperature change trend features. Through the nonlinear transformation of multi-layer neural networks, the fully connected layer can learn complex feature representations, thereby more accurately predicting the cell temperature change trend over the preset time period. The output of the fully connected layer is a multi-dimensional feature vector, where each dimension represents the cell temperature change trend at different time scales. For example, some dimensions may represent the cell temperature change over the next few minutes, while other dimensions may represent the temperature change over the next few hours.

[0085] Step S202: Input the charging and discharging power and liquid cooling system operating condition data into the spatial feature extraction layer of the preset temperature control model for pattern recognition to determine the priority features of the current operating condition.

[0086] It should be noted that in order to extract key features from these data, a spatial feature extraction layer is usually used for pattern recognition, and priority features of the current operating condition are determined.

[0087] Further, step S202 also includes: inputting the charging / discharging power and liquid cooling system operating condition data into the spatial feature extraction layer to extract spatial features. Specifically, the charging / discharging power and liquid cooling system operating condition data are input into the spatial feature extraction layer in a preset temperature control model. This network consists of a first convolutional block and a second convolutional block, each containing two one-dimensional convolutional layers, a batch normalization layer, and an activation function layer (such as ReLU). The one-dimensional convolutional layer can capture local patterns, such as the changing trend of charging / discharging power or the operating state of the liquid cooling system (such as flow rate, pressure, etc.). Specifically, the one-dimensional convolutional layer scans the input data through a sliding window mechanism to extract local features at different time points. These local features contain important information about charging / discharging behavior and the operating state of the liquid cooling system. For example, during fast charging, the charging / discharging power may increase significantly, and the flow rate and pressure of the liquid cooling system will adjust accordingly to cope with this change. Through convolution operations, the network can automatically learn these typical patterns and form a preliminary spatial feature representation. Subsequently, the convolutional data enters a batch normalization layer to ensure that each batch of data has the same distribution, which helps to accelerate the training process and improve the stability of the model. Next, a non-linear factor is introduced through an activation function layer, enabling the model to learn more complex feature representations. Finally, after processing by two sets of convolutional blocks, spatial features are generated.

[0088] Next, the spatial features are classified to obtain the current operating condition. Specifically, the extracted spatial features are then input into a classifier for classification. The classifier's role is to map the extracted spatial features to different operating condition categories. For example, the classifier can classify features into different operating condition categories such as "high load discharge," "low load charging," and "normal operation." In this embodiment, the classifier is based on a machine learning classification algorithm (such as support vector machine or random forest), which learns feature patterns under different operating conditions through training, thereby accurately classifying the input features.

[0089] Then, the system matches the current operating conditions to obtain priority labels. Specifically, after obtaining the current operating conditions, the system matches them according to preset priority rules to obtain priority labels. These priority rules are pre-set based on the operational safety and efficiency of the energy storage system. For example, high-load discharge conditions typically have higher priority because the battery temperature rises rapidly under these conditions, requiring timely cooling control. Normal operating conditions, on the other hand, have relatively lower priority because the battery temperature is relatively stable under these conditions, reducing the pressure on thermal management control. The matching process is implemented through a lookup table or rule engine, searching for the corresponding priority label based on the current operating condition. For example, if the current operating condition is "high-load discharge," the matched priority label will be "high priority."

[0090] Finally, the current operating condition priority features are obtained by converting the operating condition priority labels. Specifically, based on the matched operating condition priority labels, the priority labels are converted into current operating condition priority features. This conversion process is usually an encoding process, converting the priority labels into a numerical feature vector. For example, "high priority" can be encoded as the value 1, "medium priority" as the value 0.5, and "low priority" as the value 0. These numerical feature vectors can be directly used for subsequent thermal management control decisions.

[0091] Step S203: The cell temperature change trend features and the current operating condition priority features are spliced ​​together through the feature fusion layer in the preset temperature control model to obtain fused features.

[0092] It should be noted that the cell temperature change trend feature and the current operating condition priority feature reflect the dynamic changes in battery temperature and the importance of the current operating condition, respectively. The cell temperature change trend feature is obtained through time series analysis and includes predictions of cell temperature changes over a future period; while the current operating condition priority feature is obtained through spatial feature extraction and classification, reflecting the urgency of the current operating condition. The function of the feature fusion layer is to concatenate these two features to form a comprehensive feature vector. The concatenation process is usually a simple vector join operation, that is, arranging the two feature vectors in order to form a longer feature vector. For example, if the cell temperature change trend feature is a vector of length 10 and the current operating condition priority feature is a vector of length 3, then the concatenated fused feature will be a vector of length 13.

[0093] Step S204: The fused features are passed through the dynamic decision layer in the preset temperature control model to obtain the dynamic temperature control range and liquid cooling system adjustment parameters.

[0094] It should be noted that the fused features are input into the dynamic decision-making layer. These fused features contain all information regarding the cell temperature change trend and the priority of current operating conditions, providing comprehensive data support for the dynamic decision-making layer. The task of the dynamic decision-making layer is to generate the dynamic temperature control range and liquid cooling system adjustment parameters based on this comprehensive information.

[0095] Further, step S204 includes: inputting the fused features into the temperature control sub-network of the dynamic decision layer, extracting primary features related to the temperature control range, and obtaining the primary features by preliminary processing of the fused features through the first fully connected network of the temperature control sub-network; inputting the primary features into the second fully connected network of the temperature control sub-network for further processing, extracting intermediate features related to the temperature control range; inputting the intermediate features into the third fully connected network of the temperature control sub-network for processing, and outputting the dynamic temperature control range, which is used to guide the real-time adjustment range of battery temperature. Specifically, firstly, the fused features are input into the first fully connected network of the temperature control sub-network. The fused features contain all the information on the cell temperature change trend and the priority of the current operating condition, providing comprehensive data support for the temperature control sub-network. The task of the first fully connected network is to perform preliminary processing on these fused features to extract primary features related to the temperature control range. Then, the extracted primary features are subsequently input into the second fully connected network of the temperature control sub-network. The second fully connected network further processes the primary features to extract intermediate features related to the temperature control range. Intermediate features are a further abstraction and integration of primary features, more accurately reflecting the requirements of the temperature control range. This layer, through more neurons and complex nonlinear transformations, reduces the dimensionality of the primary features and integrates information to extract higher-level features. For example, it might identify the degree of influence of cell temperature change trends on the temperature control range under different operating conditions. Finally, the intermediate features are input into the third fully connected layer of the temperature control sub-network. The third fully connected layer processes the intermediate features and outputs the dynamic temperature control range. This layer, through a large number of neurons and complex nonlinear transformations, maps the intermediate features to specific temperature control ranges. The dynamic temperature control range provides clear guidance for real-time adjustment of battery temperature, ensuring that the battery operates within a safe and efficient temperature range. The output of the dynamic temperature control range not only considers the current cell temperature and operating conditions but also predicts the temperature change trend over a future period. For example, if the model predicts that the cell temperature is about to rise, it will adjust the temperature control range in advance to ensure that the battery temperature does not exceed the safe threshold. This dynamic adjustment mechanism can respond to battery temperature changes in real time, ensuring that the battery maintains optimal performance under various operating conditions.

[0096] Furthermore, the fused features are input into the liquid cooling regulation sub-network of the dynamic decision layer to generate an initial policy. This initial policy is then processed by the policy network within the reinforcement learning strategy within the liquid cooling regulation sub-network, resulting in the initial policy. Based on this initial policy and a preset reward function, an optimized regulation policy is obtained. Finally, the liquid cooling system regulation parameters are calculated based on this optimized policy. Specifically, the liquid cooling regulation sub-network processes the fused features through the policy network of the reinforcement learning strategy to generate the initial policy. The initial policy typically represents a set of continuous values, indicating suggested regulation parameters for the liquid cooling system, such as target cooling power and coolant circulation speed. The initial policy reflects the model's preliminary judgment on the optimal cooling measures to be taken under the current conditions. To further improve the accuracy and adaptability of the policy, the initial policy is optimized based on a preset reward function. This reward function provides a reward value based on the actual operating performance of the liquid cooling system, such as whether the battery temperature is within the target range and whether energy consumption is minimized. A higher reward value indicates better policy performance. Based on the feedback from the reward function, the initial policy is evaluated and adjusted to generate the optimized regulation policy. Based on the optimized regulation policy, the liquid cooling regulation sub-network calculates and transforms this into specific liquid cooling system regulation parameters. These adjustment parameters, including specific cooling power and circulation speed, guide the real-time operation of the liquid cooling system. The generated adjustment parameters not only consider the current cell temperature and operating conditions but also predict temperature change trends over a future period, ensuring that the liquid cooling system can adjust in advance to avoid excessively high or low battery temperatures.

[0097] This embodiment obtains the cell temperature change trend features by inputting cell temperature and ambient temperature into the spatial feature extraction layer, and simultaneously inputs charge / discharge power and liquid cooling system operating condition data into the spatial feature extraction layer to determine the priority features of the current operating condition. Then, these two types of features are concatenated in the feature fusion layer to generate a fused feature. Finally, the dynamic decision layer outputs the dynamic temperature control range and liquid cooling system adjustment parameters based on the fused feature, realizing intelligent thermal management of the energy storage system, ensuring that the cell temperature is always within the optimal operating range, extending battery life and improving system efficiency.

[0098] Based on the first embodiment of this application, in the third embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 4 The deep learning-based energy storage system thermal management control step S30 further includes steps S301 to S305:

[0099] Step S301: Obtain the real-time cell temperature of the battery.

[0100] It's important to note that this process utilizes a high-precision temperature sensor network to monitor real-time temperature changes in the battery cells. Temperature sensors are precisely installed at key locations within the battery module, ensuring coverage of all areas of the battery cells. These sensors acquire cell temperature data at a high frequency (e.g., once per second), capturing minute temperature changes. For example, during battery charging, the cell temperature may rise rapidly, and the high-precision sensors can promptly detect this change, providing real-time feedback for thermal management control. The acquired cell temperature data undergoes preliminary preprocessing. Preprocessing steps include data cleaning to remove noise and outliers, ensuring data accuracy and reliability. For instance, filtering algorithms can remove temperature fluctuations caused by sensor interference or environmental factors, making the data smoother. The preprocessed cell temperature data is stored in system memory in real-time for subsequent analysis and processing.

[0101] Step S302: Determine the highest and lowest cell temperatures based on the dynamic temperature control range.

[0102] It should be noted that the dynamic temperature control range is dynamically generated by a preset temperature control model based on real-time data (such as cell temperature and ambient temperature). This range reflects the ideal range of battery cell temperature under current operating conditions. For example, under high-load discharge conditions, the model may generate a wider temperature range to accommodate rapidly changing temperature requirements; while under normal operating conditions, the temperature range will be relatively narrower to ensure stable battery operation. The specific values ​​for the highest and lowest temperatures are extracted from the dynamic temperature control range. These values ​​are derived based on the current operating conditions and predicted temperature change trends.

[0103] Step S303: Determine the operating parameters of the liquid cooling system based on the adjustment parameters of the liquid cooling system.

[0104] It should be noted that the operating parameters include the target cooling power and the target circulation speed. The liquid cooling system adjustment parameters are generated by a preset temperature control model. These parameters are dynamically generated based on the current charge / discharge power and the operating conditions of the liquid cooling system. Determining the specific operating parameters of the liquid cooling system involves translating the adjustment parameters into actual executable operating commands. For example, the target cooling power can be converted into the compressor's operating frequency, while the target circulation speed can be achieved by adjusting the speed of the circulation pump. Furthermore, other environmental factors can be considered to further optimize the operating commands.

[0105] Step S304: When the real-time cell temperature exceeds the maximum cell temperature or does not exceed the minimum cell temperature, a pre-cooling command is generated.

[0106] It should be noted that if the real-time cell temperature exceeds the maximum cell temperature, it indicates that the battery temperature has approached or exceeded the upper limit for safe operation. In this case, immediate measures are needed to prevent the battery temperature from rising further and to avoid thermal runaway. Conversely, if the real-time cell temperature does not exceed the minimum cell temperature, it indicates that the battery temperature may be too low, and measures should be taken in advance to prevent the battery temperature from dropping further and affecting battery performance.

[0107] Therefore, when the real-time cell temperature exceeds the maximum cell temperature or does not exceed the minimum cell temperature, the system will generate a pre-cooling command. The pre-cooling command is an adjustment suggestion generated based on the real-time cell temperature, and its specific content includes:

[0108] Cooling power adjustment: The cooling power of the liquid cooling system is dynamically adjusted based on the difference between the real-time cell temperature and the maximum cell temperature. If the temperature difference is large, the cooling power is increased; if the temperature difference is small, the cooling power is increased appropriately.

[0109] Circulation speed adjustment: The circulation speed of the coolant is dynamically adjusted based on the difference between the real-time cell temperature and the maximum cell temperature. If the temperature difference is large, the circulation speed is increased; if the temperature difference is small, the circulation speed is increased moderately.

[0110] Activation of auxiliary cooling devices: If the cell temperature continues to rise and existing measures are insufficient to cope with it, it is recommended to activate additional auxiliary cooling devices, such as fans or external coolers.

[0111] Energy-saving measures: When the cell temperature is lower than the set minimum temperature, it is recommended to take measures to reduce unnecessary cooling, such as reducing the operating frequency of the compressor, slowing down the speed of the circulating pump, or even shutting down some cooling modules.

[0112] Step S305: Integrate the pre-cooling command and the liquid cooling system operating parameters to generate the target control command.

[0113] It's important to note that the pre-cooling command includes specific adjustment suggestions based on real-time cell temperature, such as increasing cooling power, accelerating coolant circulation, or activating auxiliary cooling devices. Meanwhile, the liquid cooling system operating parameters provide the current system's actual operating status, including compressor operating frequency, circulation pump speed, and fan operating mode. These two pieces of information need to be integrated into a unified target control command set. During integration, target control commands are generated based on the adjustment suggestions in the pre-cooling command and the current liquid cooling system operating parameters. For example, if the pre-cooling command suggests increasing cooling power from 80% to 90%, and the current compressor operating frequency is at its default value, the target control command will explicitly instruct the compressor frequency to be increased to 90%. If it's necessary to accelerate coolant circulation to enhance cooling, the target control command will set a specific circulation pump speed (e.g., from 2000 RPM to 2500 RPM) to ensure the coolant can more efficiently remove heat. When additional auxiliary cooling devices need to be activated, the target control command will include specific commands to turn on the fan or external cooler and set the corresponding operating parameters (e.g., fan speed, cooler power).

[0114] This embodiment determines the highest and lowest cell temperatures by real-time monitoring of the battery cell temperature and combining this with a dynamic temperature control range. Based on the liquid cooling system adjustment parameters, target cooling power and circulation speed are set. When the cell temperature exceeds the set range, a pre-cooling command is generated and integrated into the target control command, achieving precise thermal management and ensuring the cells are always within their optimal operating temperature range, thus improving system efficiency and stability.

[0115] Based on the first embodiment of this application, this application also provides a thermal management control device for an energy storage system based on deep learning. Please refer to... Figure 5 The device includes:

[0116] The acquisition module 10 is used to collect data on cell temperature, ambient temperature, charging and discharging power, and liquid cooling system operating conditions.

[0117] Processing module 20 is used to input cell temperature, ambient temperature, charging and discharging power, and liquid cooling system operating condition data into a preset temperature control model to obtain dynamic temperature control range and liquid cooling system adjustment parameters. The preset temperature control model is a hybrid neural network model, which includes a spatial feature extraction layer, a temporal feature extraction layer, a feature fusion layer, and a dynamic decision layer. The spatial feature extraction layer includes a first convolutional block and a second convolutional block, each of which contains two one-dimensional convolutional layers, a batch normalization layer, and an activation function layer. The temporal feature extraction layer includes a long short-term memory network, an attention layer, and a fully connected layer. The long short-term memory network includes a three-layer bidirectional long short-term memory sub-network. The dynamic decision layer includes a temperature control sub-network and a liquid cooling adjustment sub-network. The temperature control sub-network outputs the dynamic temperature control range through the three-layer fully connected network, and the liquid cooling adjustment sub-network outputs the liquid cooling system adjustment parameters through a reinforcement learning strategy.

[0118] The instruction generation module 30 is used to generate target control instructions based on the dynamic temperature control range and the liquid cooling system adjustment parameters.

[0119] The sending module 40 is used to send target control commands to the liquid cooling system so that the liquid cooling system can make adjustments to complete thermal management control.

[0120] The deep learning-based thermal management control device for energy storage systems provided in this application employs the deep learning-based thermal management control method for energy storage systems described in the above embodiments, and can solve the technical problem of how to improve the safety and reliability of energy storage systems. Compared with the prior art, the beneficial effects of the deep learning-based thermal management control device for energy storage systems provided in this application are the same as those of the deep learning-based thermal management control method for energy storage systems provided in the above embodiments, and other technical features in the deep learning-based thermal management control device for energy storage systems are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0121] In one embodiment, the processing module 20 is further configured to input the cell temperature and ambient temperature into the time-series feature extraction layer of the preset temperature control model for analysis to obtain the cell temperature change trend features; input the charging and discharging power and liquid cooling system operating condition data into the spatial feature extraction layer of the preset temperature control model for pattern recognition to determine the priority features of the current operating condition; splice the cell temperature change trend features and the priority features of the current operating condition through the feature fusion layer of the preset temperature control model to obtain fused features; and pass the fused features through the dynamic decision layer of the preset temperature control model to obtain the dynamic temperature control range and liquid cooling system adjustment parameters.

[0122] In one embodiment, the processing module 20 is further configured to obtain a time series based on the cell temperature and the ambient temperature; input the time series into a long short-term memory network to generate time series features; perform weighted processing on the time series features through an attention layer to obtain key temperature change features; and input the key temperature change features into a fully connected layer to obtain cell temperature change trend features for a preset duration.

[0123] In one embodiment, the processing module 20 is further configured to input the charging and discharging power and liquid cooling system operating condition data into the spatial feature extraction layer to extract spatial features; classify the spatial features to obtain the current operating condition; match the current operating condition to obtain the operating condition priority label; and convert the operating condition priority label to obtain the current operating condition priority feature.

[0124] In one embodiment, the processing module 20 is further configured to: input the fused features into the temperature control subnetwork of the dynamic decision layer to extract primary features related to the temperature control range; the primary features are obtained by preliminary processing of the fused features through a first fully connected network in the temperature control subnetwork; input the primary features into a second fully connected network in the temperature control subnetwork for further processing to extract intermediate features related to the temperature control range; input the intermediate features into a third fully connected network in the temperature control subnetwork for processing to output a dynamic temperature control range, which is used to guide the real-time adjustment range of the battery temperature; input the fused features into the liquid cooling regulation subnetwork of the dynamic decision layer to generate an initial strategy; the initial strategy is obtained by processing the fused features through a policy network in the reinforcement learning strategy in the liquid cooling regulation subnetwork; optimize the initial strategy based on a preset reward function to obtain an optimized regulation strategy; and calculate and generate liquid cooling system regulation parameters according to the optimized regulation strategy.

[0125] In one embodiment, the instruction generation module 30 is further configured to acquire the real-time cell temperature of the battery; determine the highest and lowest cell temperatures based on the dynamic temperature control range; determine the liquid cooling system operating parameters based on the liquid cooling system adjustment parameters, including the target cooling power and the target circulation speed; generate a pre-cooling instruction when the real-time cell temperature exceeds the highest cell temperature or does not exceed the lowest cell temperature; and integrate the pre-cooling instruction and the liquid cooling system operating parameters to generate a target control instruction.

[0126] In one embodiment, the sending module 40 is further configured to receive feedback signals from the liquid cooling system; parse the feedback signals to obtain actual operating parameters; compare and analyze the actual operating parameters with the liquid cooling system operating parameters in the target control command to obtain comparison results; when there is a deviation in the comparison results, obtain the deviation situation and generate a new adjustment command to send to the liquid cooling system for adjustment.

[0127] This application provides a deep learning-based thermal management control device for an energy storage system. The deep learning-based thermal management control device for an energy storage system includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the deep learning-based thermal management control method for an energy storage system in the first embodiment described above.

[0128] The following is for reference. Figure 6 This document illustrates a structural schematic diagram of a deep learning-based thermal management control device suitable for implementing embodiments of this application. The deep learning-based thermal management control device for energy storage systems in these embodiments may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 6 The deep learning-based thermal management control device for energy storage systems shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0129] like Figure 6As shown, the deep learning-based energy storage system thermal management control device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the deep learning-based energy storage system thermal management control device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the deep learning-based energy storage system thermal management control device to communicate wirelessly or wiredly with other devices to exchange data. Although various deep learning-based energy storage system thermal management control devices are shown in the figures, it should be understood that implementation or possession of all shown is not required. More or fewer may be implemented alternatively.

[0130] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable storage medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0131] The deep learning-based thermal management control device for energy storage systems provided in this application, employing the deep learning-based thermal management control method for energy storage systems described in the above embodiments, can solve the technical problem of how to improve the safety and reliability of energy storage systems. Compared with the prior art, the beneficial effects of the deep learning-based thermal management control device for energy storage systems provided in this application are the same as those of the deep learning-based thermal management control method for energy storage systems provided in the above embodiments, and other technical features of this deep learning-based thermal management control device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0132] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0133] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0134] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the deep learning-based energy storage system thermal management control method in the above embodiments.

[0135] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible storage medium containing or storing a program that can be executed by instructions, used by a device, or used in conjunction with it. The program code contained on the computer-readable storage medium may be transmitted using any suitable storage medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0136] The aforementioned computer-readable storage medium may be included in a deep learning-based energy storage system thermal management control device; or it may exist independently and not be assembled into a deep learning-based energy storage system thermal management control device.

[0137] The aforementioned computer-readable storage medium carries one or more programs that, when executed by a deep learning-based energy storage system thermal management control device, enable the deep learning-based energy storage system thermal management control device to write computer program code for performing the operations of this application in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0138] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using dedicated hardware-based implementations that perform the specified functions or operations, or can be implemented using a combination of dedicated hardware and computer instructions.

[0139] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0140] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described deep learning-based thermal management control method for energy storage systems, thereby solving the technical problem of how to improve the safety and reliability of energy storage systems. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the deep learning-based thermal management control method for energy storage systems provided in the above embodiments, and will not be repeated here.

[0141] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the deep learning-based thermal management control method for energy storage systems as described above.

[0142] The computer program product provided in this application can solve the technical problem of how to improve the safety and reliability of energy storage systems. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the deep learning-based thermal management control method for energy storage systems provided in the above embodiments, and will not be repeated here.

[0143] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A thermal management control method for an energy storage system based on deep learning, characterized in that, include: Collect data on cell temperature, ambient temperature, charging and discharging power, and liquid cooling system operating conditions; The cell temperature, ambient temperature, charging / discharging power, and liquid cooling system operating data are input into a preset temperature control model to obtain a dynamic temperature control range and liquid cooling system adjustment parameters. The preset temperature control model is a hybrid neural network model, which includes a spatial feature extraction layer, a temporal feature extraction layer, a feature fusion layer, and a dynamic decision layer. The spatial feature extraction layer includes a first convolutional block and a second convolutional block, each containing two one-dimensional convolutional layers, a batch normalization layer, and an activation function layer. The temporal feature extraction layer includes a long short-term memory network, an attention layer, and a fully connected layer. The long short-term memory network includes a three-layer bidirectional long short-term memory sub-network. The dynamic decision layer includes a temperature control sub-network and a liquid cooling adjustment sub-network. The temperature control sub-network outputs a dynamic temperature control range through the three-layer fully connected network, and the output of the dynamic temperature control range predicts the temperature change trend over a future period. The liquid cooling adjustment sub-network outputs liquid cooling system adjustment parameters through a reinforcement learning strategy. Based on the dynamic temperature control range and the liquid cooling system adjustment parameters, a target control command is generated; The target control command is sent to the liquid cooling system to cause the liquid cooling system to adjust and complete the thermal management control; The step of inputting the cell temperature, ambient temperature, charging and discharging power, and liquid cooling system operating condition data into a preset temperature control model to obtain the dynamic temperature control range and liquid cooling system adjustment parameters includes: The cell temperature and the ambient temperature are input into the time-series feature extraction layer of the preset temperature control model for analysis to obtain the cell temperature change trend characteristics. The charging and discharging power and the liquid cooling system operating data are input into the spatial feature extraction layer of the preset temperature control model for pattern recognition to determine the priority features of the current operating condition. The cell temperature change trend feature and the current operating condition priority feature are spliced ​​together through the feature fusion layer in the preset temperature control model to obtain the fused feature; The fused features are passed through the dynamic decision layer in the preset temperature control model to obtain the dynamic temperature control range and liquid cooling system adjustment parameters.

2. The method as described in claim 1, characterized in that, The step of inputting the cell temperature and the ambient temperature into the time-series feature extraction layer of a preset temperature control model for analysis to obtain the cell temperature change trend characteristics includes: A time series is obtained based on the cell temperature and the ambient temperature; The time series data is input into a long short-term memory network to generate time series features; The time-series features are weighted through an attention layer to obtain key temperature change features; The key temperature change characteristics are input into the fully connected layer to obtain the cell temperature change trend characteristics for a preset time period.

3. The method as described in claim 1, characterized in that, The step of inputting the charging and discharging power and the liquid cooling system operating condition data into the spatial feature extraction layer of the preset temperature control model for pattern recognition to determine the priority features of the current operating condition includes: The charging and discharging power and the liquid cooling system operating data are input into the spatial feature extraction layer to extract spatial features. The spatial features are classified to obtain the current operating conditions; Based on the current operating conditions, a condition priority label is obtained through matching. The current operating condition priority feature is obtained by converting the operating condition priority label.

4. The method as described in claim 1, characterized in that, The step of obtaining the dynamic temperature control range and liquid cooling system adjustment parameters by passing the fused features through the dynamic decision layer in the preset temperature control model includes: The fused features are input into the temperature control sub-network of the dynamic decision layer to extract primary features related to the temperature control range. The primary features are obtained by the temperature control sub-network through a first fully connected network to perform preliminary processing on the fused features. The primary features are input into the second fully connected layer of the temperature control sub-network for further processing to extract intermediate features related to the temperature control range. The intermediate features are input into the third fully connected network of the temperature control subnetwork for processing, and a dynamic temperature control range is output. The dynamic temperature control range is used to guide the real-time adjustment range of the battery temperature. The fused features are input into the liquid-cooled regulation subnetwork of the dynamic decision layer to generate an initial policy. The initial policy is obtained by processing the fused features through the policy network in the reinforcement learning policy via the liquid-cooled regulation subnetwork. Based on the initial strategy and a preset reward function, an optimized adjustment strategy is obtained. The liquid cooling system adjustment parameters are generated based on the optimized adjustment strategy.

5. The method as described in claim 1, characterized in that, The step of generating a target control command based on the dynamic temperature control range and the liquid cooling system adjustment parameters includes: Obtain the real-time cell temperature of the battery; The highest and lowest cell temperatures are determined based on the dynamic temperature control range. Based on the liquid cooling system adjustment parameters, the operating parameters of the liquid cooling system are determined, including the target cooling power and the target circulation speed. A pre-cooling command is generated when the real-time cell temperature exceeds the maximum cell temperature or does not exceed the minimum cell temperature. The pre-cooling command and the liquid cooling system operating parameters are integrated to generate the target control command.

6. The method as described in claim 1, characterized in that, After the step of sending the target control command to the liquid cooling system to enable the liquid cooling system to adjust and complete the thermal management control, the following steps are included: Receive feedback signals from the liquid cooling system; The feedback signal is analyzed to obtain the actual operating parameters; The actual operating parameters and the liquid cooling system operating parameters in the target control command are compared and analyzed to obtain the comparison results; When there is a deviation in the comparison results, the deviation is obtained and a new adjustment command is generated and sent to the liquid cooling system for adjustment.

7. A thermal management control device for an energy storage system based on deep learning, characterized in that, The device is applied to the deep learning-based thermal management control method for energy storage systems as described in any one of claims 1-6, and the device comprises: The acquisition module is used to collect data on cell temperature, ambient temperature, charging and discharging power, and liquid cooling system operating conditions. The processing module is used to input the cell temperature, ambient temperature, charging and discharging power, and liquid cooling system operating condition data into a preset temperature control model to obtain a dynamic temperature control range and liquid cooling system adjustment parameters. The preset temperature control model is a hybrid neural network model, which includes a spatial feature extraction layer, a temporal feature extraction layer, a feature fusion layer, and a dynamic decision layer. The spatial feature extraction layer includes a first convolutional block and a second convolutional block, each of which contains two one-dimensional convolutional layers, a batch normalization layer, and an activation function layer. The temporal feature extraction layer includes a long short-term memory network, an attention layer, and a fully connected layer. The long short-term memory network includes a three-layer bidirectional long short-term memory sub-network. The dynamic decision layer includes a temperature control sub-network and a liquid cooling adjustment sub-network. The temperature control sub-network outputs the dynamic temperature control range through the three-layer fully connected network, and the liquid cooling adjustment sub-network outputs the liquid cooling system adjustment parameters through a reinforcement learning strategy. The instruction generation module is used to generate target control instructions based on the dynamic temperature control range and the liquid cooling system adjustment parameters; The sending module is used to send the target control command to the liquid cooling system so that the liquid cooling system can adjust and complete the thermal management control.

8. A thermal management control device for an energy storage system based on deep learning, characterized in that, The device includes: a memory, a processor, and a deep learning-based thermal management control program for an energy storage system stored in the memory and running on the processor, the deep learning-based thermal management control program for an energy storage system configured to implement the steps of the deep learning-based thermal management control method for an energy storage system as described in any one of claims 1-6.

9. A storage medium, characterized in that, The storage medium stores a deep learning-based thermal management control program for an energy storage system. When the deep learning-based thermal management control program is executed by a processor, it implements the steps of the deep learning-based thermal management control method for an energy storage system as described in any one of claims 1-6.

Images