Intelligent energy storage temperature control optimization method based on fuzzy control

Abstract

The invention discloses an intelligent energy storage temperature control optimization method based on fuzzy control, and the method comprises the following steps: S1, collecting the operation parameters of an energy storage unit, constructing a state vector set, and generating a heat influence graph structure; s2, normalizing parameters of each dimension of the state vector, and generating a graph structure input graph in combination with the heat influence graph; s3, constructing a graph structure fuzzy neural network model comprising a fuzzy input layer, a graph attention propagation layer, a fuzzy rule layer and a control output layer based on the graph structure input graph, and outputting a control instruction set; s4, executing the control instruction to obtain a temperature control execution result; s5, acquiring a temperature control execution result and a control instruction to form a supervision sample, and updating the fuzzy membership function and the graph attention propagation weight to obtain an updated model; and S6, performing cycle evaluation on the model, and if a performance evaluation value is lower than a threshold value, re-estimating a thermal coupling edge weight, and reconstructing a graph structure input graph. According to the method, the graph structure fuzzy neural network is adopted, and accurate cooperation and self-adaptive optimization of energy storage and temperature control are achieved.

Description

Technical Field

[0001] The present invention relates to the field of smart energy management technology, and in particular to a smart energy storage temperature control optimization method based on fuzzy control. Background Art

[0002] Amid the rapid development of new energy systems, energy storage systems have become a key component in supporting renewable energy consumption, improving grid regulation capabilities, and ensuring energy supply stability. As the scale of battery energy storage systems continues to expand, operational safety, stability, and economical issues are becoming increasingly prominent. As a core auxiliary module in energy storage systems, temperature control systems directly impact the lifespan, performance consistency, and thermal safety of battery cells. To this end, the industry generally uses temperature sensors to collect temperature data at the cell or module level, dynamically adjusting the system through air cooling, liquid cooling, or heating devices to ensure that the system's operating temperature remains within a safe range.

[0003] Traditional energy storage temperature control strategies are mostly based on rule-driven control or PID closed-loop control. Some schemes introduce fuzzy control methods to adapt to the nonlinear characteristics and uncertainty effects in energy storage working conditions. The fuzzy control system realizes logical reasoning of multiple input states by building an empirical rule base, thereby improving the robustness of the control system. However, the above control methods generally rely on manual experience to set fuzzy rules and membership functions, lack structural adaptability, and are difficult to cope with the complex thermal coupling relationship changes between multi-node energy storage units. In addition, the existing methods mostly treat the temperature control objects as independent individuals, without considering the thermal influence diffusion path between units under spatial layout. This often leads to the failure of control instructions at the overall level and delayed or imbalanced local temperature control responses in energy storage systems with high integration, high density, and close coupling of cooling channels.

[0004] In recent years, graph neural networks have been gradually applied to the modeling of energy systems with structural topological information, propagating relationships between nodes through adjacent structural modeling. Some studies have attempted to introduce graph structures into battery management systems to characterize thermal diffusion or state consistency issues. However, most related work focuses on state prediction or data completion and has not yet been deeply integrated with fuzzy control models, making it difficult to directly affect the generation of actual control instructions. At the same time, existing graph neural network models lack a mechanism to handle fuzzy and uncertain states in temperature control scenarios, and cannot effectively express the impact of imprecise information on control results, limiting their application value in fine temperature control.

[0005] Therefore, how to provide a smart energy storage temperature control optimization method based on fuzzy control is an urgent problem that technicians in this field need to solve. Summary of the Invention

[0006] One purpose of the present invention is to propose an intelligent energy storage temperature control optimization method based on fuzzy control. The present invention fully integrates the information propagation mechanism of thermal influence diagram modeling, fuzzy logic reasoning and graph neural network, and describes in detail the entire process from multi-source state acquisition, thermal coupling weight calculation, graph structure input construction to fuzzy reasoning control instruction generation. It has the advantages of high structural modeling accuracy, strong control robustness and strong adaptive optimization capability.

[0007] According to an embodiment of the present invention, a smart energy storage temperature control optimization method based on fuzzy control includes the following steps:

[0008] S1. Collect the operating parameters of each energy storage unit in the energy storage system, construct a state vector set, calculate the thermal coupling relationship weight based on the physical layout between the energy storage units, and construct a thermal influence diagram structure;

[0009] S2. Perform normalization processing on the parameters of each dimension in the state vector set to generate a normalized state matrix, and construct a graph structure input graph together with the heat influence map structure;

[0010] S3. Based on the graph structure input graph, construct a graph structure fuzzy neural network model, wherein the graph structure fuzzy neural network model includes a fuzzy input layer, a graph attention propagation layer, a fuzzy rule layer, and a control output layer. The graph attention propagation layer is used to perform cross-node fuzzy state propagation based on edge weights in the heat influence graph structure and output a control instruction set.

[0011] S4. Based on the control instruction set, perform adjustment operations to obtain temperature control execution results;

[0012] S5. Collect the actual temperature feedback data corresponding to the temperature control execution result, and form a supervision sample with the control instruction set. Input the data into the graph structure fuzzy neural network model, update the fuzzy membership function parameters and graph attention propagation weights based on the error back propagation algorithm, and output the updated graph structure fuzzy neural network model.

[0013] S6. Periodically evaluate the updated graph structure fuzzy neural network model and calculate the performance evaluation value. If the performance evaluation value is lower than the set threshold, the edge weight of the heat impact graph structure is re-estimated and the graph structure input graph is reconstructed.

[0014] Optionally, the S1 specifically includes:

[0015] S11. Collecting real-time operating parameters of each energy storage unit in the energy storage system to construct the state vector, where the state vector includes the unit temperature, ambient temperature, state of charge, current load power, current change rate, and predicted electricity price;

[0016] S12, summarizing the state vectors of all energy storage units to form the state vector set X;

[0017] S13. Based on the shared heat conduction surface between the energy storage units in the energy storage system, the heat flux density distribution between each pair of energy storage units is simulated and calculated using finite element thermal simulation technology to obtain the heat flux density function and calculate the thermal coupling weight:

[0018]

[0019] Among them, ω ij represents the thermal coupling weight between energy storage unit i and energy storage unit j, A represents the normalized area constant of the normalized heat flow integral result, S ij Indicates the boundary area where there is direct thermal coupling between energy storage unit i and energy storage unit j, k ij (x,y) represents the effective thermal conductivity at the boundary point (x,y), represents the symbol of partial derivative, i and j represent the index numbers of the two energy storage units respectively, T represents the temperature, and n represents the unit normal direction vector at the boundary;

[0020] S14. Construct all energy storage units and their corresponding thermal coupling weights into a graph structure G = (V, E, W), where V = {v1, v2, ..., v n} represents the node set composed of energy storage units, Indicates the edge set with heat-affected connection relationship, W={ω ij} represents the set of thermal coupling weights, forming the thermal influence diagram structure.

[0021] Optionally, the S2 specifically includes:

[0022] S21. Performing dimensional normalization processing on each state vector in the state vector set X, and mapping each dimensional parameter to the interval [0, 1] using a minimum-maximum normalization method;

[0023] S22, recombining all normalized state vectors to form a normalized state matrix X norm ;

[0024] S23, the normalized state matrix X norm Perform joint modeling with the heat influence diagram structure G;

[0025] S24, constructing a graph structure input graph G based on the normalized state matrix and the heat influence map structure X .

[0026] Optionally, the S3 specifically includes:

[0027] S31. Input graph G based on graph structure XConstructing a graph-structured fuzzy neural network model, wherein the graph-structured fuzzy neural network model includes a fuzzy input layer, a graph attention propagation layer, a fuzzy rule layer, and a control output layer;

[0028] S32. In the fuzzy input layer, a fuzzy membership value is set for each dimension parameter in the normalized state vector of each energy storage unit node. The membership function adopts the form of a Gaussian function:

[0029]

[0030] in, represents the membership value of the kth normalized state vector of the i-th energy storage unit node to the j-th fuzzy set, i represents the index number of the energy storage unit node, j represents the fuzzy set number, k represents the dimension index in the state vector, exp(·) represents the natural exponential function, represents the value of the k-th dimension parameter in the normalized state vector of the energy storage unit node i, c k,j The Gaussian function center value of the jth fuzzy set representing the kth dimension parameter, σ k,j The standard deviation of the Gaussian function corresponding to the j-th fuzzy set of the k-th dimension parameter;

[0031] S33. In the graph attention propagation layer, based on the thermal coupling weights in the heat influence graph structure, the fuzzy membership information of adjacent nodes is weightedly propagated and feature aggregated;

[0032] S34. In the fuzzy rule layer, define the combination rules between the fuzzy states of multiple nodes, establish the mapping relationship between the fuzzy conditions and the control output, and generate the fuzzy reasoning results;

[0033] S35. In the control output layer, a control instruction set is output based on the fuzzy reasoning result and the feature information after graph propagation aggregation. The control instruction set includes an air cooling intensity control signal, a liquid cooling pump speed adjustment signal, a heating start instruction or a power current limiting parameter.

[0034] Optionally, the S33 specifically includes:

[0035] S331, obtaining the node set V and the edge weight set thermal coupling weight set W in the heat influence graph structure G, and determining the neighboring energy storage unit node set of each energy storage unit node;

[0036] S332. Perform graph attention propagation update on the fuzzy feature vector of each energy storage unit node in the graph structure input graph:

[0037]

[0038] in, Represents the energy storage unit node v iThe output feature vector after attention propagation in the t+1th layer, σ(·) represents the nonlinear activation function, i represents the index number of the energy storage unit node, and j represents the energy storage unit node v i The neighbor node index number, represents the set of neighboring energy storage unit nodes, t represents the propagation layer index of the graph structure fuzzy neural network, Represents the energy storage unit node v in the tth layer of the graph propagation i For neighboring energy storage unit node v j The allocated attention coefficient, W (t) represents the learnable weight matrix used by the t-th layer of graph attention propagation, Represents the neighboring energy storage unit node v j Fuzzy feature representation at layer t;

[0039] S333, calculate the energy storage unit node v according to the correlation between the thermal coupling weight and the fuzzy characteristics between the energy storage unit nodes i With neighboring energy storage unit node v j The attention coefficient between them is used to measure the energy storage unit node v in the process of graph attention propagation j For energy storage unit node v i The information contribution degree of the attention coefficient is normalized within the neighborhood of each node to ensure that the information aggregation result has local selectivity under the thermal topology constraint;

[0040] S334. Repeat the updating process of the graph attention propagation layer for all energy storage unit nodes, and output a graph structure feature set that integrates the fuzzy states of the neighbors for use by the fuzzy rule layer.

[0041] Optionally, the S34 specifically includes:

[0042] S341, fusing the graph structure feature set of the fused neighbor fuzzy states with the fuzzy membership vector corresponding to the energy storage unit node to construct a fuzzy state combination input for each energy storage unit node;

[0043] S342. Based on the fuzzy state combination input, a multi-input fuzzy inference rule is established on each energy storage unit node. The fuzzy inference rule includes antecedent conditions and consequent conclusions:

[0044]

[0045] Among them, R r represents the rth fuzzy inference rule, Represents the energy storage unit node v i The normalized monomer temperature parameter, Represents the fuzzy inference rule R r Input variables The fuzzy set label to which it belongs, Represents the energy storage unit node v i The normalized state of charge parameter, Represents the fuzzy inference rule R r Input variables The fuzzy set label to which it belongs, Represents the energy storage unit node v i Normalized current load power parameter, Represents the fuzzy inference rule R r Input variables The fuzzy set label to which it belongs, u i Represents the energy storage unit node v i The control output variable, B r Indicates the control output variable u in the rth rule i The corresponding fuzzy set label;

[0046] S343. For each fuzzy inference rule, use a product-type membership calculation method to calculate the activation strength of the fuzzy rule;

[0047] S344. Perform weighted synthesis on all fuzzy rule activation intensities and corresponding output fuzzy set labels to generate a fuzzy output result.

[0048] Optionally, the temperature control execution result includes temperature change, cooling energy consumption change and current limiting power change.

[0049] Optionally, the S5 specifically includes:

[0050] S51, collecting the temperature control execution results of each energy storage unit node and forming a supervision sample pair together with the control instructions generated by the control output layer;

[0051] S52. Based on each group of supervised samples, construct the error evaluation function of the graph structure fuzzy neural network:

[0052]

[0053] in, represents the loss function value of the entire graph structure fuzzy neural network model, n represents the total number of energy storage unit nodes, i represents the index number of the energy storage unit node, α1 is the weighted coefficient of the temperature accuracy term, Represents the energy storage unit node v i The desired target temperature value, Represents the energy storage unit node v i The feedback temperature value after the actual temperature control is executed, α2 is the weighted coefficient of the energy consumption item, Represents the energy storage unit node v iThe change in cooling energy consumption during air cooling or liquid cooling, α3 is the weighted coefficient of the power control item, Indicates the degree of influence of power control on system operation;

[0054] S53, using an error back propagation algorithm to update the trainable parameters in the graph structure fuzzy neural network, where the updated content includes fuzzy membership function parameters, graph attention propagation weight matrix, and attention coefficient parameters;

[0055] S54. After completing the parameter update, construct an updated graph structure fuzzy neural network model for the next round of control instruction generation and model closed-loop optimization.

[0056] Optionally, the S6 specifically includes:

[0057] S61. After the graph structure fuzzy neural network model operation cycle ends, based on the temperature control execution results and target performance requirements of each energy storage unit, calculate the comprehensive performance evaluation value η of the graph structure fuzzy neural network model;

[0058] S62, comparing the comprehensive performance evaluation value η with the preset performance threshold η th Compare, if η<η th , then it is determined that the performance of the current graph structure fuzzy neural network model has degraded, triggering the model structure adaptive update operation;

[0059] S63. In the structure update operation, based on the performance correlation of the thermal coupling weights in the heat influence diagram structure of the current temperature control execution result, re-estimate the thermal coupling weights between each pair of energy storage units and update them;

[0060] S64. According to the updated thermal coupling weight set, the graph structure is input into the graph and graph structure fuzzy neural network model to form a new model version after structural adaptive optimization for the next round of control process.

[0061] Optionally, the calculation of the comprehensive performance evaluation value:

[0062]

[0063] Where η represents the comprehensive performance evaluation value, n represents the total number of energy storage units participating in the evaluation, i represents the index number of the energy storage unit node, and β1 is the temperature accuracy weighting coefficient. Represents the energy storage unit node v i The desired target temperature value, Represents the energy storage unit node v i Feedback temperature value after actual temperature control is executed, ΔT max It represents the maximum tolerance value of temperature deviation set by the system, β2 is the energy consumption weighting coefficient, Represents the energy storage unit node vi The change in cooling energy consumption during air cooling or liquid cooling, E max It represents the maximum cooling energy consumption reference value allowed by the system, β3 is the weighted coefficient of power control, Indicates the impact of power control on system operation, P max Indicates the maximum allowable current limiting power value set by the system.

[0064] The beneficial effects of the present invention are:

[0065] The intelligent energy storage temperature control collaborative optimization method proposed by the present invention, which introduces a graph-structured fuzzy neural network, has achieved significant improvements in the thermal management and control accuracy of multi-node energy storage units compared to existing temperature control solutions based on independent fuzzy control or graph neural modeling. First, the present invention can accurately reflect the thermal coupling relationship between different units in the energy storage system by constructing a thermal influence diagram structure based on the physical layout of the energy storage unit and the heat conduction path, avoiding the problem of local overheating or response lag caused by the traditional method of ignoring the influence of heat diffusion between nodes. Secondly, the fuzzy input layer introduces a Gaussian membership function to process the normalized multi-dimensional state vector, so that the system can cope with the nonlinear and fuzzy characteristics in the working conditions, and enhance the adaptability of the control model in complex operating environments.

[0066] The graph attention propagation layer integrates the thermal coupling weights and node fuzzy state information, realizes the feature aggregation of neighbor fuzzy states, and breaks through the limitation of traditional fuzzy control's lack of structural modeling capabilities. This structure further enhances the spatial perception capability of information propagation, enabling control instructions to be dynamically adjusted according to the overall thermal distribution of the system. Through the combination of the fuzzy rule layer and the control output layer, the present invention implements a fuzzy reasoning control strategy based on graph structure perception, thereby making the output temperature control instructions more accurate and globally coordinated. In addition, the supervisory feedback training mechanism and structural adaptive optimization module introduced by the present invention can actively adjust the fuzzy membership function and graph structure edge weights when the control execution effect is poor, realize dynamic optimization of the control model, and ensure that the control accuracy and system efficiency are in the optimal range for a long time.

[0067] In summary, the present invention not only improves the intelligence level of the energy storage temperature control system, but also significantly enhances its robustness and control stability in multi-node coupling and dynamically changing environments. It has comprehensive advantages such as strong structural expression ability, self-consistent control logic, and feedback closed-loop optimization, and has good engineering application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0069] Figure 1This is a flow chart of a smart energy storage temperature control optimization method based on fuzzy control proposed by the present invention;

[0070] Figure 2 This is a diagram showing the relationship between graph structure input construction and graph structure fuzzy neural network input mapping in a fuzzy control-based intelligent energy storage temperature control optimization method proposed in the present invention;

[0071] Figure 3 This is a schematic diagram of the model structure adaptive optimization process in the intelligent energy storage temperature control optimization method based on fuzzy control proposed in the present invention. DETAILED DESCRIPTION

[0072] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0073] refer to Figure 1-3 , a smart energy storage temperature control optimization method based on fuzzy control, comprising the following steps:

[0074] S1. Collect the operating parameters of each energy storage unit in the energy storage system, construct a state vector set, calculate the thermal coupling relationship weight based on the physical layout between the energy storage units, and construct a thermal influence diagram structure;

[0075] S2. Perform normalization processing on the parameters of each dimension in the state vector set to generate a normalized state matrix, and construct a graph structure input graph together with the heat influence map structure;

[0076] S3. Based on the graph structure input graph, construct a graph structure fuzzy neural network model, wherein the graph structure fuzzy neural network model includes a fuzzy input layer, a graph attention propagation layer, a fuzzy rule layer, and a control output layer. The graph attention propagation layer is used to perform cross-node fuzzy state propagation based on edge weights in the heat influence graph structure and output a control instruction set.

[0077] S4. Based on the control instruction set, perform adjustment operations to obtain temperature control execution results;

[0078] S5. Collect the actual temperature feedback data corresponding to the temperature control execution result, and form a supervision sample with the control instruction set. Input the data into the graph structure fuzzy neural network model, update the fuzzy membership function parameters and graph attention propagation weights based on the error back propagation algorithm, and output the updated graph structure fuzzy neural network model.

[0079] S6. Periodically evaluate the updated graph structure fuzzy neural network model and calculate the performance evaluation value. If the performance evaluation value is lower than the set threshold, the edge weight of the heat impact graph structure is re-estimated and the graph structure input graph is reconstructed.

[0080] This invention constructs a fuzzy-control-based intelligent energy storage temperature control optimization method and introduces a graph-structured fuzzy neural network model. This method integrates structured modeling of the thermal coupling relationship between multiple energy storage units with fuzzy reasoning control, breaking through the limitations of traditional fuzzy control, which ignores heat diffusion paths and lacks adaptive capabilities. By using a graph attention propagation mechanism to enhance the cross-node perception of fuzzy information, and combining supervised feedback training with a structural self-optimization strategy, the control model can dynamically adapt to changes in operating conditions, significantly improving temperature control accuracy, consistency, and energy efficiency, and possessing strong engineering practical value and system robustness.

[0081] In this embodiment, S1 specifically includes:

[0082] S11. Collecting real-time operating parameters of each energy storage unit in the energy storage system to construct the state vector, where the state vector includes the unit temperature, ambient temperature, state of charge, current load power, current change rate, and predicted electricity price;

[0083] S12, summarizing the state vectors of all energy storage units to form the state vector set X;

[0084] S13. Based on the shared heat conduction surface between the energy storage units in the energy storage system, the heat flux density distribution between each pair of energy storage units is simulated and calculated using finite element thermal simulation technology to obtain the heat flux density function and calculate the thermal coupling weight:

[0085]

[0086] Among them, ω ij represents the thermal coupling weight between energy storage unit i and energy storage unit j, A represents the normalized area constant of the normalized heat flow integral result, S ij Indicates the boundary area where there is direct thermal coupling between energy storage unit i and energy storage unit j, k ij (x,y) represents the effective thermal conductivity at the boundary point (x,y), represents the symbol of partial derivative, i and j represent the index numbers of the two energy storage units respectively, T represents the temperature, and n represents the unit normal direction vector at the boundary;

[0087] S14. Construct all energy storage units and their corresponding thermal coupling weights into a graph structure G = (V, E, W), where V = {v1, v2, ..., v n} represents the node set composed of energy storage units, Indicates the edge set with heat-affected connection relationship, W={ω ij} represents the set of thermal coupling weights, forming the thermal influence diagram structure.

[0088] The present invention models the heat conduction process between the units in the energy storage system, calculates the thermal coupling weights between them based on the integral results of the heat flux density between the physical boundaries of the energy storage units, and constructs a graph structure that reflects the heat diffusion path and intensity. This structure not only takes into account thermophysical factors such as the actual temperature gradient, thermal conductivity, and contact area, but also effectively captures the spatial thermal influence range, providing accurate structural input for temperature control decisions. On this basis, combined with the fuzzy reasoning mechanism and the graph structure propagation method, local precise adjustment and global coordinated control of the temperature control instructions are achieved, improving the system's adjustment responsiveness, temperature balance, and energy efficiency under complex multi-node working conditions.

[0089] In this embodiment, S2 specifically includes:

[0090] S21. Performing dimensional normalization processing on each state vector in the state vector set X, and mapping each dimensional parameter to the interval [0, 1] using a minimum-maximum normalization method;

[0091] S22, recombining all normalized state vectors to form a normalized state matrix X norm ;

[0092] S23, the normalized state matrix X norm Perform joint modeling with the heat influence diagram structure G;

[0093] S24, constructing a graph structure input graph G based on the normalized state matrix and the heat influence map structure X .

[0094] The present invention normalizes the state parameters of each unit of the energy storage system and maps the multi-dimensional heterogeneous state quantities to a unified numerical interval, thereby improving the comparability and stability of the input data in the fuzzy reasoning process. The normalized state vectors are recombined into a state matrix and jointly modeled with the thermal coupling graph structure to construct a graph structure input graph with dual information of structural attributes and node attributes. As the basic input of the fuzzy neural network, this input graph not only retains the key state characteristics of each node, but also embeds the thermal coupling topological relationship between nodes, effectively improving the accuracy and global coordination of temperature control decisions.

[0095] In this embodiment, S3 specifically includes:

[0096] S31. Input graph G based on graph structure X Constructing a graph-structured fuzzy neural network model, wherein the graph-structured fuzzy neural network model includes a fuzzy input layer, a graph attention propagation layer, a fuzzy rule layer, and a control output layer;

[0097] S32. In the fuzzy input layer, a fuzzy membership value is set for each dimension parameter in the normalized state vector of each energy storage unit node. The membership function adopts the form of a Gaussian function:

[0098]

[0099] in, represents the membership value of the kth normalized state vector of the i-th energy storage unit node to the j-th fuzzy set, i represents the index number of the energy storage unit node, j represents the fuzzy set number, k represents the dimension index in the state vector, exp(·) represents the natural exponential function, represents the value of the k-th dimension parameter in the normalized state vector of the energy storage unit node i, c k,j The Gaussian function center value of the jth fuzzy set representing the kth dimension parameter, σ k,j The standard deviation of the Gaussian function corresponding to the j-th fuzzy set of the k-th dimension parameter;

[0100] S33. In the graph attention propagation layer, based on the thermal coupling weights in the heat influence graph structure, the fuzzy membership information of adjacent nodes is weightedly propagated and feature aggregated;

[0101] S34. In the fuzzy rule layer, define the combination rules between the fuzzy states of multiple nodes, establish the mapping relationship between the fuzzy conditions and the control output, and generate the fuzzy reasoning results;

[0102] S35. In the control output layer, a control instruction set is output based on the fuzzy reasoning result and the feature information after graph propagation aggregation. The control instruction set includes an air cooling intensity control signal, a liquid cooling pump speed adjustment signal, a heating start instruction or a power current limiting parameter.

[0103] The present invention introduces a graph-structured fuzzy neural network when constructing a temperature control reasoning model, performs fuzzy processing on each normalized state parameter through the fuzzy input layer, and introduces a membership function based on Gaussian distribution to make the fuzzy partitioning continuous and differentiable, thereby enhancing the expression accuracy and training stability of the model. The fusion of cross-node fuzzy information is achieved through the graph attention propagation mechanism, which can introduce the thermal coupling influence of neighboring nodes while maintaining the state characteristics of local nodes, thereby improving the coordination of the overall temperature control instructions. In the fuzzy rule layer, the system constructs multi-conditional fuzzy inference rules based on the combination relationship between fuzzy sets, and generates specific control instructions including air cooling wind speed, liquid cooling flow rate or current limiting strategy in the output layer, effectively supporting multi-strategy collaborative regulation.

[0104] In this embodiment, the S33 specifically includes:

[0105] S331, obtaining the node set V and the edge weight set thermal coupling weight set W in the heat influence graph structure G, and determining the neighboring energy storage unit node set of each energy storage unit node;

[0106] S332. Perform graph attention propagation update on the fuzzy feature vector of each energy storage unit node in the graph structure input graph:

[0107]

[0108] in, Represents the energy storage unit node v i The output feature vector after attention propagation in the t+1th layer, σ(·) represents the nonlinear activation function, i represents the index number of the energy storage unit node, and j represents the energy storage unit node v i The neighbor node index number, represents the set of neighboring energy storage unit nodes, t represents the propagation layer index of the graph structure fuzzy neural network, Represents the energy storage unit node v in the tth layer of the graph propagation i For neighboring energy storage unit node v j The allocated attention coefficient, W (t) represents the learnable weight matrix used by the t-th layer of graph attention propagation, Represents the neighboring energy storage unit node v j Fuzzy feature representation at layer t;

[0109] S333, calculate the energy storage unit node v according to the correlation between the thermal coupling weight and the fuzzy characteristics between the energy storage unit nodes i With neighboring energy storage unit node v j The attention coefficient between them is used to measure the energy storage unit node v in the process of graph attention propagation j For energy storage unit node v i The information contribution degree of the attention coefficient is normalized within the neighborhood of each node to ensure that the information aggregation result has local selectivity under the thermal topology constraint;

[0110] S334. Repeat the updating process of the graph attention propagation layer for all energy storage unit nodes, and output a graph structure feature set that integrates the fuzzy states of the neighbors for use by the fuzzy rule layer.

[0111] During the graph attention propagation process, the present invention clarifies the neighbor relationship of the energy storage unit by extracting the node set and edge weight set in the heat influence graph, and introduces the attention mechanism and learnable weights in the information update process to achieve weighted aggregation of the fuzzy features of the neighbor nodes. The state of each energy storage node is not only determined by its own characteristics, but also integrates the fuzzy information of the surrounding units, thereby enhancing the model's perception of local thermal coupling effects. The attention allocation strategy based on thermal coupling strength and feature similarity enables information propagation to have thermophysical responsiveness and structural adaptability. After completing feature aggregation, the fused graph structure feature set is output to provide high-quality input for the fuzzy reasoning layer, ensuring that the control decision has spatial coordination and dynamic adaptability.

[0112] In this embodiment, the S34 specifically includes:

[0113] S341, fusing the graph structure feature set of the fused neighbor fuzzy states with the fuzzy membership vector corresponding to the energy storage unit node to construct a fuzzy state combination input for each energy storage unit node;

[0114] S342. Based on the fuzzy state combination input, a multi-input fuzzy inference rule is established on each energy storage unit node. The fuzzy inference rule includes antecedent conditions and consequent conclusions:

[0115]

[0116] Among them, R r represents the rth fuzzy inference rule, Represents the energy storage unit node v i The normalized monomer temperature parameter, Represents the fuzzy inference rule R r Input variables The fuzzy set label to which it belongs, Represents the energy storage unit node v i The normalized state of charge parameter, Represents the fuzzy inference rule R r Input variables The fuzzy set label to which it belongs, Represents the energy storage unit node v i Normalized current load power parameter, Represents the fuzzy inference rule R r Input variables The fuzzy set label to which it belongs, u i Represents the energy storage unit node v i The control output variable, B r Indicates the control output variable u in the rth rule i The corresponding fuzzy set label;

[0117] S343. For each fuzzy inference rule, use a product-type membership calculation method to calculate the activation strength of the fuzzy rule;

[0118] S344. Perform weighted synthesis on all fuzzy rule activation intensities and corresponding output fuzzy set labels to generate a fuzzy output result.

[0119] The present invention introduces structured fuzzy state combination inputs in the fuzzy rule layer, and constructs a fuzzy reasoning rule system under multiple antecedent conditions based on normalized key variables such as temperature, state of charge and power. Each fuzzy rule clearly stipulates the fuzzy set label to which the state input belongs, and corresponds to the fuzzy set of output control instructions, forming a clearly structured and computable fuzzy control logic. By weighted fusion of the membership degrees of all activated rules, the system can generate fuzzy control outputs that meet multi-state constraints, effectively improving the personalization and adaptability of the temperature control strategy. At the same time, the reasoning structure supports fuzzy rule strength evaluation and dynamic adjustment, making the control behavior more stable and flexible under complex working conditions.

[0120] In this embodiment, the temperature control execution result includes the temperature change, the cooling energy consumption change and the current limiting power change.

[0121] The present invention comprehensively reflects the control effect and energy consumption cost by quantifying the temperature changes, cooling energy consumption and current limiting power in the temperature control execution results, provides accurate feedback for model training and performance evaluation, and enhances the system's adaptive optimization capabilities.

[0122] In this embodiment, the S5 specifically includes:

[0123] S51, collecting the temperature control execution results of each energy storage unit node and forming a supervision sample pair together with the control instructions generated by the control output layer;

[0124] S52. Based on each group of supervised samples, construct the error evaluation function of the graph structure fuzzy neural network:

[0125]

[0126] in, represents the loss function value of the entire graph structure fuzzy neural network model, n represents the total number of energy storage unit nodes, i represents the index number of the energy storage unit node, α1 is the weighted coefficient of the temperature accuracy term, Represents the energy storage unit node v i The desired target temperature value, Represents the energy storage unit node v i The feedback temperature value after the actual temperature control is executed, α2 is the weighted coefficient of the energy consumption item, Represents the energy storage unit node v iThe change in cooling energy consumption during air cooling or liquid cooling, α3 is the weighted coefficient of the power control item, Indicates the degree of influence of power control on system operation;

[0127] S53, using an error back propagation algorithm to update the trainable parameters in the graph structure fuzzy neural network, where the updated content includes fuzzy membership function parameters, graph attention propagation weight matrix, and attention coefficient parameters;

[0128] S54. After completing the parameter update, construct an updated graph structure fuzzy neural network model for the next round of control instruction generation and model closed-loop optimization.

[0129] The present invention collects the temperature control execution results and corresponding control instructions of the energy storage unit to construct a supervision sample, establishes a weighted loss function that includes temperature control error, cooling energy consumption cost, and power current limiting amplitude, and comprehensively evaluates the control effect and system cost. This loss function can quantify the balance between control accuracy and resource utilization efficiency, and jointly trains the fuzzy membership function and graph attention weights through the error backpropagation algorithm. Through continuous supervision and feedback, the present invention can achieve synchronous self-learning of model structure and control parameters, significantly improving the adaptability and long-term stability of the control strategy to complex environments, and supporting closed-loop self-optimization of graph-structured fuzzy neural networks.

[0130] In this embodiment, S6 specifically includes:

[0131] S61. After the graph structure fuzzy neural network model operation cycle ends, based on the temperature control execution results and target performance requirements of each energy storage unit, calculate the comprehensive performance evaluation value η of the graph structure fuzzy neural network model;

[0132] S62, comparing the comprehensive performance evaluation value η with the preset performance threshold η th Compare, if η<η th , then it is determined that the performance of the current graph structure fuzzy neural network model has degraded, triggering the model structure adaptive update operation;

[0133] S63. In the structure update operation, based on the performance correlation of the thermal coupling weights in the heat influence diagram structure of the current temperature control execution result, re-estimate the thermal coupling weights between each pair of energy storage units and update them;

[0134] S64. According to the updated thermal coupling weight set, the graph structure is input into the graph and graph structure fuzzy neural network model to form a new model version after structural adaptive optimization for the next round of control process.

[0135] The present invention determines the fitness level of the current model by calculating a comprehensive performance evaluation value of the control effect and energy consumption after the model operation cycle. If the performance is insufficient, the system will automatically re-evaluate the thermal coupling relationship between the energy storage units, adjust the edge weight configuration in the heat impact diagram, reconstruct the graph structure input, and generate an updated neural network model for the next round of training and control. This mechanism enables the temperature control system to have structural self-adaptation capabilities, dynamically respond to environmental changes and operational fluctuations, and achieve long-term, efficient, and stable coordinated control.

[0136] In this embodiment, the comprehensive performance evaluation value is calculated as follows:

[0137]

[0138] Where η represents the comprehensive performance evaluation value, n represents the total number of energy storage units participating in the evaluation, i represents the index number of the energy storage unit node, and β1 is the temperature accuracy weighting coefficient. Represents the energy storage unit node v i The desired target temperature value, Represents the energy storage unit node v i Feedback temperature value after actual temperature control is executed, ΔT max It represents the maximum tolerance value of temperature deviation set by the system, β2 is the energy consumption weighting coefficient, Represents the energy storage unit node v i The change in cooling energy consumption during air cooling or liquid cooling, E max It represents the maximum cooling energy consumption reference value allowed by the system, β3 is the weighted coefficient of power control, Indicates the impact of power control on system operation, P max Indicates the maximum allowable current limiting power value set by the system.

[0139] This invention constructs a comprehensive performance evaluation value, normalizing and weighting three indicators: temperature control accuracy, cooling energy consumption, and power current limiting. This comprehensive measure measures the control effectiveness and resource efficiency of the energy storage system. This evaluation mechanism dynamically reflects the model's adaptability and optimization level under actual operating conditions, providing a quantitative basis for subsequent structural adjustments and parameter updates, thereby achieving closed-loop self-optimization and continuous performance improvement.

[0140] Example 1:

[0141] In order to verify the feasibility of the present invention in practice, the present invention was applied to the temperature control optimization task of a certain energy storage power station. The power station is composed of hundreds of lithium iron phosphate energy storage units, which are compactly arranged and have a significant thermal coupling effect. In actual operation, the power station has long adopted fixed threshold control and fuzzy control methods based on expert rules. However, due to the lack of inter-node heat conduction modeling and adaptive control capabilities, the system is prone to problems such as excessive temperature of some battery cells, poor battery consistency, and excessive cooling energy consumption under high temperature and peak load operating conditions in summer.

[0142] To address the above issues, a graph-structured fuzzy neural network model was constructed in the power station to replace the original fuzzy control module. First, based on the physical layout of the energy storage battery and the thermal simulation results, a thermal impact diagram structure containing 416 energy storage nodes was established, and the thermal coupling weights between adjacent units were calculated. The operating parameters include single cell temperature, state of charge, load power, current change rate and ambient temperature, a total of 6-dimensional state vectors, which are used as input after normalization. A three-membership function is used for modeling in the fuzzy input layer, and three levels of low, medium and high are set for each parameter. The graph attention propagation layer uses the thermal coupling weight as the attention guide, integrates the fuzzy state information of the neighboring nodes, and then outputs the temperature control strategy through the fuzzy rule layer.

[0143] Control commands are applied to the air-cooling fan, liquid-cooling circulation pump, and power management module, respectively, to adjust air speed, coolant flow rate, and maximum charge and discharge power limits. The system completes a complete state acquisition, control inference, and execution feedback loop every two minutes. The experiment involved a high-temperature scenario with 72 hours of continuous operation, during which the ambient temperature remained between 32°C and 36°C, and the load periodically fluctuated between 60% and 95%.

[0144] By comparing the operating data of the original control system and the system of the present invention under the same conditions, it was found that the control system of the present invention effectively reduced the temperature peak of the battery pack, with the maximum temperature dropping from 45.3°C to 41.7°C, and the standard deviation of the temperature difference dropping from 6.1°C to 3.4°C; the cooling energy consumption was reduced by an average of 12.7% per hour, from 2.37kWh to 2.07kWh; without affecting the output capacity, excessive current limiting was avoided through graph structure propagation, and the number of current limiting starts was reduced by 28%; after the system automatically trained and adjusted the membership function, the control deviation converged to within 0.8°C within 48 hours.

[0145] Based on the experimental results, the following table is compiled to demonstrate the advantages of the present invention in terms of temperature control accuracy, energy consumption performance, and execution stability.

[0146] Table 1 Comparison of temperature control performance between graph structure fuzzy neural network and traditional fuzzy control

[0147]

[0148] As can be seen from the above table, the present invention has significant improvements over traditional fuzzy control methods in multiple key performance indicators of the smart energy storage temperature control system. First, in terms of temperature control accuracy, the present invention uses a graph-structured fuzzy neural network to achieve joint modeling and reasoning of multi-node states, significantly reducing the overall thermal drift of the system. Under the original fuzzy control, the maximum temperature of the battery pack can reach 45.3°C, while the system of the present invention effectively suppresses local overheating through thermal coupling perception and graph attention propagation mechanism, so that the maximum temperature is controlled within 41.7°C, a decrease of 3.6°C. At the same time, the standard deviation of the temperature difference dropped from 6.1°C to 3.4°C, a reduction of more than 44%, indicating that the temperature control of each node tends to be consistent and the heat distribution is more balanced.

[0149] In terms of energy consumption performance, the present invention adaptively optimizes the cooling strategy through a fuzzy rule layer, and combines execution feedback to update the graph structure, making the allocation of cooling resources more targeted and efficient. Test results show that the hourly cooling energy consumption is reduced from 2.37kWh to 2.07kWh, with an energy saving of 12.7%, which reduces the system operating costs while ensuring thermal safety. In addition, the graph structure propagation mechanism effectively reduces unnecessary power current limiting responses, and the number of current limiting starts is reduced from 39 to 28, a reduction of 28.2%, which helps to improve the continuity of the battery system output capacity and the stability of user-side energy supply.

[0150] It is worth noting that the present invention realizes the periodic self-update of fuzzy membership functions and thermal coupling structures by introducing supervised training and performance evaluation mechanisms. Compared with the inherent limitation of traditional systems that cannot adapt to changes in operating conditions, the model of the present invention can automatically complete structural optimization once every 24 hours, and in actual applications, the control deviation can converge to within ±1°C within 48 hours, with stable and fast self-regulation capabilities. Based on the above data results, the present invention shows significant advantages in control accuracy, energy efficiency performance and model stability, and is suitable for the intelligent thermal management needs of large-scale energy storage systems.

[0151] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A smart energy storage temperature control optimization method based on fuzzy control, characterized in that: The steps include: S1. Collect the operating parameters of each energy storage unit in the energy storage system, construct a state vector set, calculate the thermal coupling relationship weight based on the physical layout between the energy storage units, and construct a thermal influence diagram structure; S2. Perform normalization processing on the parameters of each dimension in the state vector set to generate a normalized state matrix, and construct a graph structure input graph together with the heat influence map structure; S3. Based on the graph structure input graph, a graph structure fuzzy neural network model is constructed. The graph structure fuzzy neural network model includes a fuzzy input layer, a graph attention propagation layer, a fuzzy rule layer, and a control output layer. The graph attention propagation layer performs cross-node fuzzy state propagation based on edge weights in the heat influence graph structure and outputs a set of control instructions. S4. Based on the control instruction set, perform adjustment operations to obtain temperature control execution results; S5. Collect the actual temperature feedback data corresponding to the temperature control execution result, and form a supervision sample with the control instruction set. Input the data into the graph structure fuzzy neural network model, update the fuzzy membership function parameters and graph attention propagation weights based on the error back propagation algorithm, and output the updated graph structure fuzzy neural network model. S6. Periodically evaluate the updated graph structure fuzzy neural network model and calculate the performance evaluation value. If the performance evaluation value is lower than the set threshold, the edge weight of the heat impact graph structure is re-estimated and the graph structure input graph is reconstructed.

2. The intelligent energy storage temperature control optimization method based on fuzzy control according to claim 1 is characterized in that: Said S1 specifically includes: S11. Collecting real-time operating parameters of each energy storage unit in the energy storage system and constructing a state vector, wherein the state vector includes the unit temperature, ambient temperature, state of charge, current load power, current change rate, and predicted electricity price; S12. Summarize the state vectors of all energy storage units to form a state vector set X; S13. Based on the shared heat conduction surface between the energy storage units in the energy storage system, the heat flux density distribution between each pair of energy storage units is simulated and calculated using finite element thermal simulation technology to obtain the heat flux density function and calculate the thermal coupling weight: Among them, ω ij represents the thermal coupling weight between energy storage unit i and energy storage unit j, A represents the normalized area constant of the normalized heat flow integral result, S ij Indicates the boundary area where there is direct thermal coupling between energy storage unit i and energy storage unit j, k ij (x,y) represents the effective thermal conductivity at the boundary point (x,y), represents the symbol of partial derivative, i and j represent the index numbers of the two energy storage units respectively, T represents the temperature, and n represents the unit normal direction vector at the boundary; S14. Construct all energy storage units and their corresponding thermal coupling weights into a graph structure G = (V, E, W), where V = {v1, v2, ..., v n } represents the node set composed of energy storage units, Indicates the edge set with heat-affected connection relationship, W={ω ij } represents the set of thermal coupling weights, forming the thermal influence diagram structure.

3. The intelligent energy storage temperature control optimization method based on fuzzy control according to claim 1 is characterized in that: The S2 specifically includes: S21. Perform dimensional normalization on each state vector in the state vector set X, and map each dimensional parameter to the interval [0, 1] using a minimum-maximum normalization method; S22, recombining all normalized state vectors to form a normalized state matrix X norm ; S23, normalize the state matrix X norm Joint modeling with the heat influence diagram structure G; S24. Based on the normalized state matrix and the heat impact diagram structure, construct the graph structure input graph G X .

4. The intelligent energy storage temperature control optimization method based on fuzzy control according to claim 1 is characterized in that: The S3 specifically includes: S31. Input graph G based on graph structure X Constructing a graph-structured fuzzy neural network model, wherein the graph-structured fuzzy neural network model includes a fuzzy input layer, a graph attention propagation layer, a fuzzy rule layer, and a control output layer; S32. In the fuzzy input layer, a fuzzy membership value is set for each dimension parameter in the normalized state vector of each energy storage unit node, and the membership function adopts the form of a Gaussian function; S33. In the graph attention propagation layer, based on the thermal coupling weights in the heat influence graph structure, the fuzzy membership information of adjacent nodes is weightedly propagated and feature aggregated; S34. In the fuzzy rule layer, define the combination rules between the fuzzy states of multiple nodes, establish the mapping relationship between the fuzzy conditions and the control output, and generate the fuzzy reasoning results; S35. In the control output layer, a control instruction set is output based on the fuzzy reasoning result and the feature information after graph propagation aggregation. The control instruction set includes an air cooling intensity control signal, a liquid cooling pump speed adjustment signal, a heating start instruction or a power current limiting parameter.

5. The intelligent energy storage temperature control optimization method based on fuzzy control according to claim 4 is characterized in that: The S33 specifically includes: S331, obtaining the node set V and the edge weight set thermal coupling weight set W in the heat influence graph structure G, and determining the neighboring energy storage unit node set of each energy storage unit node; S332. Perform graph attention propagation update on the fuzzy feature vector of each energy storage unit node in the graph structure input graph: in, Represents the energy storage unit node v i The output feature vector after attention propagation in the t+1th layer, σ(·) represents the nonlinear activation function, i represents the index number of the energy storage unit node, and j represents the energy storage unit node v i The neighbor node index number, represents the set of neighboring energy storage unit nodes, t represents the propagation layer index of the graph structure fuzzy neural network, Represents the energy storage unit node v in the tth layer of the graph propagation i For neighboring energy storage unit node v j The allocated attention coefficient, W (t) represents the learnable weight matrix used by the t-th layer of graph attention propagation, Represents the neighboring energy storage unit node v j Fuzzy feature representation at layer t; S333, calculate the energy storage unit node v according to the correlation between the thermal coupling weight and the fuzzy characteristics between the energy storage unit nodes i With neighboring energy storage unit node v j The attention coefficient between , where the attention coefficient is normalized within the neighborhood of each node; S334. Repeat the updating process of the graph attention propagation layer for all energy storage unit nodes, and output a graph structure feature set that integrates the fuzzy states of the neighbors.

6. The intelligent energy storage temperature control optimization method based on fuzzy control according to claim 4 is characterized in that: The S34 specifically includes: S341, fusing the graph structure feature set of the fused neighbor fuzzy states with the fuzzy membership vector corresponding to the energy storage unit node to construct a fuzzy state combination input for each energy storage unit node; S342. Based on the fuzzy state combination input, a multi-input fuzzy inference rule is established on each energy storage unit node. The fuzzy inference rule includes antecedent conditions and consequent conclusions: Among them, R r represents the rth fuzzy inference rule, Represents the energy storage unit node v i The normalized monomer temperature parameter, Represents the fuzzy inference rule R r Input variables The fuzzy set label to which it belongs, Represents the energy storage unit node v i The normalized state of charge parameter, Represents the fuzzy inference rule R r Input variables The fuzzy set label to which it belongs, Represents the energy storage unit node v i Normalized current load power parameter, Represents the fuzzy inference rule R r Input variables The fuzzy set label to which it belongs, u i Represents the energy storage unit node v i The control output variable, B r Indicates the control output variable u in the rth rule i The corresponding fuzzy set label; S343. For each fuzzy inference rule, use a product-type membership calculation method to calculate the activation strength of the fuzzy rule; S344. Perform weighted synthesis on all fuzzy rule activation intensities and corresponding output fuzzy set labels to generate a fuzzy output result.

7. The intelligent energy storage temperature control optimization method based on fuzzy control according to claim 1 is characterized in that: The temperature control execution result includes the temperature change, the cooling energy consumption change and the current limiting power change.

8. The intelligent energy storage temperature control optimization method based on fuzzy control according to claim 1 is characterized in that: The S5 specifically includes: S51, collecting the temperature control execution results of each energy storage unit node and forming a supervision sample pair together with the control instructions generated by the control output layer; S52. Based on each group of supervised samples, construct the error evaluation function of the graph structure fuzzy neural network: in, represents the loss function value of the entire graph structure fuzzy neural network model, n represents the total number of energy storage unit nodes, i represents the index number of the energy storage unit node, α1 is the weighted coefficient of the temperature accuracy term, Represents the energy storage unit node v i The desired target temperature value, Represents the energy storage unit node v i The feedback temperature value after the actual temperature control is executed, α2 is the weighted coefficient of the energy consumption item, Represents the energy storage unit node v i The change in cooling energy consumption during air cooling or liquid cooling, α3 is the weighted coefficient of the power control item, Indicates the degree of influence of power control on system operation; S53, using an error back propagation algorithm to update the trainable parameters in the graph structure fuzzy neural network, where the updated content includes fuzzy membership function parameters, graph attention propagation weight matrix, and attention coefficient parameters; S54. After completing the parameter update, construct an updated graph structure fuzzy neural network model.

9. The intelligent energy storage temperature control optimization method based on fuzzy control according to claim 1 is characterized in that: The S6 specifically includes: S61. After the operation cycle of the graph structure fuzzy neural network model ends, based on the temperature control execution results and target performance requirements of each energy storage unit, calculate the comprehensive performance evaluation value η of the graph structure fuzzy neural network model; S62, compare the comprehensive performance evaluation value η with the preset performance threshold η th Compare, if η<η th , then it is determined that the performance of the current graph structure fuzzy neural network model has degraded, triggering the model structure adaptive update operation; S63. In the structure update operation, based on the performance correlation of the thermal coupling weights in the heat influence diagram structure of the current temperature control execution result, re-estimate the thermal coupling weights between each pair of energy storage units and update them; S64. According to the updated thermal coupling weight set, the graph structure input graph is reconstructed and put into the graph structure fuzzy neural network model to form a new model version after structural adaptive optimization.

10. The intelligent energy storage temperature control optimization method based on fuzzy control according to claim 9 is characterized in that: The calculation of the comprehensive performance evaluation value includes: Where η represents the comprehensive performance evaluation value, n represents the total number of energy storage units participating in the evaluation, i represents the index number of the energy storage unit node, and β1 is the temperature accuracy weighting coefficient. Represents the energy storage unit node v i The desired target temperature value, Represents the energy storage unit node v i Feedback temperature value after actual temperature control is executed, ΔT max It represents the maximum tolerance value of temperature deviation set by the system, β2 is the energy consumption weighting coefficient, Represents the energy storage unit node v i The change in cooling energy consumption during air cooling or liquid cooling, E max It represents the maximum cooling energy consumption reference value allowed by the system, β3 is the weighted coefficient of power control, Indicates the impact of power control on system operation, P max Indicates the maximum allowable current limiting power value set by the system.

Images