Intelligent planning configuration management system for household energy storage system

Abstract

The invention relates to the technical field of energy storage and utilization, in particular to an intelligent planning configuration management system for a household energy storage system, which comprises a data acquisition and transmission unit, a data analysis and processing unit and a decision execution unit, the data acquisition and transmission unit acquires and transmits related data of a battery and cooling liquid through a sensor and a thermal imager, and the data analysis and processing unit preprocesses, fuses and screens the data by using a denoising algorithm based on quantum noise cancellation and a multi-physical field data fusion algorithm based on holographic entanglement. A thermal health factor is calculated to evaluate the thermal health state of the battery, meanwhile, model predictive control and a multi-stage early warning algorithm are used for optimizing parameters of a thermal management system and early warning thermal faults of the battery, and a decision execution unit controls a coolant pump and regulates the thermal state of the battery according to the optimized parameters, so that the problem of inaccurate data processing of an existing household energy storage system is effectively solved. The safety and stability of the system are improved, and the service life of the battery is prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage and utilization, and specifically, to an intelligent planning, configuration and management system for a household energy storage system. Background Art

[0002] Energy storage and utilization is an important technology, and the importance of household energy storage systems has become increasingly prominent. It provides a flexible and efficient solution for household energy management. With the continuous popularization of its application, the requirements for system stability and safety are becoming more and more stringent.

[0003] Due to the single means of collecting data in the traditional method, it is vulnerable to environmental interference. The data such as battery temperature and internal thermal strain collected are often mixed with a large amount of noise, seriously affecting the data quality. Before these raw data enter the thermal management system, there is a lack of effective preprocessing and fusion mechanisms, resulting in the isolation of data in different physical fields and being unable to comprehensively and accurately reflect the battery thermal state. This directly leads to missing and inaccurate information when the thermal management system obtains data. And the control effect of the thermal management system depends on accurate data. Due to the defects in the collected and processed data, the thermal management system cannot accurately regulate according to the real battery thermal state. When the internal thermal strain of the battery increases, indicating that the battery is about to have a thermal fault, but due to the lack of data collection and processing, it is impossible to adjust the rotation speed of the coolant pump and the opening degree of the valve in advance, making the battery thermal management in a lag state. If this continues for a long time, it will not only accelerate battery aging and reduce its service life, but also may cause potential safety hazards, seriously affecting the stable operation of the household energy storage system. To solve this technical problem, we provide an intelligent planning, configuration and management system for a household energy storage system. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent planning, configuration and management system for a household energy storage system to solve the problems raised in the above background art.

[0005] To achieve the above purpose, an intelligent planning, configuration and management system for a household energy storage system is provided, including a data acquisition and transmission unit, a data analysis and processing unit, and a decision-making and execution unit;

[0006] The data acquisition and transmission unit arranges sensors and thermal imagers on the energy storage battery module to monitor the battery temperature, internal thermal strain, battery surface temperature distribution image, and the pressure, flow rate and temperature of the coolant in real time, and transmits them to the data analysis and processing unit;

[0007] The data analysis and processing unit preprocesses the collected raw data through a data preprocessing module, and inputs the processed data into a multi-physical field data fusion module. The multi-physical field data fusion module fuses temperature, thermal strain, and surface temperature distribution data to construct a battery thermal state dataset, then extracts the fused data from the battery thermal state dataset and inputs it into a thermal health factor calculation module to calculate the thermal health factor to evaluate the battery thermal health state, and then transmits the thermal health factor and related data to a model predictive control module and a fault warning module. The model predictive control module optimizes the control parameters of the thermal management system based on the mathematical model of the battery thermal management system using a model predictive control algorithm, and feeds the results back to the decision-making and execution unit. The fault warning module uses a long short-term memory network model to perform multi-stage warnings on battery thermal faults, and judges the fault stage based on the data of the thermal health factor calculation module;

[0008] The thermal management system control module in the decision-making and execution unit controls the operation of the coolant pump according to the optimized control parameters to regulate the battery thermal state.

[0009] As a further improvement of this technical solution, the data preprocessing module of the data analysis and processing unit uses a denoising algorithm based on quantum noise cancellation to denoise the raw data. The specific steps are as follows:

[0010] Map the raw data to the quantum state space and construct the quantum representation of the data, and encode it into the superposition state of quantum bits according to the eigenvalues and eigenvectors of the data. Each quantum bit represents a dimensional information of the data;

[0011] Introduce a quantum noise model to simulate the noise introduced during data acquisition and transmission, use quantum error correction codes to perform error correction operations on the superposition state, correct the quantum state errors caused by noise, restore the original information of the data, decode the corrected quantum state, and convert it back to the classical data space to obtain the denoised data.

[0012] As a further improvement of this technical solution, the multi-physical field data fusion module of the data analysis and processing unit uses a multi-physical field data fusion algorithm based on holographic entanglement. The specific steps are as follows:

[0013] Encode the temperature, thermal strain, and surface temperature distribution data into the quantum entanglement state, and use the characteristics of quantum entanglement to establish a correlation relationship between the data of different physical fields;

[0014] Construct a holographic projection model, project the quantum entanglement state into the holographic space. In the holographic space, by designing a specific quantum circuit, realize the information exchange and integration of the data of different physical fields, perform an inverse projection on the evolved quantum state, and convert it back to the classical data space to obtain the fused battery thermal state dataset.

[0015] As a further improvement of this technical solution, after the multi-physical field data fusion module performs fusion, it also adopts a fusion data screening algorithm based on fractal feature matching. The specific steps are as follows:

[0016] Calculate the fractal features of the fused data. The fractal features are used to reflect the complex structure and self-similarity of the data. Construct a fractal feature library to store the fractal features of battery thermal state data under different types and states. The fractal feature library is obtained through the analysis and learning of historical data;

[0017] Match the calculated fractal features with the features in the fractal feature library, calculate the similarity score, and use the dynamic time warping algorithm to calculate the similarity between features. This algorithm is used to process feature sequences of different lengths and time scales;

[0018] Screen out the qualified fused data according to the similarity score and set a similarity threshold. Retain the data with a score higher than the threshold for the calculation of the thermal health factor.

[0019] As a further improvement of this technical solution, the thermal health factor calculation module of the data analysis and processing unit adopts a thermal health factor calculation method based on complex network dynamics. The specific steps are as follows:

[0020] Construct a complex network of battery thermal states, regard each monitoring point of the battery as a network node, and the connection between nodes represents the heat transfer and interaction relationship between monitoring points. The weight of the connection is determined according to the heat conduction coefficient and temperature difference between monitoring points;

[0021] Analyze the clustering coefficient of the complex network and define the mapping relationship between the thermal health factor and the dynamic characteristics of the complex network. The clustering coefficient is used to reflect the distribution and propagation law of the battery thermal state. Then, through a machine learning algorithm, learn the non-linear relationship between the thermal health factor and the dynamic characteristics, and calculate the thermal health factor according to the clustering coefficient of the current battery thermal state complex network and the use of the mapping relationship.

[0022] As a further improvement of this technical solution, the thermal health factor calculation module also adopts a thermal health factor correction algorithm based on Bayesian dynamic update. The specific steps are as follows:

[0023] Determine the prior probability of the thermal health factor in different states according to historical data, and establish the prior probability distribution of the thermal health factor. Then, calculate the likelihood function according to the battery thermal state data obtained in real time. The likelihood function represents the probability of observing the current data given the state of the thermal health factor;

[0024] The posterior probability distribution of the thermal health factor is updated using Bayes' formula, which combines the prior probability and the likelihood function to obtain the posterior probability distribution. The calculation result of the thermal health factor is corrected according to the updated posterior probability distribution, and finally the expected value of the posterior probability distribution is taken as the corrected thermal health factor.

[0025] As a further improvement of this technical solution, the model predictive control module of the data analysis and processing unit adopts a model predictive control algorithm based on quantum reinforcement learning, and the specific steps are as follows:

[0026] Define the quantum environment and the quantum state space, encode the state and environmental information of the battery thermal management system into quantum states to construct the quantum state space, and at the same time define the quantum action space, which includes the speed adjustment of the coolant pump and the opening adjustment of the valve;

[0027] Construct a quantum reinforcement learning agent and use the quantum variational algorithm to construct the policy network and value network of the agent. The agent interacts with the quantum environment, obtains environmental feedback through quantum measurement, the agent selects quantum actions according to the current quantum state, and the environment updates the quantum state according to the action and returns a reward signal;

[0028] Use the quantum gradient descent algorithm to update the network parameters of the agent, and finally perform model predictive control in the actual system according to the trained agent. The agent selects the optimal action according to the real-time quantum state and optimizes the control parameters of the thermal management system in real time.

[0029] As a further improvement of this technical solution, the fault warning module of the data analysis and processing unit adopts a multi-stage warning algorithm based on the spatio-temporal graph neural network, and the specific steps are as follows:

[0030] Construct a spatio-temporal graph of the battery thermal state, where each monitoring point of the battery is regarded as a node of the graph, the connection between nodes represents the spatio-temporal relationship between monitoring points, the features of nodes include temperature and thermal strain data, and the features of edges include thermal conductivity and time delay;

[0031] Design a spatio-temporal graph neural network architecture, which includes a spatial convolutional layer and a temporal convolutional layer. The spatial convolutional layer is used to extract the spatial features between nodes, and the temporal convolutional layer is used to extract the time series features of nodes;

[0032] Use historical data to train the spatio-temporal graph neural network, use the known fault stages as labels, train the network to learn the mapping relationship between the battery thermal state and the fault stages, input the real-time spatio-temporal graph of the battery thermal state into the trained network, output the predicted stage of the battery thermal fault, and set different warning levels according to different fault stages.

[0033] Compared with the prior art, the beneficial effects of the present invention:

[0034] In the intelligent planning, configuration and management system of a household energy storage system, the multi-physical field data fusion module uses a fusion algorithm based on holographic entanglement. By leveraging the characteristics of quantum entanglement, it correlates data from different physical fields and achieves deep fusion through holographic projection and specific quantum circuits, constructing a dataset that comprehensively and accurately reflects the thermal state of the battery. After fusion, a screening algorithm based on fractal feature matching is also used to screen out high-quality data for calculating the thermal health factor, enhancing the effectiveness of the data. The thermal health factor calculation module uses a method based on complex network dynamics and combines machine learning algorithms to accurately evaluate the thermal health state of the battery. Moreover, through a correction algorithm based on Bayesian dynamic update, the evaluation results are continuously optimized according to real-time data. The model predictive control module uses an algorithm based on quantum reinforcement learning to effectively improve the control ability of the battery thermal management system and intelligently optimize the control parameters. Brief Description of the Drawings

[0035] Figure 1 It is the overall block diagram of the present invention.

[0036] The meanings of each label in the figure are as follows:

[0037] 1. Data acquisition and transmission unit; 2. Data analysis and processing unit; 21. Data preprocessing module; 22. Multi-physical field data fusion module; 23. Thermal health factor calculation module; 24. Model predictive control module; 25. Fault warning module; 3. Decision execution unit; 31. Thermal management system control module. Detailed Embodiments

[0038] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0039] The present invention provides an intelligent planning, configuration and management system for a household energy storage system. Please refer to Figure 1 as shown, which includes a data acquisition and transmission unit 1, a data analysis and processing unit 2, and a decision execution unit 3;

[0040] The data acquisition and transmission unit 1 arranges sensors and thermal imagers on the energy storage battery module to monitor the battery temperature, internal thermal strain, the temperature distribution image on the battery surface, as well as the pressure, flow rate, and temperature of the coolant in real time, and transmits them to the data analysis and processing unit 2;

[0041] The data analysis and processing unit 2 preprocesses the collected raw data through the data preprocessing module 21, and inputs the processed data into the multi-physical field data fusion module 22. The multi-physical field data fusion module 22 fuses the temperature, thermal strain, and surface temperature distribution data to construct a battery thermal state dataset, and then extracts the fused data from the battery thermal state dataset and inputs it into the thermal health factor calculation module 23 to calculate the thermal health factor to evaluate the battery thermal health state. Then, the thermal health factor and related data are transmitted to the model predictive control module 24 and the fault warning module 25. The model predictive control module 24 optimizes the control parameters of the thermal management system based on the mathematical model of the battery thermal management system using the model predictive control algorithm, and feeds the results back to the decision-making and execution unit 3. The fault warning module 25 uses the long short-term memory network model to perform multi-stage warnings on battery thermal faults, and judges the fault stage based on the data of the thermal health factor calculation module 23.

[0042] The data preprocessing module 21 of the data analysis and processing unit 2 uses a denoising algorithm based on quantum noise cancellation to denoise the raw data. The specific steps are as follows:

[0043] The raw data has limitations in dealing with complex noise under traditional processing methods. It is mapped to the quantum state space, encoded into a superposition state of qubits according to the eigenvalues and eigenvectors of the data, and each qubit represents a dimensional information of the data, providing a basis for subsequent quantum noise processing, enabling it to utilize the advantages of quantum algorithms for denoising. Then, a quantum noise model is introduced to simulate the noise introduced during data acquisition and transmission. By introducing the quantum noise model, the influence of actual noise on the quantum state can be accurately simulated, providing a clear target for subsequent error correction operations. Finally, quantum error correction codes are used to perform error correction operations on the superposition state, correcting the quantum state errors caused by noise and restoring the original information of the data, enabling the quantum state after noise interference to be restored to a state close to the original, laying a foundation for obtaining accurate denoised data through subsequent decoding. The corrected quantum state is decoded and converted back to the classical data space to obtain the denoised data, improving the quality of the data.

[0044] The multi-physical field data fusion module 22 of the data analysis and processing unit 2 uses a multi-physical field data fusion algorithm based on holographic entanglement. The specific steps are as follows:

[0045] Traditional data fusion methods are difficult to fully explore the deep associations between data from different physical fields. Quantum entanglement has non-locality and strong correlation. Encoding temperature, thermal strain, and surface temperature distribution data into quantum entanglement states and utilizing the characteristics of quantum entanglement can establish correlation relationships between data from different physical fields, providing a good foundation for subsequent in-depth data fusion in the holographic space. Then, a holographic projection model is constructed to project the quantum entanglement state into the holographic space, obtaining the projected quantum state. In the holographic space, the information of data from different physical fields can be more conveniently interacted and integrated, improving the effect of data fusion. Finally, the inverse projection of the evolved quantum state is performed to convert it back to the classical data space, obtaining the fused battery thermal state dataset. This dataset comprehensively integrates the information of multi-physical field data such as temperature, thermal strain, and surface temperature distribution, more comprehensively and accurately reflecting the thermal state of the battery.

[0046] After fusion, the multi-physical field data fusion module 22 also adopts a fusion data screening algorithm based on fractal feature matching. The specific steps are as follows:

[0047] Calculate the fractal features of the fused data. Fractal features are used to reflect the complex structure and self-similarity of the data. The obtained fractal features can more accurately describe the characteristics of the fused data, facilitating subsequent precise matching with the feature library. To judge the effectiveness and reliability of the current fused data, a reference standard is needed. A fractal feature library is constructed to store the fractal features of battery thermal state data under different types and states. The fractal feature library is obtained through the analysis and learning of historical data. With the feature library, the fractal features of the current fused data can be compared with it to determine whether it conforms to the feature patterns under normal or specific states. Match the calculated fractal features with the features in the fractal feature library and calculate the similarity score. The dynamic time warping algorithm is used to calculate the similarity between features. This algorithm is used to process feature sequences of different lengths and time scales. By calculating the similarity score, the similarity degree between the fused data and the typical features in the feature library can be quantitatively evaluated, providing a clear quantitative index for subsequent data screening. Screen out the fused data that meets the requirements according to the similarity score and set the similarity threshold. Retain the data with a score higher than the threshold for calculating the thermal health factor, making the thermal health factor calculated based on these data more reliable and providing stronger support for the state monitoring and fault warning of the battery.

[0048] The thermal health factor calculation module 23 of the data analysis and processing unit 2 adopts a thermal health factor calculation method based on complex network dynamics. The specific steps are as follows:

[0049] The battery thermal state is a complex system, with mutual heat transfer and interaction among various monitoring points. By constructing a complex network of the battery thermal state, each monitoring point of the battery is regarded as a network node, and the connection between nodes represents the heat transfer and interaction relationship between monitoring points. The weight of the connection is determined according to the thermal conductivity coefficient and temperature difference between monitoring points. The constructed complex network provides an effective model basis for subsequent analysis of the distribution and propagation law of the battery thermal state. The clustering coefficient is an important index in the complex network, which can reflect the aggregation degree of nodes and local connection characteristics in the network. Analyze the clustering coefficient of the complex network and define the mapping relationship between the thermal health factor and the dynamic characteristics of the complex network. The clustering coefficient is used to reflect the distribution and propagation law of the battery thermal state. Then, through the machine learning algorithm, learn the non-linear relationship between the thermal health factor and the dynamic characteristics. The trained model can accurately calculate the thermal health factor according to the dynamic characteristics of the current battery thermal state complex network, providing a more scientific basis for the evaluation of the battery thermal state. Calculate the thermal health factor according to the clustering coefficient of the current battery thermal state complex network and using the mapping relationship, as follows:

[0050] Obtain the clustering coefficient of the current complex network of electro-thermal state in real time , and form an input feature vector , and take the input feature vector and substitute it into the trained machine learning model. By calculating , obtain the current value of the thermal health factor. By calculating the thermal health factor in real time, abnormal changes in the battery thermal state can be detected in time, and corresponding measures can be taken to ensure the safe and stable operation of the battery.

[0051] The thermal health factor calculation module 23 also adopts a thermal health factor correction algorithm based on Bayesian dynamic update, and the specific steps are as follows:

[0052] Determine the prior probability of the thermal health factor in different states according to historical data, and establish the prior probability distribution of the thermal health factor. Assume that the thermal health factor has different states . Collect a large amount of historical data, and count the number of times the thermal health factor is in each state in these historical data . The total quantity is . Then the prior probability of the thermal health factor in state can be calculated by the formula , and thus the prior probability distribution of the thermal health factor is established , it provides a reasonable initial probability distribution for subsequent updates based on Bayes' formula, which helps to more accurately correct the calculation results of the thermal health factor. Then, the likelihood function is calculated according to the battery thermal state data obtained in real time. The likelihood function represents the probability of observing the current data given the thermal health factor state. Let the electro-thermal state data obtained in real time be , for each thermal health factor state , it is necessary to determine the probability of observing the data in this state , that is, the likelihood function. If the thermal state data follows a normal distribution and the parameters of this distribution are known in the state and , then the likelihood function can be calculated according to the probability density function of the normal distribution, that is . By calculating the likelihood function, it provides the information of the current data for Bayesian update, making the subsequent posterior probability distribution update more accurate. The posterior probability distribution of the thermal health factor is updated using Bayes' formula. Bayes' formula combines the prior probability and the likelihood function to obtain the posterior probability distribution. Bayes' formula can integrate the prior probability and the likelihood information of the real-time data to obtain a more accurate posterior probability distribution, which reflects the probability of the thermal health factor being in different states after considering the new data. Specifically as follows:

[0053] According to Bayes' formula, the posterior probability that the thermal health factor is in the state is calculated as ; where is the prior probability, is the likelihood function, and the denominator is to ensure the normalization of the posterior probability distribution. Through this formula, the posterior probability is calculated for each thermal health factor state to obtain the updated posterior probability distribution . The obtained posterior probability distribution takes into account the influence of the real-time data and can more accurately reflect the state distribution of the current battery thermal health factor than the prior probability distribution. The calculation results of the thermal health factor are corrected according to the updated posterior probability distribution. Finally, the expected value of the posterior probability distribution is taken as the corrected thermal health factor, which can make the correction result fully consider various possible states and their probabilities to obtain a more reasonable estimated value. Let the thermal health factor value corresponding to each thermal health factor state be , and the expected value of the posterior probability distribution is calculated as . Taking as the corrected thermal health factor provides stronger support for the evaluation and decision-making of the battery thermal state.

[0054] The model predictive control module 24 of the data analysis and processing unit 2 adopts a model predictive control algorithm based on quantum reinforcement learning, and the specific steps are as follows:

[0055] Traditional reinforcement learning may have problems such as low computational efficiency and difficulty in capturing deep characteristics of the system when dealing with complex battery thermal management systems. Define a quantum environment and a quantum state space, encode the state and environmental information of the battery thermal management system into quantum states to construct the quantum state space, and at the same time define a quantum action space. The quantum action space includes the speed regulation of the coolant pump and the opening regulation of the valve, as follows:

[0056] Let the state information of the battery thermal management system include the battery temperature , the coolant temperature , the coolant pressure and the ambient temperature . Discretize these continuous state and environmental information. For each discretized state and environmental information, encode it into a superposition state of quantum bits to obtain the entire quantum state space . For the speed regulation of the coolant pump, discretize its speed range into gears, and each gear corresponds to a quantum action. For the opening regulation of the valve, discretize the opening range into levels, and each level corresponds to a quantum action . The quantum action space is composed of the combination of all coolant pump speed regulation actions and valve opening regulation actions, which improves the representation ability of the state and environmental information of the battery thermal management system and provides a richer information basis for the decision-making of the intelligent agent. Construct a quantum reinforcement learning intelligent agent and use the quantum variational algorithm to construct the policy network and value network of the intelligent agent. The intelligent agent interacts with the quantum environment, obtains environmental feedback through quantum measurement. The intelligent agent selects a quantum action according to the current quantum state, the environment updates the quantum state according to the action and returns a reward signal. The policy network and value network can be represented by variational quantum circuits. For the policy network, input the current quantum state , and obtain the output quantum action probability distribution through the evolution of the variational quantum circuit . The value network calculates the value estimate of the current state through the variational quantum circuit . Initially, the intelligent agent is in a quantum state . The intelligent agent selects a quantum action from the quantum action space according to the policy network . The environment updates the quantum state according to the action selected by the intelligent agent, and the environment returns a reward signal , the reward signal can be defined according to the performance metrics of the battery thermal management system. The agent can continuously adjust its decisions based on environmental feedback and gradually learn the strategy of choosing the optimal quantum action in different quantum states.

[0057] Use the quantum gradient descent algorithm to update the network parameters of the agent. Finally, perform model predictive control in the actual system based on the trained agent. The agent selects the optimal action according to the real-time quantum state and optimizes the control parameters of the thermal management system in real time, as follows:

[0058] Define the loss function , the quantum gradient descent algorithm updates the parameters by calculating the gradients of the loss function with respect to the variational parameters and . In the actual battery thermal management system, the state and environmental information of the system are obtained in real time and encoded as a quantum state . The trained agent selects the optimal quantum action from the quantum action space according to the policy network , and converts the optimal quantum action into actual control parameters, such as the specific rotation speed of the coolant pump and the specific opening of the valve, to perform real-time control on the thermal management system and make the battery thermal management system achieve better performance.

[0059] The fault warning module 25 of the data analysis and processing unit 2 adopts a multi-stage warning algorithm based on a spatio-temporal graph neural network, and the specific steps are as follows:

[0060] The battery thermal state is a complex system with spatio-temporal characteristics. There are spatial heat transfer and temporal dynamic changes among various monitoring points. Represent these relationships by constructing a spatio-temporal graph of the battery thermal state. Among them, each monitoring point of the battery is regarded as a node of the graph, and the connections between nodes represent the spatio-temporal relationships between monitoring points. The features of nodes include temperature and thermal strain data, and the features of edges include thermal conductivity and time delay. The spatio-temporal graph can comprehensively capture the spatio-temporal characteristics of the battery thermal state, transform complex physical relationships into a graph structure, and facilitate the use of graph neural networks for processing. The change of the battery thermal state is related to both the spatial relationship between monitoring points and the temporal dynamic change. Designing a spatial graph neural network can mine the spatial dependence relationship between nodes, and the temporal convolutional layer can capture the change law of node features over time. The combination of the two can more comprehensively extract spatio-temporal features and provide a powerful feature representation for accurately predicting the fault stage.

[0061] Use historical data to train the spatio-temporal graph neural network. Take the known fault stage as a label, and train the network to learn the mapping relationship between the battery thermal state and the fault stage, so as to be able to accurately predict the fault stage of unknown real-time data. Collect a large amount of historical battery thermal state data and construct a spatio-temporal graph sequence , where is the node feature matrix at the moment, is the edge feature tensor at the moment. Meanwhile, corresponding fault stage labels are marked for each spatio-temporal graph , and the fault stages can be divided into normal, early fault, intermediate fault, and severe fault. The cross-entropy loss function is adopted. For a batch of samples, the loss function is ; where is the -th component of the true fault stage label of the -th sample, is the number of categories of the fault stage. Stochastic gradient descent is used to update the parameters of the network. The trained network can accurately predict the stage of the battery thermal fault according to the input spatio-temporal graph, providing a reliable basis for early warning. The real-time spatio-temporal graph of the battery thermal state is input into the trained network, and the predicted stage of the battery thermal fault is output. Different early warning levels are set according to different fault stages. Real-time monitoring of the battery thermal state and timely early warning can help users take corresponding measures to avoid losses caused by battery faults. Real-time data such as the temperature and thermal strain of each monitoring point of the battery are obtained, and a real-time spatio-temporal graph of the battery thermal state is constructed. It is input into the trained spatio-temporal graph neural network to obtain the prediction result of the fault stage. Different early warning levels are set according to different fault stages.

[0062] When the prediction is in the normal stage, no early warning is issued. When the prediction is in the early fault stage, a low-level early warning is issued to prompt the user to pay attention to observing the battery state. When the prediction is in the intermediate fault stage, a medium-level early warning is issued, suggesting that the user take some preventive measures. When the prediction is in the severe fault stage, a high-level early warning is issued to remind the user to take immediate emergency measures. The user can take corresponding measures in a timely manner according to the early warning information to reduce the risk of battery faults and ensure the safe operation of the battery.

[0063] The thermal management system control module 31 in the decision execution unit 3 controls the operation of the coolant pump according to the optimized control parameters to regulate the battery thermal state.

[0064] In the present invention, the data acquisition and transmission unit 1 collects data related to the battery and coolant through sensors and thermal imagers and transmits the data. The data analysis and processing unit 2 preprocesses, fuses, and filters the data by using a denoising algorithm based on quantum noise cancellation and a multi-physical field data fusion algorithm based on holographic entanglement, calculates the thermal health factor to evaluate the battery thermal health state, and simultaneously optimizes the parameters of the thermal management system and warns of battery thermal faults by using model predictive control and multi-stage warning algorithms. The decision-making and execution unit 3 controls the coolant pump according to the optimized parameters to regulate the battery thermal state, effectively solving the problem of inaccurate data processing in existing household energy storage systems, improving the safety and stability of the system, and prolonging the service life of the battery.

[0065] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and the above embodiments and descriptions in the specification are only preferred examples of the present invention and do not limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and all these changes and improvements fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. A smart planning, configuration and management system for a household energy storage system, characterized in that, It includes a data acquisition and transmission unit (1), a data analysis and processing unit (2), and a decision execution unit (3); The data acquisition and transmission unit (1) arranges sensors and thermal imagers in the energy storage battery module to monitor the battery temperature, internal thermal strain, the battery surface temperature distribution image, as well as the pressure, flow rate, and temperature of the coolant in real time, and transmits them to the data analysis and processing unit (2); The data analysis and processing unit (2) preprocesses the collected raw data through a data preprocessing module (21), and inputs the processed data into a multi-physical field data fusion module (22). The multi-physical field data fusion module (22) fuses the temperature, thermal strain, and surface temperature distribution data to construct a battery thermal state dataset, then extracts the fusion data from the battery thermal state dataset and inputs it into a thermal health factor calculation module (23) to calculate the thermal health factor to evaluate the battery thermal health state, and then transmits the thermal health factor and related data to a model prediction control module (24) and a fault warning module (25). The model prediction control module (24) optimizes the control parameters of the thermal management system based on the battery thermal management system mathematical model using the model predictive control algorithm, and feeds the result back to the decision execution unit (3). The fault warning module (25) uses a long short-term memory network model to conduct multi-stage early warning of battery thermal faults, and judges the fault stage based on the data of the thermal health factor calculation module (23); The thermal management system control module (31) in the decision execution unit (3) controls the operation of the coolant pump according to the optimized control parameters to regulate the battery thermal state.

2. The intelligent planning, configuration and management system for a household energy storage system according to claim 1, wherein: The data preprocessing module (21) of the data analysis and processing unit (2) uses a denoising algorithm based on quantum noise cancellation to denoise the raw data. The specific steps are as follows: Map the raw data to the quantum state space and construct the quantum representation of the data. Encode it into the superposition state of quantum bits according to the eigenvalues and eigenvectors of the data, and each quantum bit represents a dimension information of the data; Introduce a quantum noise model to simulate the noise introduced during data acquisition and transmission. Use quantum error correction codes to perform error correction operations on the superposition state, correct the quantum state errors caused by noise, restore the original information of the data, decode the corrected quantum state, and convert it back to the classical data space to obtain the denoised data.

3. The intelligent planning, configuration and management system for a household energy storage system according to claim 1, wherein: The multi-physical field data fusion module (22) of the data analysis and processing unit (2) uses a multi-physical field data fusion algorithm based on holographic entanglement. The specific steps are as follows: Encode the temperature, thermal strain, and surface temperature distribution data into the quantum entanglement state, and use the characteristics of quantum entanglement to establish a correlation relationship between the data of different physical fields; Construct a holographic projection model, project the quantum entanglement state into the holographic space. In the holographic space, through designing a specific quantum circuit, realize the information exchange and integration of the data of different physical fields, perform inverse projection on the evolved quantum state, and convert it back to the classical data space to obtain the fused battery thermal state dataset.

4. The intelligent planning, configuration and management system for the household energy storage system according to claim 3, wherein: After fusion, the multi-physical field data fusion module (22) also uses a fusion data screening algorithm based on fractal feature matching. The specific steps are as follows: Calculate the fractal features of the fused data. The fractal features are used to reflect the complex structure and self-similarity of the data. Construct a fractal feature library to store the fractal features of battery thermal state data under different types and states. The fractal feature library is obtained through the analysis and learning of historical data; Match the calculated fractal features with the features in the fractal feature library, calculate the similarity score, and use the dynamic time warping algorithm to calculate the similarity between features. This algorithm is used to process feature sequences of different lengths and time scales; Filter out the fused data that meets the requirements according to the similarity score and set a similarity threshold. Retain the data with a score higher than the threshold for the calculation of the thermal health factor.

5. The intelligent planning, configuration and management system for a household energy storage system according to claim 1, wherein: The thermal health factor calculation module (23) of the data analysis and processing unit (2) adopts a thermal health factor calculation method based on complex network dynamics. The specific steps are as follows: Construct a complex network of battery thermal state. Regard each monitoring point of the battery as a network node. The connection between nodes represents the heat transfer and interaction relationship between monitoring points. The weight of the connection is determined according to the heat conduction coefficient and temperature difference between monitoring points; Analyze the clustering coefficient of the complex network and define the mapping relationship between the thermal health factor and the dynamic characteristics of the complex network. The clustering coefficient is used to reflect the distribution and propagation law of the battery thermal state. Then, learn the non-linear relationship between the thermal health factor and the dynamic characteristics through a machine learning algorithm, and calculate the thermal health factor according to the clustering coefficient of the current battery thermal state complex network and using the mapping relationship.

6. The intelligent planning, configuration and management system for a household energy storage system according to claim 5, characterized in that: The thermal health factor calculation module (23) also adopts a thermal health factor correction algorithm based on Bayesian dynamic update. The specific steps are as follows: Determine the prior probability of the thermal health factor in different states according to historical data and establish the prior probability distribution of the thermal health factor. Then, calculate the likelihood function according to the real-time obtained battery thermal state data. The likelihood function represents the probability of observing the current data given the state of the thermal health factor; Use Bayes' formula to update the posterior probability distribution of the thermal health factor. Bayes' formula combines the prior probability and the likelihood function to obtain the posterior probability distribution. Correct the calculation result of the thermal health factor according to the updated posterior probability distribution. Finally, take the expected value of the posterior probability distribution as the corrected thermal health factor.

7. The intelligent planning, configuration and management system for a household energy storage system according to claim 1, wherein: The model predictive control module (24) of the data analysis and processing unit (2) adopts a model predictive control algorithm based on quantum reinforcement learning. The specific steps are as follows: Define the quantum environment and the quantum state space. Encode the state and environment information of the battery thermal management system into a quantum state to construct the quantum state space. At the same time, define the quantum action space. The quantum action space includes the speed adjustment of the coolant pump and the opening adjustment of the valve; Construct a quantum reinforcement learning agent and use the quantum variational algorithm to construct the policy network and value network of the agent. The agent interacts with the quantum environment, obtains environmental feedback through quantum measurement. The agent selects quantum actions according to the current quantum state. The environment updates the quantum state according to the action and returns a reward signal; Update the network parameters of the agent using the quantum gradient descent algorithm. Finally, perform model predictive control in the actual system based on the trained agent. The agent selects the optimal action according to the real-time quantum state and optimizes the control parameters of the thermal management system in real time.

8. The intelligent planning, configuration and management system for the household energy storage system according to claim 1, characterized in that: The fault warning module (25) of the data analysis and processing unit (2) adopts a multi-stage warning algorithm based on a spatio-temporal graph neural network. The specific steps are as follows: Construct a spatio-temporal graph of the battery thermal state, where each monitoring point of the battery is regarded as a node of the graph. The connection between nodes represents the spatio-temporal relationship between monitoring points. The features of nodes include temperature and thermal strain data, and the features of edges include thermal conductivity and time delay. Design a spatio-temporal graph neural network architecture, which includes a spatial convolutional layer and a temporal convolutional layer. The spatial convolutional layer is used to extract the spatial features between nodes, and the temporal convolutional layer is used to extract the time series features of nodes. Use historical data to train the spatio-temporal graph neural network. Take the known fault stages as labels, train the network to learn the mapping relationship between the battery thermal state and the fault stages. Input the real-time spatio-temporal graph of the battery thermal state into the trained network, output the predicted stage of the battery thermal fault, and set different warning levels according to different fault stages.

Images