Battery energy storage system dynamic reconstruction switching method based on multistage monitoring

Abstract

The invention discloses a battery energy storage system dynamic reconstruction switching method based on multistage monitoring. The method belongs to the technical field of battery energy storage system intelligent monitoring and management. According to the method, a device-level, module-level and system-level multi-level sensing monitoring network is constructed, local acquisition, preprocessing and preliminary anomaly screening of data are realized through distributed intelligent nodes, and high-sensitivity detection and rapid positioning of faults are realized by adopting multi-layer Bayesian data fusion and an improved graph structure attention network. Based on a dynamic decision algorithm of an evolutionary game, the health degree, the energy state and the system safety margin of an energy storage unit are comprehensively considered, an optimal reconstruction switching strategy is generated and executed, and automatic isolation of a fault unit and access of a standby module are realized. The method is suitable for a modularly designed complex energy storage system, and has dynamic self-healing and efficient operation capabilities under various abnormal and sudden working conditions.

Description

Technical Field

[0001] This invention belongs to the field of intelligent monitoring and management technology of battery energy storage systems, and more specifically relates to a dynamic reconfiguration and switching method for battery energy storage systems based on multi-level monitoring. Background Technology

[0002] With the rapid development of renewable energy and the increasing demands for flexibility and reliability in power systems, Battery Energy Storage Systems (BESS) are widely used in grid peak shaving, distributed energy management, microgrids, and emergency power supply. Traditional BESS management methods primarily focus on monitoring individual cells and basic protection measures. However, as energy storage scales up and system structures become increasingly complex, single-level monitoring is insufficient to meet the higher demands for safety, reliability, and intelligent operation in practical applications. In large-scale energy storage systems, anomalies in individual cells or local modules can easily trigger cascading failures, even leading to systemic failures, directly impacting the stable operation and lifespan of the energy storage device.

[0003] In existing technologies, some battery management systems improve monitoring accuracy by increasing the number of sensors and data acquisition frequency. However, this is usually limited to single-level or decentralized data acquisition, lacking deep integration of multi-level and multi-dimensional information. On the other hand, fault detection and system reconfiguration largely rely on manual intervention or preset static strategies, making it difficult to achieve rapid and accurate diagnosis of multiple faults under complex operating conditions and flexible system-level reconfiguration. This not only leads to delayed system maintenance response but also increases manual maintenance costs and error risks, failing to fully realize the self-healing capabilities and economic value of energy storage systems.

[0004] Furthermore, traditional energy storage systems mostly have fixed structures, lacking the ability to proactively adapt to cell degradation, failures, and dynamic load changes. Faced with real-world conditions such as battery performance fluctuations, abnormal temperatures, and increased internal resistance, existing systems exhibit slow response and coarse control in terms of module isolation, backup cell integration, and overall performance optimization, often failing to meet the goals of efficient, safe, and intelligent operation. Simultaneously, data transmission and centralized processing methods also face communication bottlenecks and processing pressures in large-scale applications, restricting the system's scalability and real-time performance requirements.

[0005] Therefore, there is an urgent need to propose a dynamic monitoring and intelligent reconfiguration switching method for multi-level distributed battery energy storage systems. This method aims to achieve comprehensive, high-frequency real-time perception of the system's operating status. Through mechanisms such as hierarchical data processing and fusion, intelligent fault identification and location, dynamic reconfiguration, and closed-loop self-healing control, the safety, reliability, and intelligence level of battery energy storage systems can be improved, promoting efficient and stable operation of energy storage systems in various application scenarios. This is precisely the key technical problem that this invention seeks to solve. Summary of the Invention

[0006] This invention aims to address the shortcomings of existing battery energy storage systems in terms of multi-level monitoring, real-time fault diagnosis, system dynamic reconfiguration, and self-healing control. In particular, it addresses the inability to achieve comprehensive and timely perception of the operating status of the energy storage system, the lack of rapid location and response capabilities for abnormal operating conditions, and the challenges of system reconfiguration and continuous, efficient, and reliable operation under conditions of cell failure, performance degradation, or load fluctuations. This invention aims to improve the safety, reliability, and intelligence level of energy storage systems.

[0007] To achieve the above objectives, the present invention employs the following technical solution: the method comprises: Construct a multi-level monitoring system, including a multi-layered sensor monitoring network at the equipment, module, and system levels. Deploy various sensors at key nodes of the battery energy storage system and realize local preprocessing and preliminary anomaly screening of data based on distributed intelligent nodes. Implement hierarchical data fusion. Based on the collected data, progressive data fusion is carried out between different monitoring levels through a spatiotemporal correlation model. An adaptive weighted multi-layer Bayesian fusion algorithm is adopted to achieve high-sensitivity dynamic detection and reliability verification of abnormal signals. Dynamic fault diagnosis and localization utilizes fused multi-level data and employs a fast battery fault localization algorithm based on an improved graph structure attention network to visually identify and locate faulty units within the energy storage system. Generate a reconfiguration switching strategy. After obtaining accurate fault information and system status, call the dynamic reconfiguration switching decision algorithm based on evolutionary game theory, and take into account the health status of energy storage units, energy balance requirements and system safety margin to generate the optimal reconfiguration switching scheme. Perform reconfiguration switching and real-time evaluation. Based on the generated switching strategy, execute the reconfiguration operation of the battery energy storage system, including automatically connecting / disconnecting energy storage units and switching related control logic.

[0008] In one scheme, the construction of a multi-level monitoring system includes: dividing the system into three levels—equipment level, module level, and system level—based on the structural characteristics of the battery energy storage system. Various types of sensors, such as temperature sensors, current and voltage sensors, and internal resistance sensors, are deployed at key nodes at each level, and their layout is scientifically optimized in conjunction with the system's heat dissipation structure and electrical connection methods. After hardware configuration, an edge computing unit based on distributed intelligent nodes is introduced to realize local data collection, processing and storage, as well as preliminary abnormal data screening, and localized data processing and compression.

[0009] In one approach, the implementation of hierarchical data fusion includes: taking data collected from a previous multi-level monitoring system as input, and processing the data stream sequentially at the device level, module level, and system level; The monitoring levels are progressively integrated using a spatiotemporal correlation model, taking into account both the temporal variation characteristics of nodes and the spatial correlation between different nodes. An adaptive weighted multi-level Bayesian fusion algorithm is adopted to dynamically integrate information from each level. The fusion weights are adaptively adjusted based on the Bayesian posterior probability of the decision results at each level, and the historical state is dynamically updated to reflect the changes in the accuracy of anomaly detection at each level.

[0010] In one scheme, the dynamic fault diagnosis and localization includes: using multi-level monitoring data obtained by hierarchical Bayesian fusion as input, constructing a multi-attribute graph structure that reflects the real state of the battery energy storage system, with nodes carrying fused feature vectors and edges having initial weights for coupling relationships; In the fault localization stage, an improved graph structure attention network algorithm is adopted. The weights of each neighborhood are dynamically adjusted based on a multi-head adaptive attention mechanism for graph data. Data-driven generation of abnormal association scores is used to compensate edge weights, thereby enhancing the model's feature representation of abnormal regions.

[0011] In one scheme, the generation and reconfiguration switching strategy includes: receiving fault units located by the improved graph structure attention network model and fusion monitoring data, and acquiring information such as the health status, energy distribution and safety margin of the energy storage system in real time; A dynamic reconfiguration and switching decision algorithm based on evolutionary game theory is adopted, which treats each unit as a game participant and establishes a payoff function by combining health status, energy balance and system safety margin. Under system constraints, the strategy of each unit is adaptively evolved by replicating dynamic equations. The final output includes unit operating status instructions, standby unit access sequence, and the optimal reconfiguration and switching scheme for the entire system's energy and security allocation.

[0012] In one scheme, the execution of reconfiguration switching and real-time evaluation includes: according to the optimal reconfiguration switching scheme output by the game decision module, issuing instructions to the energy storage system interface layer, automatically completing the isolation of faulty units, the access of backup units and the adaptive switching of circuit switches, and dynamically adjusting energy management, load distribution and balancing control to achieve coordinated switching at the software and hardware levels. A multi-level monitoring system is used to continuously collect and track the operational status after reconstruction at high frequency, and the performance of the reconstructed system is dynamically evaluated by using hierarchical data fusion and anomaly detection mechanisms.

[0013] In one embodiment, the method is applicable to battery energy storage systems with multiple energy storage units, modular and backup unit designs, and can achieve dynamic self-healing and efficient and reliable operation of the system in the face of complex operating conditions such as unit failure, performance degradation or load changes.

[0014] In one scheme, the dynamic reconfiguration and switching decision algorithm of the evolutionary game establishes a multi-objective weighted payoff function based on the health, energy state and safety margin of each energy storage unit, and drives the strategy of each unit to evolve towards the global optimum by replicating the dynamic equation, thereby achieving the isolation of damaged units and the optimal access of backup units.

[0015] Beneficial effects of this invention: This invention, by introducing a multi-level monitoring system, intelligent data fusion analysis, and a dynamic reconfiguration switching mechanism, achieves comprehensive, real-time monitoring of the operating status of battery energy storage systems and multi-level adaptive fault handling. Compared with existing technologies, this invention can significantly improve the fault early warning and location accuracy of battery energy storage systems, shorten fault response and repair time, and greatly enhance the safety and stability of the system.

[0016] Meanwhile, this invention, through a distributed dynamic reconfiguration strategy, can dynamically adjust the energy storage system structure according to changes in unit performance, faults, or load demands, improving the system's flexibility and scalability, extending the lifespan of battery modules, and reducing operation and maintenance costs. Furthermore, the efficient data transmission and hierarchical processing scheme of this invention effectively alleviates the data communication and processing pressure in large-scale application scenarios, significantly improving the intelligence, automation, and reliable operation level of the energy storage system, and has broad application prospects and significant economic and social benefits. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0018] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Typical embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0019] Unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. To facilitate understanding, the invention will now be described more fully with reference to the accompanying drawings. Typical embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to make the disclosure of the invention more thorough and complete.

[0020] Figure 1 As shown, a dynamic reconfiguration and switching method for a battery energy storage system based on multi-level monitoring specifically includes: Step 1: Construct a multi-level monitoring system. Design a multi-layered sensor monitoring network including device level, module level, and system level. Deploy various sensors (temperature, current, voltage, internal resistance) at key nodes of the battery energy storage system, and realize local preprocessing and preliminary anomaly screening of data based on distributed intelligent nodes.

[0021] In constructing a multi-level monitoring system, the first step is to divide the battery energy storage system into three levels: device level, module level, and system level, based on its structural characteristics. The device level primarily targets individual batteries or key components, the module level corresponds to battery modules composed of several individual batteries, and the system level encompasses the entire energy storage device and its external interfaces. Next, various types of sensors, including temperature sensors, current and voltage sensors, and internal resistance sensors, are deployed at key nodes in each level to achieve multi-dimensional and comprehensive real-time monitoring of the battery's operating status.

[0022] Sensor selection must balance environmental adaptability and measurement accuracy, and sensor layout should be scientifically optimized in conjunction with the system's heat dissipation structure and electrical connection methods. After the hardware configuration is in place, an edge computing unit based on distributed intelligent nodes is introduced, enabling each node to have local data acquisition, processing, and storage capabilities, and to perform preliminary anomaly data screening.

[0023] These nodes employ embedded algorithms to perform functions such as data denoising, filtering, and anomaly early warning, and to perform localized processing and compression of raw monitoring data, effectively reducing data transmission pressure and the load on the central processing unit.

[0024] When monitoring data exceeds set thresholds or typical abnormal patterns are detected, intelligent nodes can automatically generate preliminary anomaly reports and report them to the superior management unit, providing basic support for subsequent hierarchical data fusion and fault response. Through the establishment of this system, multi-level, distributed, and intelligent monitoring of the energy storage system's operating status can be achieved, providing a highly reliable data foundation for dynamic system reconfiguration and switching.

[0025] Step 2: Implement hierarchical data fusion. Based on the data collected by the multi-level monitoring system, progressive data fusion is implemented between different monitoring levels using a spatiotemporal correlation model. An adaptive weighted multi-level Bayesian fusion algorithm is employed to achieve highly sensitive dynamic detection and reliability verification of abnormal signals. This algorithm combines approximate Bayesian inference with a dynamic weight allocation mechanism, adjusting the weights of each monitoring level based on historical data and real-time data, ensuring both accuracy and real-time performance in the detection results.

[0026] First, data collected from the previous multi-level monitoring system is used as input. The data flow passes through the device level, module level, and system level processing stages sequentially from bottom to top. Between each monitoring level, a spatiotemporal correlation model is used to progressively fuse the data, considering both the temporal variation characteristics of individual nodes and capturing the spatial correlation between different nodes, thereby achieving a comprehensive expression of information.

[0027] Specifically, assuming at a certain level Below, the observations for each node are used Let be the node number and t be the time index. First, the temporal features of the observed sequence are extracted using Short-Time Fourier Transform (STFT) or sliding window averaging, and then the Pearson correlation coefficient is used. The spatial correlation between different nodes i and j is measured. To achieve highly reliable data fusion, an adaptive weighted multi-level Bayesian fusion algorithm is proposed to dynamically integrate information from various levels. Here, let... Indicates the first The results of hierarchical observation and decision-making. The adaptive weights for this level at time t are given, and the fused output is... , where l is the total number of monitoring levels. Weight The probability is obtained dynamically from the Bayesian posterior probability, i.e. in This means that given all the observation data and historical status Next, the The posterior probability of the hierarchical decision outcome. To improve real-time performance, approximate Bayesian inference (variational Bayes or particle filtering) is used to efficiently estimate the posterior probability. The core of the entire algorithm lies in the historical states. The dynamic updates can adaptively adjust the importance weights of each level based on the prior probabilities of past and current anomaly detections. If a level has a high recent anomaly detection accuracy, its weight is automatically increased to reflect its reliability; conversely, its contribution is decreased. Ultimately, after multi-level probability weighted fusion, the system can achieve high-sensitivity detection of anomaly signals and generate anomaly reports with confidence levels, providing a reliable basis for subsequent refined fault diagnosis and system switching.

[0028] Step 3, Dynamic Fault Diagnosis and Localization. Utilizing the fused multi-level data, a fast battery fault localization algorithm based on an Improved Graph Attention Network (IGATN) is employed to visually identify and locate faulty units within the energy storage system. This algorithm can automatically adjust adjacency weights under complex topologies, giving enhanced attention to abnormal areas and improving fault isolation accuracy.

[0029] In the process of dynamic fault diagnosis and localization, the multi-level monitoring data obtained by the aforementioned hierarchical Bayesian fusion is first used as input to construct a multi-attribute graph structure reflecting the true state of the battery energy storage system. Let the entire energy storage system be represented by a graph. It means that, among them For a set of nodes, each node Represents a battery unit or functional module. This is an edge set, reflecting the physical connections or functional relationships between units. Each node carries a fused feature vector. It includes key monitoring parameters such as temperature, voltage, and internal resistance of the unit, and also for each edge Includes initial weights reflecting the strength of the coupling relationship. In the fault localization phase, the graph data is processed using the Improved Graph Attention Network (IGATN) algorithm.

[0030] The core of IGATN is a multi-head adaptive attention mechanism that dynamically adjusts the neighborhood weights of each node and highlights the feature representation of abnormal regions, especially for complex topologies and heterogeneous edge weights. Specifically, nodes... The feature update process at layer k is as follows in, For nodes Feature representation at layer k, Let i represent the set of neighbors of node i. For trainable linear transformation weight matrix, The activation function. Attention coefficients. : here For trainable attention vectors, This represents vector concatenation. It is a structure-sensitive adjustment parameter. This represents the anomaly correlation score generated by data-driven methods, used for dynamic compensation of edge weights, enabling the model to automatically strengthen its representation of anomalous regions when potential faults occur. The multi-head attention mechanism enhances model robustness by aggregating neighborhood features in parallel across different subspaces. Labeled historical fault cases can be introduced during model training, and supervised or semi-supervised graph classification loss is used for optimization. After training, the current fused feature map is input to output the anomaly probability score for each node. .

[0031] Based on a threshold setting, nodes with anomaly probabilities exceeding the warning threshold are highlighted, thereby achieving accurate visual identification and location of faulty units. The IGATN algorithm's innovations in dynamic edge weight adjustment and anomaly pattern enhancement effectively improve the accuracy and response speed of fault isolation in complex battery system topologies, providing precise fault information support for subsequent system reconfiguration and switching.

[0032] Step 4: Generate reconfiguration switching strategy. After obtaining accurate fault information and system status, the dynamic reconfiguration switching decision algorithm based on evolutionary game theory is invoked. This algorithm can comprehensively consider the health status of energy storage units, energy balance requirements, and system safety margins to generate the optimal reconfiguration switching scheme, thereby achieving isolation of damaged units and rapid access of backup units, while also taking into account the overall system operating efficiency.

[0033] In the process of generating the reconfiguration switching strategy, the system first receives faulty units precisely located by the IGATN model and fusion monitoring data, acquiring comprehensive information such as the current health status, energy distribution, and safety margin of the energy storage system in real time. Based on this, a dynamic reconfiguration switching decision algorithm based on Evolutionary Game Theory (EGT) is used to generate the scheme. Let each unit in the energy storage system be a player in the game, denoted as the set. Each unit It has an optional set of strategies. "Participating in power supply", "Standby", or "Isolated". The system state at time t is determined by the unit health vector. Energy state vector and fault indicators Joint description.

[0034] The profit function of each unit It takes into account three aspects: State of Health (SoH), energy balance (meeting the energy supply and demand constraints of the entire system), and system safety margin (avoiding local overloads of current and voltage). Specifically, it can be modeled as follows: in, and These are the strategies adopted by the unit itself and the other units, respectively. , , These are the weighting coefficients. Reflecting health factors, using The lower the health level, the lower the reward. It represents the contribution to energy balance, and the amount of change is dynamically adjusted according to the strategy ("power supply" or "standby"); This measures the degree to which safety margins and system constraints are met. Positive values ​​are awarded when current and voltage meet the constraints at each node, with lower scores as they approach the safety limit. The global reconfiguration switching objective is to maximize the overall weighted benefit of the system; its optimization problem is: in To meet load requirements, As a quantification of safety margin, As a safety constraint threshold, This indicates that the faulty (isolated) unit cannot participate in power supply.

[0035] Evolutionary game theory achieves adaptive evolution of unit strategies by replicating dynamic equations—for each strategy... Its adoption probability Evolving over time to satisfy in For strategy Instant benefits This represents the average benefit of all units. As the evolution process progresses, inefficient or constraint-violating strategies will be gradually eliminated, and the system will eventually tend towards a Nash equilibrium solution. This means that, based on individual health, energy coordination, and safety margin, each unit can effectively isolate damaged units and quickly connect backup units in a timely manner, ensuring overall operational efficiency and system safety.

[0036] Ultimately, the optimal reconfiguration and switching scheme output by the algorithm includes detailed unit operating status instructions, standby unit access sequences, and energy and safety allocation for the entire system, guiding the energy storage system to achieve dynamic self-healing and efficient and reliable operation.

[0037] Step 5: Perform reconfiguration switching and real-time evaluation. Based on the generated switching strategy, perform the reconfiguration operation of the battery energy storage system, including automatically connecting / disconnecting energy storage units and switching related control logic. Simultaneously, a multi-level monitoring system continuously tracks the operating status of the system after reconfiguration, and adjusts the reconfiguration strategy parameters in real time based on feedback data to ensure the energy storage system maintains high reliability and high efficiency.

[0038] During the reconfiguration switching and real-time evaluation process, the system first issues instructions to the energy storage system's interface layer based on the optimal reconfiguration switching scheme output by the game theory decision-making module. The reconfiguration operation encompasses the automated connection and disconnection of the target energy storage unit, specifically including the isolation of faulty units, the access of backup units, and the adaptive switching of related circuit switches. Simultaneously, the control system dynamically adjusts core logic such as energy management, load distribution, and balancing control based on the latest reconfiguration scheme, ensuring seamless and coordinated switching at both the hardware and software levels after reconfiguration.

[0039] After completing the reconstruction at the physical and control logic levels, the system continuously collects and tracks the reconstructed operating status in real time through a multi-level monitoring system. This multi-level monitoring covers the device level, string level, and system level, focusing on observing health and safety parameters such as voltage, current, temperature, internal resistance, and SoH (Solar Hourly Rate), promptly capturing abnormal signals and operational deviations. The collected operating data is transmitted back in real time in the background. Through the aforementioned layered data fusion and anomaly detection mechanism, the system's performance after reconstruction is dynamically evaluated, including energy output stability, safety margin, health level, and system load response.

[0040] Once the evaluation process detects operational anomalies, load changes, standby unit performance fluctuations, or strategy deviations, the system automatically triggers adaptive parameter adjustments: promptly adjusting reconfiguration algorithm parameters based on the latest monitoring data, updating unit status weights, optimizing the standby unit access sequence, and even restarting the reconfiguration game process based on the fault evolution trend, achieving closed-loop self-healing control. This entire process ensures that the energy storage system maintains a highly reliable and efficient collaborative operating state when facing complex operating conditions and unexpected events, effectively guaranteeing the safe, stable, and intelligent operation goals of the battery energy storage system.

[0041] Example: I. System Structure and Monitoring System Construction This embodiment uses a certain type of distributed battery energy storage system as an example. The system consists of 6 modules (Module 1~6), each module containing 4 energy storage units (Cell 1~4), and the system has 2 backup modules. The overall architecture is as follows: Figure 1 As shown in the table below, the typical monitoring node deployments at each level are illustrated.

[0042] Table 1: Examples of Multi-Level Monitoring Node and Sensor Configurations II. Monitoring and Data Collection After the system starts up, sensor nodes at all levels collect various types of sensor data at a frequency of 1Hz, and upload the data after preprocessing at the local edge node. Taking a routine inspection as an example, some of the collected data is as follows: Table 2: Partial Data Monitoring Results at a Certain Moment Note: M = Module, C = Cell III. Layered Data Fusion and Anomaly Detection The edge intelligent nodes first perform local anomaly screening on the data they collect. In the table above, the temperature of cell M1-C4 is significantly higher than that of the surrounding cells, and its internal resistance is also abnormally high, so it is initially identified as a suspicious cell.

[0043] After entering the module and system levels, an adaptive weighted multilayer Bayesian fusion algorithm was adopted. Combined with the correlation of spatiotemporal data, the M1-C4 unit was finally identified as a "high temperature / high resistance unit" with an anomaly probability P>0.98 and an impact level of severe.

[0044] Table 3: Results of the Fusion Algorithm IV. Fault Location and Visualization The system employs a fault location algorithm based on an improved graph-structured attention network (GAT). By inputting multi-level fusion data, it can locate the M1-C4 unit (a precursor to thermal runaway) within seconds. The diagnostic map is shown below.

[0045] Table 4: Fault Location and Risk Level V. Generation of Dynamic Reconfiguration and Switching Strategies Diagnostic information is input into the game decision-making module, and a weighted payoff function is established by combining the health status, energy balance, and standby module information of each module. By replicating the dynamic equations, the following switching strategy is derived: Isolate units M1-C4, while the remaining units of M1 continue to be connected. The backup module B1 will be used to supplement the current energy and safety requirements. Table 5: Restructuring Decision Output VI. Refactoring Switchover Execution and Real-time Evaluation The reconfiguration command is issued by the system, automatically completing the circuit switch, disconnecting M1-C4, and smoothly connecting module B1. After the switch, the system's operating status is dynamically monitored, and the following data is continuously collected and evaluated: Table 6: Comparison of some system parameters before and after the switch In this embodiment, the system rapidly identifies abnormal battery cells based on multi-level monitoring and fusion, and completes self-healing switching within seconds, ensuring stable output and high safety redundancy. After switching, the system's energy distribution is more balanced, and no new anomalies appear.

[0046] Table 7: Comparison of Operation and Maintenance Efficiency Before and After Implementation The above embodiments demonstrate that the dynamic reconfiguration switching method of the present invention greatly improves the fault self-healing capability and operating efficiency of the battery energy storage system, significantly shortens the fault response and switching time, realizes efficient and reliable system operation, and adapts to various complex operating conditions.

[0047] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0048] It should be understood that the above detailed description of the technical solutions of the present invention with reference to preferred embodiments is illustrative and not restrictive. Those skilled in the art can modify the technical solutions described in the embodiments or make equivalent substitutions for some of the technical features based on reading this specification; however, these modifications or substitutions do not cause the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A dynamic reconfiguration and switching method for a battery energy storage system based on multi-level monitoring, characterized in that: The method includes: Construct a multi-level monitoring system, including a multi-layered sensor monitoring network at the equipment, module, and system levels. Deploy various sensors at key nodes of the battery energy storage system and realize local preprocessing and preliminary anomaly screening of data based on distributed intelligent nodes. Implement hierarchical data fusion. Based on the collected data, progressive data fusion is carried out between different monitoring levels through a spatiotemporal correlation model. An adaptive weighted multi-layer Bayesian fusion algorithm is adopted to achieve high-sensitivity dynamic detection and reliability verification of abnormal signals. Dynamic fault diagnosis and localization utilizes fused multi-level data and employs a fast battery fault localization algorithm based on an improved graph structure attention network to visually identify and locate faulty units within the energy storage system. Generate a reconfiguration switching strategy. After obtaining accurate fault information and system status, call the dynamic reconfiguration switching decision algorithm based on evolutionary game theory, and take into account the health status of energy storage units, energy balance requirements and system safety margin to generate the optimal reconfiguration switching scheme. Perform reconfiguration switching and real-time evaluation. Based on the generated switching strategy, execute the reconfiguration operation of the battery energy storage system, including automatically connecting / disconnecting energy storage units and switching related control logic.

2. The method for dynamic reconfiguration and switching of a battery energy storage system based on multi-level monitoring according to claim 1, characterized in that: The construction of the multi-level monitoring system includes: dividing the system into three levels—equipment level, module level, and system level—based on the structural characteristics of the battery energy storage system; Various types of sensors, including temperature sensors, current and voltage sensors, and internal resistance sensors, are deployed at key nodes at each level, and their layout is scientifically optimized in conjunction with the system's heat dissipation structure and electrical connection methods. After hardware configuration, an edge computing unit based on distributed intelligent nodes is introduced to realize local data collection, processing and storage, as well as preliminary abnormal data screening, and localized data processing and compression.

3. The method for dynamic reconfiguration and switching of a battery energy storage system based on multi-level monitoring according to claim 1, characterized in that: The implementation of hierarchical data fusion includes: taking data collected from previous multi-level monitoring systems as input, and processing the data stream sequentially at the device level, module level, and system level; The monitoring levels are progressively integrated using a spatiotemporal correlation model, taking into account both the temporal variation characteristics of nodes and the spatial correlation between different nodes. An adaptive weighted multi-level Bayesian fusion algorithm is adopted to dynamically integrate information from each level. The fusion weights are adaptively adjusted based on the Bayesian posterior probability of the decision results at each level, and the historical state is dynamically updated to reflect the changes in the accuracy of anomaly detection at each level.

4. The method for dynamic reconfiguration and switching of a battery energy storage system based on multi-level monitoring according to claim 1, characterized in that: The dynamic fault diagnosis and localization includes: using multi-level monitoring data obtained by hierarchical Bayesian fusion as input, constructing a multi-attribute graph structure that reflects the real state of the battery energy storage system, with nodes carrying fused feature vectors and edges having initial weights for coupling relationships; In the fault localization stage, an improved graph structure attention network algorithm is adopted. The weights of each neighborhood are dynamically adjusted based on a multi-head adaptive attention mechanism for graph data. Data-driven generation of abnormal association scores is used to compensate edge weights, thereby enhancing the model's feature representation of abnormal regions.

5. The method for dynamic reconfiguration and switching of a battery energy storage system based on multi-level monitoring according to claim 1, characterized in that: The aforementioned generation, reconfiguration, and switching strategy includes: receiving faulty units located by the improved graph structure attention network model and fusion monitoring data, and acquiring real-time information on the health status, energy distribution, and safety margin of the energy storage system; A dynamic reconfiguration and switching decision algorithm based on evolutionary game theory is adopted, which treats each unit as a game participant and establishes a payoff function by combining health status, energy balance and system safety margin. Under system constraints, the strategy of each unit is adaptively evolved by replicating dynamic equations. The final output includes unit operating status instructions, standby unit access sequence, and the optimal reconfiguration and switching scheme for the entire system's energy and security allocation.

6. The method for dynamic reconfiguration and switching of a battery energy storage system based on multi-level monitoring according to claim 1, characterized in that: The aforementioned execution of reconfiguration switching and real-time evaluation includes: issuing instructions to the energy storage system interface layer based on the optimal reconfiguration switching scheme output by the game decision module, automatically completing the isolation of faulty units, access of backup units and adaptive switching of circuit switches, and dynamically adjusting energy management, load distribution and balancing control to achieve coordinated switching at the software and hardware levels. A multi-level monitoring system is used to continuously collect and track the operational status after reconstruction at high frequency, and the performance of the reconstructed system is dynamically evaluated by using hierarchical data fusion and anomaly detection mechanisms.

7. A dynamic reconfiguration and switching method for a battery energy storage system based on multi-level monitoring, as described in any one of claims 1-6, characterized in that: The method is applicable to battery energy storage systems with multiple energy storage units, modular and backup unit designs, and can achieve dynamic self-healing and efficient and reliable operation of the system in the face of complex operating conditions such as unit failure, performance degradation or load changes.

8. The method for dynamic reconfiguration and switching of a battery energy storage system based on multi-level monitoring according to claim 5, characterized in that: The dynamic reconfiguration and switching decision algorithm of the evolutionary game establishes a multi-objective weighted payoff function based on the health, energy state and safety margin of each energy storage unit, and drives the strategy of each unit to evolve adaptively towards the global optimum by replicating the dynamic equation, so as to achieve the isolation of damaged units and the optimal access of backup units.

Images