Modular energy storage method and system

Abstract

The application provides a modular energy storage method and system, belonging to the technical field of power energy storage scheduling, which comprises the following steps: collecting operation data of power generation equipment, a large power grid and power consumption equipment, and basic operation parameters and extended parameters of each energy storage module; performing standardized processing and mining on the operation data, and extracting the correlation characteristics between power generation and power consumption; fusing the basic operation parameters and the extended parameters to determine the health status of each energy storage module; inputting the correlation characteristics into a preset power generation and power consumption prediction model to obtain predicted power generation and power consumption demand; determining the initial charging and discharging strategy of each energy storage module based on the predicted power generation and power consumption demand; distributing the initial charging and discharging strategy according to the health status of each energy storage module to generate a coordinated control scheme; inputting the coordinated control scheme and the health status into a preset fault analysis model to obtain predicted fault types and fault occurrence probabilities; and performing hierarchical processing on the coordinated control scheme based on the fault occurrence probabilities.

Description

Technical Field

[0001] This application relates to the field of power energy storage and dispatch technology, and in particular to a modular energy storage method and system. Background Technology

[0002] With the large-scale integration of distributed power generation equipment in smart industrial parks, the supply and demand of energy sources in these parks exhibits fluctuating and random characteristics, highlighting the shortcomings of traditional energy storage dispatching technologies. Currently, energy storage systems in smart industrial parks mostly adopt a unified dispatching mode, failing to consider the differentiated health status of each energy storage module and exhibiting poor distribution balance, leading to overload operation and accelerated aging of some modules. Furthermore, existing technologies lack in-depth data mining of power generation and consumption, failing to capture source-load correlation patterns, resulting in low accuracy in power generation and consumption prediction and insufficient adaptability of initial charging and discharging strategies. In addition, existing dispatching methods are mostly reactive protection after a fault, lacking a proactive fault prediction mechanism. This prevents the early identification of potential fault risks during the coordinated operation of energy storage modules and the dynamic optimization of control strategies based on fault probability and type, leading to equipment failures and decreased operational stability. Summary of the Invention

[0003] To address the aforementioned technical problems, this application provides a modular energy storage method and system.

[0004] A first aspect of this application provides a modular energy storage method, comprising: Collect operational data from power generation equipment, the main power grid, and power consumption equipment in the smart industrial park, as well as the basic and extended operational parameters of each energy storage module; The operational data is standardized and data-mined to extract the correlation features between power generation and power consumption; the basic operational parameters and extended parameters are fused, and the health status of each energy storage module is determined. The associated features are input into a preset power generation and electricity consumption prediction model to obtain the predicted power generation and predicted electricity consumption. Based on the predicted power generation and the predicted electricity demand, the initial charging and discharging strategy for each energy storage module is determined. Based on the health status of each energy storage module, the initial charging and discharging strategy is assigned tasks and power balanced to generate a collaborative control scheme among the energy storage modules. The collaborative control scheme and the health status are input into a preset fault analysis model to obtain the predicted fault types and fault occurrence probabilities that will occur when each energy storage module executes the collaborative control scheme. When the probability of the fault occurrence is greater than a preset first fault occurrence probability threshold and less than a preset second fault occurrence probability threshold, the collaborative control scheme is modified based on the fault type to obtain a target collaborative control scheme, and the charging and discharging power of each energy storage module is adjusted based on the target collaborative control scheme; when the probability of the fault occurrence is greater than the preset second fault occurrence probability threshold, a warning message is generated and output.

[0005] A second aspect of this application provides a modular energy storage system, comprising: The multi-source data acquisition module is used to collect operating data of power generation equipment, power grid and power consumption equipment in the smart industrial park, as well as the basic operating parameters and extended parameters of each energy storage module. The data feature analysis module is used to standardize and mine the operating data, extract the correlation features between power generation and power consumption, fuse the basic operating parameters and extended parameters, and determine the health status of each energy storage module. The source-load power prediction module is used to input the associated features into a preset power generation and consumption prediction model to obtain the predicted power generation and predicted electricity demand. The initial charge and discharge strategy module is used to determine the initial charge and discharge strategy of each energy storage module based on the predicted power generation and the predicted power demand. A collaborative scheme generation module is used to perform task allocation and power balancing on the initial charging and discharging strategy based on the health status of each energy storage module, and generate a collaborative control scheme between each energy storage module. The fault prediction and calculation module is used to input the collaborative control scheme and the health status into a preset fault analysis model to obtain the predicted fault types and fault occurrence probabilities that will occur when each energy storage module executes the collaborative control scheme. The tiered risk control and handling module is used to modify the collaborative control scheme based on the fault type when the probability of the fault occurrence is greater than a preset first fault occurrence probability threshold and less than a preset second fault occurrence probability threshold, to obtain a target collaborative control scheme, and to adjust the charging and discharging power of each energy storage module based on the target collaborative control scheme; when the probability of the fault occurrence is greater than the preset second fault occurrence probability threshold, a warning message is generated and output.

[0006] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the modular energy storage method described above.

[0007] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the modular energy storage method described above.

[0008] The beneficial effects of the modular energy storage method and system provided in this application are as follows: This application extracts the correlation features between power generation and consumption in the industrial park from multi-source data, and determines the health status based on multi-dimensional parameters of energy storage modules. This effectively reduces the problems of insufficient source-load feature mining and single energy storage status assessment in traditional intelligent industrial park energy storage scheduling, and effectively improves the accuracy of source-load power prediction. Based on the prediction results, the initial charging and discharging strategy is determined, and power balancing and task allocation are completed according to the health status of each energy storage module. This achieves differentiated collaborative scheduling of multiple energy storage modules, reduces the overload operation of a single module, effectively delays the aging of energy storage equipment, and thus improves the overall service life of the energy storage system. At the same time, based on the fault analysis model, operational fault prediction is achieved in advance. Through a dual-threshold hierarchical control mechanism, the collaborative control scheme can be dynamically modified or alarm information can be output for different fault probability scenarios, changing the traditional passive fault protection mode of energy storage. This effectively improves the accuracy, balance, and operational safety of the modular energy storage system scheduling in intelligent industrial parks, and ensures the stable and efficient operation of the park's source-grid-load-storage system. Attached Figure Description

[0009] Figure 1 A schematic flowchart of a modular energy storage method provided in an embodiment of this application; Figure 2 A structural block diagram of a modular energy storage system provided in an embodiment of this application; Figure 3 This is a schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0010] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0011] To make the purpose, technical solution, and advantages of this application clearer, the following will be described in conjunction with the appendix. Figure 1 -Appendix Figure 3 The following is an explanation using specific examples.

[0012] Please refer to Figure 1 , Figure 1This is a flowchart illustrating a modular energy storage method provided in an embodiment of this application. The method includes: S101: Collects operational data from power generation equipment, the main power grid, and power consumption equipment in the smart industrial park, as well as the basic and extended operational parameters of each energy storage module.

[0013] In this embodiment, the power generation equipment is a distributed power system deployed within the smart industrial park, capable of generating electricity to provide local power support for the park's load. This equipment includes distributed photovoltaic systems, small wind turbines, and gas-fired power generation equipment. The power output of these systems is affected by external factors such as sunlight, wind speed, and ambient temperature, exhibiting strong randomness and volatility. The main power grid connects to the external public power transmission network of the smart industrial park, serving as the primary power exchange entity for replenishing the park's power supply and transmitting surplus power. It can supplement power supply when the park's power generation is insufficient and absorb excess power when there is surplus power, ensuring an overall balance between power supply and demand within the park. The electrical equipment refers to various energy-consuming production, office, and supporting equipment within the smart industrial park, forming part of the park's power load. This equipment includes industrial production equipment, air conditioning and lighting equipment, office electrical equipment, and supporting power equipment for the park. The power load exhibits dynamic fluctuations depending on production shifts, day / night cycles, and seasonal environmental conditions.

[0014] Operational data includes various operating conditions, power, and status data generated by power generation equipment, the main power grid, and electrical equipment during operation. Specifically, power generation equipment operational data includes power generation, voltage generation, current generation, equipment operating temperature, and start-up / shutdown status; main power grid operational data includes grid connection point interaction power, grid voltage, grid frequency, and electricity purchase price; and electrical equipment operational data includes time-series data such as load power, current consumption in each branch circuit, and load operating duration.

[0015] An energy storage module is an independent, controllable, and independently dispatchable basic energy storage unit in a modular energy storage system. It has independent charging and discharging capabilities, and multiple energy storage modules work together to form a modular energy storage system for the park. Basic operating parameters are the parameters of the basic operating conditions of the energy storage module, including the energy storage module's state of charge, charging and discharging power, terminal voltage, operating current, insulation status, equipment start-up and shutdown status, etc. The energy storage module can represent the current basic operating conditions of the energy storage module.

[0016] Extended parameters are auxiliary monitoring parameters that characterize the operating environment, aging status, potential losses, and operating details of the energy storage module. These include the internal temperature of the energy storage module, temperature rise rate, ambient temperature and humidity, number of charge-discharge cycles, health status attenuation coefficient, historical fault records, individual cell voltage consistency, and heat dissipation parameters.

[0017] S102: Standardize and mine the operating data to extract the correlation features between power generation and power consumption; integrate the basic operating parameters and extended parameters, and determine the health status of each energy storage module.

[0018] In this embodiment, the operating data is double-aligned based on timestamps and the spatial layout of the energy storage modules to construct a spatiotemporal data matrix. The spatiotemporal data matrix is ​​then split into time series of different nodes based on the spatial dimension. Variational mode decomposition is performed on the time series of each node to obtain the intrinsic mode components of each node. The spatiotemporal mutual information entropy of the power generation and power consumption mode components between different spatial nodes and within the same node is calculated, and a correlation strength function with spatial distance as the attenuation factor is constructed. Node pairs with a value greater than a preset node threshold and their corresponding mode component labels are extracted from the correlation strength function as the correlation features between power generation and power consumption.

[0019] Simultaneously, the basic operating parameters and extended parameters are fused. Specifically, the basic operating parameters and extended parameters undergo unified standardization preprocessing to eliminate differences in dimensions and numerical ranges between different parameters. Secondly, based on time series alignment, the basic operating parameters and extended parameters of the same energy storage module at the same time are dimensionally concatenated to obtain a multi-dimensional energy storage state parameter set including operating condition information, environmental information, aging information, and loss information. Finally, through a feature-weighted fusion algorithm, weight coefficients are assigned according to the degree of influence of various parameters on the energy storage operating state, ultimately obtaining fused state data. This fused state data includes the explicit operating conditions and implicit aging characteristics of the energy module.

[0020] After obtaining the fusion status data, this embodiment determines the health status of each energy storage module based on this data. A comprehensive quantitative assessment of the energy storage module status is performed from four dimensions: operating conditions, equipment aging degree, operating environment quality, and potential loss risks. Specifically, basic operating parameters represent the stability of the current charge and discharge conditions of the energy storage module, while extended parameters characterize the long-term aging degradation level, temperature stress, and historical operating losses of the energy storage module. Furthermore, the fusion status data is substituted into a preset energy storage health evaluation model to calculate the health status of each energy storage module.

[0021] The energy storage health assessment model is constructed using a deep learning multi-layer feature perception network model, which is adapted to the nonlinear correlation fitting requirements of multi-dimensional time-series fusion parameters. It can uncover the implicit aging and health degradation patterns of energy storage under multi-parameter coupling, and has stronger adaptability to complex operating conditions and higher assessment accuracy. The energy storage health assessment model adopts a four-layer progressive hierarchical architecture, including a data input layer, a feature extraction layer, a correlation fitting layer, and a result output layer. Specifically, the data input layer receives preprocessed fusion status data, aggregating four dimensions of parameters: energy storage module operating conditions, equipment aging, operating environment, and potential losses. It performs data dimension normalization and time-series verification, removing residual noise data. The feature extraction layer uses a convolutional neural network structure to perform deep feature mining on the multi-dimensional fusion parameters, filtering features that affect the energy storage health status, and extracting steady-state operating condition features, temperature aging features, loss attenuation features, and environmental interference features. The fitting layer uses a fully connected neural network structure to perform nonlinear coupling fitting of the extracted features, establishing a mapping relationship between multiple parameters and energy storage health attenuation, normalizing the mapping relationship, and outputting a normalized energy storage health score. The result output layer incorporates multi-level health threshold judgment rules, matching the health score with a standard health threshold range, and finally outputting the health status corresponding to each energy storage module.

[0022] The energy storage health evaluation model uses historical operating datasets for offline training and online iterative optimization. Specifically, it first collects multi-dimensional fusion status data retained from the long-term operation of various energy storage modules in the smart industrial park, including sample data under different operating conditions, environmental conditions, and aging stages. The sample data is then classified and labeled, and the corresponding energy storage health status labels are matched to obtain a complete model training sample set. Subsequently, the sample data is divided into training set, validation set, and test set. The training set is input into the network model to complete the forward propagation operation, and the network weights and bias parameters are continuously corrected through backpropagation. The nonlinear mapping relationship between the multi-dimensional fusion parameters and the energy storage health status is continuously fitted. At the same time, the validation set is used to verify the model's fitting effect in real time. Finally, the generalization ability of the model is verified through the test set.

[0023] S103: Input the associated features into the preset power generation and electricity consumption prediction model to obtain the predicted power generation and predicted electricity demand.

[0024] In this embodiment, the correlation features refer to the characteristic information mined from the standardized power generation and consumption operation data of the park, which can characterize the temporal and spatial coupling change patterns of power generation and consumption, representing the inherent linkage relationship between source and load. The predicted power generation is the electrical power output that the power generation equipment in the park can produce at each future time, calculated by the power generation and consumption prediction model. The predicted electricity demand represents the electrical load required by all electrical equipment in the park at each future time, output by the power generation and consumption prediction model.

[0025] The power generation and consumption prediction model employs a deep learning temporal hybrid network structure. The backbone uses a long short-term memory (LSM) network to extract temporal variation patterns, while the front end uses one-dimensional convolutions for local spatial correlation feature extraction. The convolutional layers aggregate local features from the coupling features of homogeneous and neighboring spatial nodes. The LSM network then receives the convolutional outputs to mine daily and weekly power generation and consumption fluctuation patterns. This is a hybrid deep learning architecture combining one-dimensional convolutions and LSM networks, rather than a pure convolutional structure. The model's training includes: dividing the historical dataset of the smart industrial park into training, validation, and test sets, using power generation and load as labels, and inputting the correlation features mined for the corresponding time periods as samples; employing supervised learning, using mean squared error as the loss function to iteratively optimize the network weights; stopping training when the validation set error no longer decreases after multiple rounds, and continuously adjusting using subsequent actual operating data.

[0026] The power generation and consumption prediction model comprises a four-layer progressive network hierarchy. Specifically, the input layer receives the effective correlated features mined earlier; the feature fusion layer uses a one-dimensional convolutional structure to perform local feature aggregation and deep coupling feature extraction on the input spatiotemporal correlated features, filtering highly correlated effective features and weakening ineffective features caused by random noise; the temporal mining layer uses a long short-term memory network structure to specifically handle the deep features after convolutional fusion, capturing the long-term variation patterns and short-term fluctuation characteristics of power generation and consumption demand, and memorizing historical temporal linkage features; the output layer is a fully connected mapping layer, responsible for mapping the high-dimensional features after temporal mining into low-dimensional quantized values, outputting the predicted power generation and predicted consumption demand for future time periods.

[0027] S104: Based on the predicted power generation and predicted electricity demand, determine the initial charging and discharging strategy for each energy storage module.

[0028] In this embodiment, the predicted power generation and predicted electricity demand are aligned based on time series, and the power difference sequence at each time segment is calculated. Based on the alternating positive and negative characteristics of the power difference sequence, the time interval of continuous unidirectional power differences is divided into a window to obtain the scheduling window. According to the power limit of the interaction between electricity price information and the main power grid, the cumulative power deviation from the start time of the scheduling window to the current time is calculated. The power difference sequence is smoothed using a moving average filter to generate the power adjustment coefficient of the scheduling window. According to the health status and the current state of charge of each energy storage module, the response priority of each energy storage module is calculated. According to the response priority and the power adjustment coefficient, the charging and discharging power of each energy storage module is allocated to obtain the charging and discharging power and charging and discharging time period of each energy storage module as the initial charging and discharging strategy. Among them, the initial charging and discharging strategy is a preliminary, group-based energy storage charging and discharging scheduling scheme formulated only from the perspective of the overall source-load balance of the smart industrial park, based on the overall power supply and demand difference of the park calculated by the predicted power generation and predicted electricity demand.

[0029] S105: Based on the health status of each energy storage module, perform task allocation and power balancing for the initial charging and discharging strategy, and generate a collaborative control scheme among the energy storage modules.

[0030] In this embodiment, the health status of each energy storage module serves as a constraint on power allocation. Different energy storage modules are affected by cyclic aging and differences in operating environment, resulting in varying health levels, differences in the maximum charge and discharge power they can bear, and differences in their sustainable operating time. Therefore, it is not possible to uniformly determine the initial charge and discharge strategy.

[0031] Specifically, when carrying out task allocation and power balancing, the health status of each energy storage module is used as the basis. First, the scheduling priority of energy storage units is divided according to their health status. Large-scale charging and discharging tasks are preferentially allocated to energy storage modules with good health status, while the task allocation of energy storage modules with poor health status is reduced and short-term high-power output is limited. Then, power balancing is carried out according to the remaining state of charge and power upper and lower limit constraints of each energy storage module to smooth out the power difference between energy storage in the same group, so as to avoid the imbalance of some energy storage modules operating at full power and some modules idling.

[0032] After completing task allocation and power balancing, the charging and discharging power, switching time periods, and operating constraints allocated to each energy storage module are integrated to generate a collaborative control scheme among the energy storage modules. This collaborative control scheme includes the start and stop times of each energy storage module, real-time charging and discharging power curves, and operating boundaries. Through multi-module staggered peak coordination and complementary output, the scheme effectively mitigates fluctuations in the park's source and load.

[0033] S106: Input the collaborative control scheme and health status into the preset fault analysis model to obtain the predicted fault types and fault probabilities that will occur when each energy storage module executes the collaborative control scheme.

[0034] In this embodiment, the predicted fault types include typical energy storage faults such as overvoltage, overcurrent, and abnormal temperature rise. The fault occurrence probability refers to the probability value of the corresponding fault type occurring during the execution of the collaborative control scheme by each energy storage module.

[0035] The fault analysis model employs a multi-layer fully connected deep neural network architecture with a shallow feature filtering layer added at the front end and no convolutional structure. It takes multi-dimensional operating conditions and aging features as input, completes non-linear feature mapping through multi-layer fully connected layers, and outputs different fault categories and corresponding probabilities at the terminal with multiple branches, achieving simultaneous output of multi-classification and probabilistic regression. The model is trained using a labeled dataset composed of historical fault data from the park's battery management system, historical execution scheduling schemes, and post-fault records. Supervised multi-task training is employed: fault type loss is jointly optimized using classification loss, and fault occurrence probability loss is jointly optimized using regression loss. Data undergoes normalization preprocessing, the dataset is split for iterative training, and overfitted samples are removed through cross-validation. Later, incremental retraining is performed periodically with newly added fault samples from the field to continuously improve the accuracy of fault prediction.

[0036] The fault analysis model comprises a four-layer structure adapted to specific operating conditions. Specifically, the input layer is responsible for uniformly accessing multi-dimensional input parameters, including parameters of the collaborative control schemes of each energy storage module, charging and discharging power allocation values, charging and discharging durations, and energy storage health status, health degradation rate, and historical operating condition data collected by the battery management system, completing the normalization and standardization preprocessing of multi-source heterogeneous data. The feature selection layer is a shallow feature optimization layer responsible for automatically identifying features strongly correlated with energy storage faults, filtering irrelevant and redundant operating condition parameters, and retaining high-risk fault-inducing features such as power over-limit, aging degradation, and long-term operating conditions. The deep mapping layer adopts a multi-layer fully connected neural network structure, responsible for performing high-dimensional nonlinear transformations and deep correlation mining on the selected fault-related features, fitting the implicit mapping relationship between scheduling conditions, equipment aging status, and various faults, and quantifying the fault risk level under different operating conditions. The multi-task output layer is a branch output structure, corresponding to fault classification and probability regression tasks respectively, responsible for outputting the fault type and corresponding fault occurrence probability of each energy storage module.

[0037] S107: When the probability of a fault occurrence is greater than the preset first fault occurrence probability threshold and less than the preset second fault occurrence probability threshold, the collaborative control scheme is modified based on the fault type to obtain the target collaborative control scheme, and the charging and discharging power of each energy storage module is adjusted based on the target collaborative control scheme; when the probability of a fault occurrence is greater than the preset second fault occurrence probability threshold, a warning message is generated and output.

[0038] In this embodiment, when the probability of a fault occurrence is greater than a preset first fault occurrence probability threshold and less than a preset second fault occurrence probability threshold, the extreme type of the current operating scenario is identified. The extreme types include sudden power change scenarios and extreme electricity price fluctuation scenarios. The fault type output by the fault analysis model is obtained. When the fault type is an overcurrent fault or an overvoltage fault, the fault type is matched and corrected with the extreme type to obtain the target cooperative control scheme. Specifically: if the fault type is an overcurrent fault and the current scenario is a sudden power change scenario, the power change rate in the cooperative control scheme is limited to a preset maximum change rate, and the power waveform is low-pass filtered; if the fault type is an overvoltage fault and the current scenario is an extreme electricity price fluctuation scenario, the voltage regulation coefficient is calculated based on the electricity price deviation, and the charging power limit of the corresponding energy storage module is reduced based on the voltage regulation coefficient; if the fault type is not directly related to the current scenario, the cooperative control scheme is corrected based on the correction strategy corresponding to the preset fault type-correction table. The first and second failure probability thresholds are two risk demarcation values ​​pre-defined based on historical failure data and the tolerance limits of each energy storage device. These two thresholds divide the medium-risk and high-risk ranges from low to high. When the failure probability falls within the range greater than the first failure probability threshold but less than the second failure probability threshold, it indicates that there is a potential for failure in the operation of each energy storage module according to the existing collaborative control scheme, but the failure risk is controllable and there is no need to directly terminate the scheduling.

[0039] When the probability of a fault exceeds the second fault probability threshold, it indicates that even adjusting the charging and discharging power cannot prevent the equipment fault. Continuing to execute the corresponding scheduling task will lead to safety hazards such as energy storage damage and power instability in the park. Therefore, no further scheme correction operations will be carried out. At this time, a warning message is generated and output. The alarm message includes the energy storage module number corresponding to the fault, the fault type, and the location of the risk. If necessary, the corresponding energy storage module will be suspended from operation. The conditions for suspending the operation of the corresponding energy storage module include: the fault type marked in the output warning message being a high-risk irreversible fault such as internal cell short circuit, insulation damage, or main control hardware abnormality; and the energy storage module's operating parameters simultaneously showing abnormal phenomena such as a sudden increase in temperature and a serious deviation of the individual cell voltage from the normal range.

[0040] As can be seen from the above, this application effectively reduces the problems of insufficient source-load feature mining and single energy storage status assessment in traditional intelligent industrial park energy storage scheduling by extracting the correlation features of power generation and consumption in the park from multi-source data and determining the health status based on multi-dimensional parameters of energy storage modules. This significantly improves the accuracy of source-load power prediction. Based on the prediction results, the initial charging and discharging strategy is determined, and power balancing and task allocation are completed according to the health status of each energy storage module. This achieves differentiated collaborative scheduling of multiple energy storage modules, reduces the overload operation of a single module, effectively delays the aging of energy storage equipment, and thus extends the overall service life of the energy storage system. Simultaneously, based on a fault analysis model, operational fault prediction is achieved in advance. Through a dual-threshold hierarchical control mechanism, the collaborative control scheme can be dynamically modified or alarm information can be output for different fault probability scenarios, changing the traditional passive fault protection mode of energy storage. This effectively improves the accuracy, balance, and operational safety of the modular energy storage system scheduling in intelligent industrial parks, ensuring the stable and efficient operation of the park's source-grid-load-storage system.

[0041] In one embodiment of this application, the operational data is standardized and data mined to extract the correlation features between power generation and power consumption, including: The operational data is double-aligned based on timestamps and the spatial layout of energy storage modules to construct a spatiotemporal data matrix. The spatiotemporal data matrix is ​​split into time series of different nodes based on the spatial dimension, and variational mode decomposition is performed on the time series of each node to obtain the eigenmode components of each node. Calculate the spatiotemporal mutual information entropy of power generation and power consumption mode components between different spatial nodes and within the same node, and construct a correlation strength function with spatial distance as the attenuation factor; Extract node pairs and their corresponding modal component labels from the correlation strength function that are greater than a preset node threshold, as the correlation features between power generation and power consumption.

[0042] In this embodiment, a unified time benchmark for data collection is first established, using the timestamp corresponding to a fixed sampling period as the time series scale to calibrate the time series of dispersedly collected power generation data, power consumption data, and grid interaction data. Simultaneously, based on the spatial layout information of each energy storage module within the smart industrial park, power generation equipment and power loads in different geographical locations are spatially bound to energy storage nodes, determining the spatial node location corresponding to each set of operational data. After completing time series alignment and spatial registration, all the regularized operational data are structured and arranged using timestamps as the row dimension and spatial nodes as the column dimension, constructing a complete, orderly, and spatiotemporally unified spatiotemporal data matrix to achieve integrated spatiotemporal storage and representation of the park's source-load operational data. The fixed sampling period is the continuous data acquisition time interval uniformly adopted when collecting various operational data from power generation equipment, power consumption equipment, grid interaction, and energy storage modules on-site in the smart industrial park. Specifically, it is comprehensively set based on the hardware sampling performance of the smart industrial park's power distribution system data acquisition device, the rate of change of source-load power fluctuations in the park, and the time granularity requirements of the energy storage system's scheduling and control.

[0043] Secondly, the spatial dimension index of the spatiotemporal data matrix is ​​traversed, and the spatial coordinates of each energy storage module are used as node identifiers. Data vectors of each node at all timestamps are extracted from the spatiotemporal data matrix to form the time series of each node. The initial modal decomposition number K and the second-order penalty factor α are calculated based on the signal length and sampling frequency of the time series of each node. Here, K represents the total number of intrinsic modal components to be decomposed. With the convergence objective that the center frequencies of each intrinsic modal component no longer overlap during the iteration process, the alternating direction multiplier method is used to perform variational modal decomposition on the time series of each node, and the calculation is performed after each iteration. The spectral overlap between any two intrinsic mode components; in the first iteration, the spectral overlap is not compared, and the next iteration begins; when the spectral overlap is greater than the spectral overlap calculated in the previous iteration, the value of k is increased by 1, and the center frequency of the newly added intrinsic mode component is initialized to the midpoint value of the maximum interval between any two adjacent center frequencies according to the current center frequency distribution of each intrinsic mode component, and the iterative decomposition continues; when the rate of change of the center frequency of the mode component in two consecutive iterations is less than the preset rate of change of the center frequency, the decomposition stops, and the intrinsic mode components of each node are obtained.

[0044] Subsequently, for all intrinsic mode components obtained from the decomposition, correlation quantification calculations were performed at two levels: within a single node and through multi-node interactions. The information dependence between the power generation mode component and the power consumption mode component was quantified using spatiotemporal mutual information entropy. A higher spatiotemporal mutual information entropy value indicates stronger synchronization and coupling between the two temporal fluctuations. Simultaneously, based on the spatial attenuation characteristics of power coupling in smart industrial parks—that is, the greater the spatial distance between nodes, the weaker the source-load linkage effect—the spatial distance between nodes was used as an attenuation factor to spatially correct the spatiotemporal mutual information entropy, constructing a correlation strength function. This correlation strength function can simultaneously consider the correlation of temporal fluctuations and spatial location differences. Specifically, correlation strength = mutual information entropy × attenuation coefficient, where the attenuation coefficient monotonically decreases as the spatial distance between the two nodes increases: the smaller the spatial distance between the two nodes and the closer the electrical connection, the closer the attenuation coefficient approaches 1.

[0045] Finally, the correlation strength function calculation results of all node pairs are compared one by one with the preset node threshold. Valid node pairings with correlation strength greater than the threshold are selected, and the modal component labels corresponding to these pairs are retained. These modal component labels are used to distinguish different correlation types, such as trend correlation, periodic correlation, and short-term disturbance correlation. Finally, the selected valid node pair relationships and corresponding modal component labels are integrated and summarized to obtain the correlation characteristics representing the spatiotemporal coupling law of power generation and consumption in the smart industrial park. The node threshold is determined through statistical analysis of historical power generation and consumption data and operational sample data of each energy storage module in the smart industrial park. Specifically, the statistical distribution law of the correlation strength function of all node pairs is first calculated in batches using historical datasets. Based on the power topology and spatial span of the smart industrial park, as well as the engineering requirements for selecting effective coupling correlations and filtering weak correlations caused by random noise, and referring to the intensity boundary between invalid stray correlations and truly effective coupling correlations under historical operating conditions, while also considering the input feature capacity constraints of subsequent prediction models, the preset node threshold is comprehensively determined. The node threshold is used to distinguish between effective source-load coupled node pairs and redundant node pairs without actual correlation.

[0046] As can be seen from the above, this embodiment constructs a spatiotemporal data matrix through dual alignment of timestamps and energy storage space layout, achieving unified and regular spatiotemporal dimensions of source and load data to eliminate data deviations caused by temporal misalignment and scattered spatial distribution. Variational mode decomposition is used to separate load and power generation at different frequencies, deconstructing multi-scale variation patterns. Furthermore, mutual information entropy with a spatial distance attenuation factor is introduced to construct a correlation strength function, quantifying the strength of source-load coupling and filtering out features exceeding the threshold. This not only eliminates invalid and redundant data interference, improving the accuracy of mining the implicit spatiotemporal coupling relationship between power generation and consumption in the park, but also improves the quality of input features for subsequent power prediction models, thereby reducing prediction errors. It also provides reliable data support for the charging and discharging scheduling of each energy storage module, improving the efficiency of resource allocation for each energy storage module.

[0047] In one embodiment of this application, the spatiotemporal data matrix is ​​split into time series of different nodes based on the spatial dimension, and variational mode decomposition is performed on the time series of each node to obtain the intrinsic mode components of each node, including: Traverse the spatial dimension index of the spatiotemporal data matrix, use the spatial coordinates of each energy storage module as node identifiers, extract the data vectors of each node at all timestamps from the spatiotemporal data matrix, and construct the time series of each node. The initial modal decomposition number K and the second-order penalty factor α are calculated based on the signal length and sampling frequency of the time series at each node; where K represents the total number of intrinsic modal components to be obtained by decomposition. The convergence objective is to ensure that the center frequencies of each intrinsic mode component no longer overlap during the iteration process. The alternating direction multiplier method is used to perform variational mode decomposition on the time series of each node, and the spectral overlap between any two intrinsic mode components is calculated after each iteration. In the first iteration, the spectral overlap is not compared, and the next iteration begins. When the spectral overlap is greater than the spectral overlap calculated in the previous iteration, the value of k will be increased by 1, and the center frequency of the newly added intrinsic mode component will be initialized to the midpoint value of the maximum interval between any two adjacent center frequencies according to the current center frequency distribution of each intrinsic mode component, and the iterative decomposition will continue. When the rate of change of the center frequency of the modal components in two consecutive iterations is less than the preset rate of change of the center frequency, the decomposition stops and the intrinsic modal components of each node are obtained.

[0048] In this embodiment, the latitude and longitude coordinates or relative plane coordinates of the factory area obtained from on-site surveying of the smart industrial park are first entered, with each set of spatial coordinates bound to a node. The process of traversing the spatial dimension index of the spatiotemporal data matrix locates the corresponding coordinate nodes sequentially according to the column number of the spatiotemporal data matrix; data vectors continuously arranged throughout the entire time period are extracted for each node, with the vectors synchronously including the node's corresponding power generation, load power, and other operating values. Null values ​​caused by short-term equipment offline are removed, and the data are then concatenated to obtain the time series of each node. Each node's time series represents a continuous data set of dynamic changes in the power generation and consumption conditions of a single node over time.

[0049] Secondly, the initial number of mode decompositions, K, and the second-order penalty factor, α, are calculated based on the signal length and sampling frequency of the time series at each node. Here, K represents the total number of intrinsic mode components to be decomposed. Specifically, the signal length is determined based on the total number of sampling points in the time series at each node, and the sampling frequency is fixed by integrating the field acquisition devices. The initial values ​​of K and α are then synchronously calculated using empirical mapping relationships calibrated in engineering. For example, the initial value of K is appropriately increased for load time series with drastic fluctuations and dense sampling, while a smaller initial value of K is selected for distributed generation time series with gradual changes. The value of α is synchronously and adaptively adjusted with the sampling frequency.

[0050] Subsequently, with the convergence objective of ensuring that the center frequencies of each intrinsic modal component no longer overlap during the iteration process, variational mode decomposition is performed on the time series of each node using the alternating direction multiplier method. During the iterative calculation, each modal component, center frequency, and Lagrange multiplier are updated step by step according to the alternating direction multiplier method. In the first iteration, only parameter iteration and update are completed, and spectral data is retained but no overlap comparison is performed. Starting from the second iteration, all generated modal intrinsic modal components are paired one by one, and the spectral overlap is quantified by the percentage of overlapping area in the spectral interval. The overlap value corresponding to each iteration is recorded, and the overlap value is used for subsequent adaptive correction of the K value.

[0051] The alternating direction multiplier method is a numerical algorithm that iteratively solves an optimization problem by alternately optimizing multiple sets of variables step by step using the augmented Lagrange operator. The algorithm's inputs are the complete time series of each node in a single node, the initial number of decompositions K, the quadratic penalty factor α, and the iteration termination criterion (rate of change of center frequency). The outputs are the converged intrinsic modal components and their corresponding center frequencies. The algorithm comprises four computational levels. Specifically, the first layer is the augmented Lagrangian construction layer, used to construct the augmented Lagrangian function corresponding to the variational problem based on the input parameters and the original time series signal, integrating three types of variables to be optimized: modal components, center frequencies, and dual multipliers. The second layer is the variable alternation optimization layer, used to fix the remaining variables in turn step by step, optimize the intrinsic modal components individually, then fix the modal results to optimize the center frequencies, and finally update the Lagrangian dual multipliers to complete one round of alternation iteration. The third layer is the index verification layer, used to verify the overlap of the spectrum of each mode and the rate of change of the center frequency after each iteration, compare it with the preset judgment rules, and determine whether it is necessary to increase K and reinitialize the center frequencies of the newly added modes. The fourth layer is the convergence judgment and result output layer, used to terminate the iteration after the convergence condition is met and output all optimized intrinsic modal components. In this embodiment, when the spectral overlap is greater than that calculated in the previous iteration, the K value is increased by 1. Based on the current center frequency distribution of each intrinsic mode component, the center frequency of the newly added intrinsic mode component is initialized to the midpoint of the largest interval between any two adjacent center frequencies, and iterative decomposition continues. During the iterative calculation, all existing center frequency values ​​are sorted in real time, and the group of adjacent components with the largest frequency spacing is searched. The midpoint of the spacing is taken as the initial center frequency of the newly added component, thereby dispersing the frequency domain distribution and alleviating spectral aliasing. After updating the parameters, the original alternating direction multiplier method is used for iterative recalculation, and the trend of the overlap index is continuously observed.

[0052] When the rate of change of the center frequency of the modal components in two consecutive iterations is less than the preset rate of change of the center frequency, the decomposition stops, and the intrinsic modal components of each node are obtained. Specifically, during the iteration process, the relative change amplitude of the center frequency in each iteration is calculated. If the change amplitude in two consecutive rounds is less than the preset rate of change of the center frequency, the decomposition is considered to have converged, and the iteration operation is terminated. Finally, all modal data in the current iteration are retained as the intrinsic modal components of each node. Each intrinsic modal component corresponds to the fluctuation component of the source load in different periods, which is used for subsequent mutual information correlation analysis. The preset rate of change of the center frequency is set comprehensively based on the characteristics of the source load time series data of the smart industrial park, the accuracy requirements of the variational mode decomposition project, and the objectives of subsequent correlation feature mining. Specifically, on the one hand, by acquiring historical power generation and consumption time series samples from smart industrial parks, batch trial decomposition simulations are conducted to statistically analyze the baseline change amplitude of the natural fluctuation of the center frequency during the convergence phase of different operating conditions. The frequency variation ranges of the time series with slow power generation and the time series with drastic load fluctuations are distinguished to obtain the critical reference value for natural data convergence. On the other hand, based on the accuracy constraints of spatiotemporal mutual information entropy calculation and associated feature extraction, the system takes into account two problems: avoiding premature truncation of decomposition leading to under-decomposition of modes and the waste of computing power due to meaningless continuous iteration. At the same time, based on the resolution requirements of different frequency components for each energy storage module scheduling, the center frequency change rate threshold is finally determined. The center frequency change rate threshold is used to uniformly determine whether the variational mode decomposition of the time series of each node meets the convergence termination condition.

[0053] As can be seen from the above, this embodiment uses variational mode decomposition to adapt the initial decomposition parameter K and penalty factor α according to the time-series data characteristics of each energy storage node, avoiding the over-decomposition and under-decomposition problems caused by traditional fixed parameter decomposition. By iteratively monitoring the spectral overlap, dynamically supplementing modal components and optimizing the initial position of the center frequency, the mode aliasing phenomenon is effectively eliminated, and the time-series components with different frequency characteristics are accurately separated. At the same time, the rate of change of the center frequency is used as the convergence criterion to ensure the stability and accuracy of the decomposition results, fully preserving the multi-scale fluctuation characteristics of the power generation and consumption data of each node, providing high-quality and highly identifiable modal data support for subsequent calculation of spatiotemporal correlation strength and mining of source-load coupling laws, and significantly improving the accuracy of feature mining.

[0054] In one embodiment of this application, based on the predicted power generation and predicted electricity demand, an initial charging and discharging strategy for each energy storage module is determined, including: The predicted power generation is aligned with the predicted electricity demand based on time series, and the power difference sequence at each time segment is calculated. Based on the alternating positive and negative characteristics of the power difference sequence, the time interval of continuous power differences in the same direction is divided into a window to obtain the scheduling window; Based on the power limit of the interaction between electricity price information and the main power grid, calculate the cumulative power deviation from the start time of the dispatch window to the current time; The power difference sequence is smoothed using a moving average filter to generate the power adjustment coefficient for the scheduling window. Calculate the response priority of each energy storage module based on its health status and the current state of charge of each module. Based on the response priority and power adjustment coefficient, the charging and discharging power of each energy storage module is allocated to obtain the charging and discharging power and charging and discharging time period of each energy storage module, which serves as the initial charging and discharging strategy.

[0055] In this embodiment, firstly, the predicted power generation and predicted electricity demand are sampled at time intervals and timestamps to eliminate time misalignment issues caused by model output delays and inconsistent sampling dimensions. After accurate time-series matching, supply-demand difference calculations are performed at each scale, matching real-time load consumption with power generation capacity to obtain the power difference at each time point. The power differences of all time-series nodes are arranged in chronological order to form a continuous, complete power difference sequence that represents the supply changes in the smart industrial park.

[0056] Secondly, the power difference sequence is iterated and judged. Consecutive positive power difference intervals are uniformly identified as power surplus intervals, corresponding to the energy storage module charging scheduling scenario; consecutive negative power difference intervals are uniformly identified as power shortage intervals, corresponding to the energy storage module discharging scheduling scenario. Whenever the sign of the power difference reverses, the previous window is truncated and a new scheduling window is opened, thus dividing the daily supply and demand curve into multiple scheduling windows with simple operating conditions and stable trends.

[0057] Subsequently, within each dispatch window, based on the fluctuation pattern of time-of-use electricity prices in the electricity price information and the upper and lower limits of the power exchange with the main power grid, the maximum power purchased and delivered between the smart industrial park and the external grid is limited. Taking the start time of the dispatch window as the calculation base point, the offset between the supply and demand difference within the window and the theoretical equilibrium value is accumulated moment by moment to obtain the cumulative power deviation. This cumulative power deviation can represent the overall supply and demand imbalance within the current window and is used to correct the energy storage dispatch intensity under extreme electricity prices and grid quota constraints.

[0058] This embodiment selects a sliding window length adapted to the source-load fluctuation characteristics of the smart industrial park, and performs sliding filtering on the high-frequency fluctuating power difference sequence to filter out prediction errors and high-frequency noise caused by instantaneous load impacts, while retaining the overall trend of power difference changes. The smoothed steady-state power difference is then compared with the original instantaneous power difference, and normalization correction is performed based on the cumulative deviation of the window, ultimately obtaining the power adjustment coefficient for the corresponding scheduling window. This power adjustment coefficient is used to adjust the overall charging and discharging capacity of each energy storage module. Based on the health status and the current state of charge of each energy storage module, the response priority of each energy storage module is calculated. The response priority is directly proportional to the health status and inversely proportional to the degree to which the current state of charge deviates from the target state of charge. Specifically, the health status of each energy storage module is used as the evaluation benchmark. The better the health status of the energy storage module, the stronger its ability to withstand charging and discharging shocks and the more scheduling tasks it can undertake. At the same time, the target state of charge is the optimal operating range for energy storage. The further the current state of charge of the energy storage module deviates from this range, the smaller its operational redundancy and the worse its scheduling adaptability. The response priority of each energy storage module is obtained, and the response priority is used to realize differentiated rating of energy storage modules with different aging levels and different remaining power states.

[0059] Finally, based on the response priority and power adjustment coefficient, the charging and discharging power of each energy storage module is allocated to obtain the charging and discharging power and charging and discharging time period of each energy storage module, which serves as the initial charging and discharging strategy.

[0060] As can be seen from the above, this embodiment improves the accuracy of time-based management and control of energy storage module scheduling by aligning power generation and consumption forecast data with time series and constructing a power difference sequence, and dividing the scheduling window according to the same-direction difference characteristics. Furthermore, by calculating the cumulative power deviation through electricity price and grid interaction limits and optimizing the power adjustment coefficient through moving average filtering, it effectively reduces drastic power fluctuations, ensures grid interaction stability, and improves scheduling economy. Simultaneously, it dynamically calculates response priorities based on the health status and state-of-charge deviation of each energy storage module, prioritizing the use of energy storage modules with excellent operating conditions and optimal status for scheduling. This strategy takes into account grid constraints, electricity price benefits, and the health and lifespan of energy storage equipment, achieving a rational and differentiated allocation of charging and discharging power for each energy storage module, and providing accurate and reliable initial scheduling basis for modular energy storage collaborative scheduling.

[0061] In one embodiment of this application, charging and discharging power is allocated to each energy storage module according to response priority and power adjustment coefficient to obtain the charging and discharging power and charging and discharging time period of each energy storage module, which serves as the initial charging and discharging strategy, including: Calculate the net total charging and discharging power within the scheduling window based on the power adjustment coefficient and the power difference within the scheduling window; The net total charging and discharging power is initially allocated based on the response priority of each energy storage module, resulting in the initial allocated power for each energy storage module. Based on the current state of charge, maximum charge / discharge power limit, and health status of each energy storage module, an adjustable range for each energy storage module is constructed. If the initial allocated power does not exceed the adjustable range of the corresponding energy storage module, then the initial allocated power shall be taken as the final allocated power of the energy storage module. If the initial allocated power exceeds the adjustable range of the corresponding energy storage module, the boundary value of the adjustable range of the energy storage module is taken as the final allocated power, and the power difference of the excess part is calculated. The total amount of power difference exceeding the limit is summed up to obtain the total amount of power to be redistributed. Based on the response priority of the remaining energy storage modules that are within the adjustable range, the total amount of power to be redistributed is then re-allocated. If the allocated power of all energy storage modules falls within the corresponding adjustable range within the preset maximum number of iterations, then the final allocated power of each energy storage module and the corresponding charging and discharging period will be used as the initial charging and discharging strategy.

[0062] In this embodiment, quantitative calculations are first performed based on the power adjustment coefficient and the power difference corresponding to the scheduling window to calculate the net total charging and discharging power within the scheduling window. This net total charging and discharging power is the overall control power benchmark that all energy storage modules in this scheduling window need to jointly bear, and it is also the basic data for subsequent power allocation calculations. During the calculation of the net total charging and discharging power, the electricity price constraints and external grid power interaction limits of this scheduling window are bound to ensure that the net total charging and discharging power conforms to the actual grid connection operation rules of the park.

[0063] After obtaining the total net charge / discharge power, the response priority of each energy storage module is used as the power allocation weight. The weight value corresponds to the power proportion of the module in the initial allocation. The initial allocation of the total net charge / discharge power is completed according to the weight, and the initial allocated power for each energy storage module is calculated one by one. Furthermore, the adjustable range of each energy storage module is constructed based on its current state of charge, maximum charge / discharge power limit, and health status.

[0064] Secondly, the initial allocated power is compared and verified with the adjustable range of the corresponding energy storage module. Specifically, for any energy storage module, when its initial allocated power is within its own adjustable range, the initial allocated power is determined as the final allocated power of the energy storage module; once the initial allocated power is greater than the upper limit of the adjustable range or less than the lower limit of the adjustable range, the original initial allocated power is discarded, and the boundary value corresponding to the adjustable range is selected as the final allocated power of the module.

[0065] Subsequently, the power differences exceeding the limits for all over-limit energy storage modules are aggregated and summed to obtain a unified total power to be redistributed. This total power to be redistributed represents the remaining controllable power that could not be carried by the original energy storage due to equipment operation limitations during the initial allocation and needs to be transferred to other compliant energy storage. Remaining energy storage modules whose allocated power after the initial allocation did not exceed their own adjustable range are selected. The response priority of these modules is used as the new allocation weight for a secondary allocation operation on the total power to be redistributed. The cyclical steps of power verification, over-limit aggregation, and remaining module redistribution are repeated, with a preset maximum number of iterations as the upper limit for the loop termination. During the iteration process, the allocated power of each energy storage module is continuously adjusted. If the allocated power of all energy storage modules falls within their respective adjustable ranges before reaching the preset maximum number of iterations, the iterative allocation process terminates, and the final allocated power of all energy storage modules is aggregated. The final allocated power of each energy storage module and its corresponding charging / discharging period are used as the initial charging / discharging strategy. The preset maximum number of iterations is determined step-by-step based on four conditions: the total number of energy storage modules in the park, the adjustable range margin of individual energy storage units, the fluctuation characteristics of source-load power difference, and the real-time constraints of scheduling tasks. Specifically, the maximum number of iterations is first determined based on the actual number of energy storage modules installed in the smart industrial park. The more modules there are and the more complex the combined scenarios of secondary power redistribution, the higher the required number of iterations will be.

[0066] As can be seen from the above, this embodiment first determines the total dispatch power based on the adjustment coefficient and the difference between the window power and the total power. The initial power allocation is then completed through priority weights. Next, the adjustable power range for each module is defined based on the state of charge, power limits, and health status. Excess power is collected and then used to conduct a secondary iterative allocation based on the priority of the remaining energy storage modules. This hierarchical allocation mode adheres to the energy storage operation boundary constraints, reducing damage to equipment from individual overcharging and over-discharging, while fully leveraging the dispatch potential of high-quality energy storage and achieving efficient reuse of power margins. Furthermore, the iterative convergence mechanism ensures compliance with allocation results, optimizing power distribution while meeting equipment safety requirements. This guarantees that the initial charging and discharging strategy balances economy and energy storage lifespan, laying a reasonable benchmark for subsequent collaborative optimization. Each individual module is a single energy storage module.

[0067] In one embodiment of this application, an adjustable range for each energy storage module is constructed based on its current state of charge, maximum charge / discharge power limit, and health status, including: Based on the maximum charge / discharge power limits and health status of each energy storage module, a steady-state rigid boundary is constructed. Based on the deviation between the state of charge of each energy storage module and the preset standard state of charge, as well as the decay between healthy states, a transient elastic boundary is constructed. Obtain the current operating mode of each energy storage module, including charging mode, discharging mode and standby mode; Based on the current operating mode, determine the enabling state of the limiting transient elastic boundary to obtain the target enabling state; Determine the elastic buffer based on the target enable state; By superimposing the steady-state rigid boundary with the elastic buffer, the adjustable range of each energy storage module is obtained; among them, the power distribution within the elastic buffer requires triggering an additional safety verification mechanism.

[0068] In this embodiment, the maximum charge and discharge power limits of each energy storage module are first calculated and corrected according to the health status to construct a steady-state rigid boundary. The steady-state rigid boundary is the absolute power limit that cannot be exceeded during long-term continuous operation. The worse the health status of the energy storage module, the more the steady-state rigid boundary will shrink inward due to the aging of the battery cells and the increase in internal resistance. From the perspective of hardware durability, the power red line that cannot be crossed during long-term stable operation of energy storage is locked. Regardless of the short-term dispatch demand, the power allocation under continuous operation conditions cannot exceed the boundary constraint.

[0069] Secondly, the deviation is calculated by subtracting the state of charge (SOC) of each energy storage module from the preset standard SOC. Then, the decay of the health status is obtained by extracting statistical data from the full life cycle operation data of each energy storage module, and a transient elastic boundary is constructed. The transient elastic boundary allows short-term exceedance of the steady-state rigid boundary, and the exceedance is proportional to the deviation and inversely proportional to the decay. The greater the deviation of the SOC from the preset standard SOC, the greater the allowable range of short-term power. The more severe the decay of the health status, the weaker the cell's ability to withstand instantaneous overload, and the corresponding short-term over-limit range will be compressed. This boundary only applies to instantaneous short-term scheduling conditions and is not used as a basis for long-term power operation. Simultaneously, the current operating modes of each energy storage module are collected, including charging mode, discharging mode, and standby mode. The heating characteristics and electrochemical reaction rates of the energy storage cells differ under different operating modes, resulting in varying tolerances to elastic power. This serves as a distinguishing condition for boundary control. Based on the safety control logic for different operating conditions and the current operating mode, the enabling state for limiting transient elastic boundaries is determined, resulting in the target enabling state. In standby mode, the enabling permissions for transient elastic boundaries are mostly disabled, while in charging or discharging conditions, some elastic permissions are enabled as needed. Specifically, in charging mode, only the elastic boundaries in the charging direction are enabled; in discharging mode, only the elastic boundaries in the discharging direction are enabled. The preset standard state of charge (SOC) is based on the midpoint of the optimal SOC range designed at the factory for the energy storage cells, and is ultimately determined after comprehensive verification based on the full-cycle charging and discharging historical operating data collected by the battery management system, long-term stable operation losses, and aging statistical experience.

[0070] Based on the target enable state, the available range of the transient elastic boundary is trimmed and constrained, and the over-limit range that is prohibited from being used under the current operating condition is eliminated. An elastic buffer is defined. When the target enable state is off, the elastic buffer is empty, and only the steady-state rigid boundary range is retained.

[0071] Finally, the steady-state rigid boundary and the elastic buffer are superimposed to obtain the adjustable range of each energy storage module; among them, the power distribution within the elastic buffer needs to trigger an additional safety verification mechanism.

[0072] As can be seen from the above, this embodiment constructs an adjustable energy storage range by dividing a steady-state rigid boundary and a transient elastic boundary. The rigid boundary sets a long-term insurmountable limit based on the rated power and health status, fundamentally reducing the damage to the battery cells caused by over-limit operation of energy storage. The over-limit range of the elastic boundary is dynamically determined based on the state of charge and the degree of health degradation, and the elastic buffer zone is flexibly enabled by charging, discharging, and standby conditions. The dual-layer interval architecture not only strengthens the long-term operational safety baseline of the equipment through rigid constraints, but also improves the energy storage scheduling margin with the help of elastic space, improving the adaptability to short-term power fluctuation scenarios of source and load in smart industrial parks. At the same time, a special verification is added to the power within the buffer zone, taking into account both scheduling flexibility and equipment safety, and optimizing the adaptability of modular energy storage power allocation.

[0073] In one embodiment of this application, power allocation within the elastic buffer requires triggering additional security verification mechanisms, including: When the power allocated to any energy storage module falls within the elastic buffer zone of the adjustable range, the energy storage module is marked as an elastic call module; Get the current temperature, temperature rise rate, and health status decay rate of the elastic call module; Calculate the first safety factor based on the ratio of the current temperature to the preset temperature safety threshold; The second safety factor is calculated based on the ratio of the rate of temperature rise to the preset maximum allowable rate of temperature rise. The third safety factor is calculated based on the ratio of the health status decay rate to the preset decay rate alarm threshold. The first security factor, the second security factor, and the third security factor are weighted and summed to obtain the comprehensive security risk index of the elastic call module; If the comprehensive safety risk index is greater than the preset safety risk index threshold, the elastic power allocation of the elastic call module will be rejected, the final allocated power of the module will be forcibly limited to its steady-state rigid boundary value, and the power gap will be redistributed to other energy storage modules within the steady-state rigid boundary. If the overall security risk index is less than or equal to the preset security risk index threshold, the elastic call module is allowed to execute based on the allocated power in the elastic buffer.

[0074] In this embodiment, when the power allocated to any energy storage module falls within the adjustable range of the elastic buffer, the energy storage module is marked as an elastic call module, and the energy storage module is simultaneously locked and enters a special security verification process.

[0075] The system collects three key operating parameters from the elastic call module: current temperature, temperature rise rate, and health status decay rate. A first safety factor is calculated based on the ratio of the current temperature to a preset temperature safety threshold; a second safety factor is calculated based on the ratio of the temperature rise rate to a preset maximum allowable temperature rise rate; and a third safety factor is calculated based on the ratio of the health status decay rate to a preset decay rate alarm threshold. Higher values ​​for these three safety factors indicate a greater potential hazard in any single operation. The preset temperature safety threshold is the factory-calibrated benchmark value for the energy storage cell's heat resistance parameters; the preset maximum allowable temperature rise rate is based on factory thermal performance test data and optimized using historical overload temperature rise failures in smart industrial park energy storage systems; and the preset decay rate alarm threshold is calibrated using accelerated aging tests at the battery factory.

[0076] This embodiment calculates a weighted sum of the first, second, and third safety factors to obtain a comprehensive safety risk index for the flexible power delivery module. This comprehensive safety risk index integrates three risk dimensions: temperature, rate of temperature rise, and health aging, and can represent the overall safety level of each energy storage module carrying flexible power. The weights of each factor in the weighted summation are determined based on historical fault statistics for each energy storage module, the frequency of fault occurrence corresponding to each safety factor, and the severity of the accident. Specifically, historical samples of risk scenarios corresponding to the first, second, and third safety factors are first analyzed, and the probability of occurrence and the degree of fault loss caused by the three factors are statistically analyzed. The weight values ​​are then obtained based on the proportion of the probability of occurrence and the degree of fault loss of the safety hazards.

[0077] Specifically, if the comprehensive safety risk index exceeds a preset safety risk index threshold, the energy storage module is deemed unable to safely handle the power within the elastic range. Therefore, the elastic power allocation for this module is rejected, and its final allocated power is forcibly limited to its steady-state rigid boundary value. The resulting power gap is separately aggregated and redistributed to other energy storage modules within the steady-state rigid boundary according to predetermined allocation rules. If the comprehensive safety risk index is less than or equal to the preset safety risk index threshold, all operating indicators of the energy storage module are deemed to be within a controllable and safe range. In this case, the elastic power allocation module is allowed to execute based on the allocated power within the elastic buffer. The preset safety risk index threshold is jointly calibrated using threshold standards corresponding to three safety factors and historical fault statistical samples. Specifically, firstly, the threshold standards corresponding to the three safety factors are weighted, with the weights determined by the historical failure probability and failure loss level. A weighted formula is used to deduce the comprehensive risk value corresponding to the critical failure. Secondly, historical data from the smart industrial park is aggregated to calculate the comprehensive safety risk index corresponding to when energy storage experiences overheating or accelerated aging failures. The value where the comprehensive risk value corresponding to the critical failure calculated using the weighted formula matches the calculated comprehensive safety risk index is used as the preset safety risk index threshold. When the comprehensive safety risk index exceeds the preset threshold, it indicates that multiple hidden dangers are nearing the failure boundary, and flexible power deployment is prohibited. The established allocation rule uses the response priority of the remaining compliant energy storage modules as the allocation weight to redistribute the power gap caused by the restriction of flexible module deployment.

[0078] As can be seen from the above, this embodiment performs a multi-dimensional safety evaluation of energy storage modules entering the elastic buffer zone. Three safety factors are constructed based on temperature, temperature rise rate, and health degradation rate, respectively. A comprehensive risk index is obtained through weighting, achieving quantitative control of elastic power dispatch. When the risk exceeds the limit, the power is forcibly locked to a rigid boundary and the power gap is transferred. Elastic power can only be activated if the indicators are qualified. This approach not only fully utilizes short-term elastic dispatch space to smooth power fluctuations and improve energy storage utilization, but also prevents overheating and accelerated aging risks caused by elastic operating conditions in real time. It avoids the aggravated losses due to excessive operation of energy storage, ensuring dispatch flexibility while maintaining the bottom line of equipment safety.

[0079] In one embodiment of this application, a modular energy storage method further includes: If the preset maximum number of iterations is reached and there is still an unconverged power difference exceeding the adjustable range, then the product of the adjustable range margin and response priority of each energy storage module in the current iteration round is obtained to construct a relaxation factor vector. The dominant charging / discharging mode of the scheduling window is determined based on the sign direction of the unconverged power difference. When the scheduling window is in charging-dominated mode, the energy storage modules are sorted from high to low according to the upper limit of the adjustable range. The power difference of the excess is redistributed to the energy storage module with the highest upper limit of the adjustable range in decreasing relaxation step size until the power difference is zero or all modules reach their expansion limit. The expansion limit is the original upper limit of the adjustable range multiplied by the preset relaxation coefficient. When the scheduling window is in discharge-dominant mode, the energy storage modules are sorted from low to high according to the lower limit of the adjustable range, and the excess power difference is allocated to the energy storage module with the lowest lower limit of the adjustable range. The final allocated power of each energy storage module after relaxation is used as the initial charging and discharging strategy, and an alarm message containing an incomplete convergence flag is generated to trigger the power adjustment coefficient correction in the next scheduling window.

[0080] In this embodiment, during the execution of the preceding energy storage power iterative allocation process, the number of iterations is continuously counted. Once the calculation reaches the preset maximum number of iterations, and there is still an unconverged power difference exceeding the adjustable range, it indicates that the remaining power to be allocated cannot be absorbed according to the allocation rules based on the adjustable range boundary and response priority. At this time, the fallback relaxation allocation logic is activated. Specifically, the adjustable range margin and response priority of each energy storage module in the current iteration are extracted one by one, and the two parameters are multiplied to complete the numerical calculation. The relaxation factor vector is constructed by orderly arranging the calculation results of all modules. This relaxation factor vector quantifies the potential capacity of each energy storage module to temporarily expand and absorb surplus power.

[0081] This embodiment distinguishes the positive and negative signs of the unconverged power difference. Based on the sign of the unconverged power difference, the charging and discharging dominant mode of the scheduling window is determined. A positive value indicates that the overall surplus power needs to be absorbed by energy storage, corresponding to the charging dominant mode. A negative value indicates that the overall power deficit needs to be supplied by energy storage, corresponding to the discharging dominant mode. This distinguishes the two different relaxation allocation logics.

[0082] Specifically, when the scheduling window is in charging-dominated mode, energy storage modules are prioritized from highest to lowest based on the upper limit of their adjustable range. This ranking serves as the power replenishment order, and the excess power difference is allocated sequentially to the energy storage module with the highest adjustable range upper limit using decreasing relaxation steps. During allocation, the power upper limit of the energy storage module is no longer constrained by the upper limit of its adjustable range; the expanded power upper limit is multiplied by a preset relaxation coefficient. This allocation operation continues cyclically until the unconverged power difference reaches zero, or all participating modules reach their own expanded power limits, terminating the current round of charging-side relaxation allocation. When the scheduling window is in discharging-dominant mode, the system switches to discharging-adaptive allocation rules. Energy storage modules are ranked from lowest to highest based on the lower limit of their adjustable range. The module with the lowest lower limit and the largest discharge margin is prioritized to take over the remaining unconverged power difference. The power to be allocated is distributed sequentially according to the ranking order, absorbing the remaining power gap on the discharging side within the boundary constraints of the relaxation expansion. The preset relaxation coefficient is determined offline based on historical health status data of each energy storage module, the adjustable range margin of each cell, temperature and temperature rise limit parameters collected by the battery management system, and the short-term overload tolerance capability of the equipment. For example, the first step involves retrieving historical health status data stored in the battery management system and calculating the health degradation rate of each energy storage module. This rate is compared with the degradation rate alarm threshold, and health levels are classified based on the degree of aging. Higher health levels allow for greater capacity expansion, serving as the basis for increasing the relaxation coefficient. The second step involves extracting measured data of the adjustable range margin of each cell. A larger margin indicates more power surplus under normal operating conditions, allowing for a moderate increase in the relaxation coefficient; a smaller margin limits the coefficient's upward movement. The third step involves comparing real-time temperature and historical temperature rise data collected by the battery management system with the temperature safety threshold and the maximum allowable temperature rise rate to determine the normal temperature rise margin of the energy storage module. Sufficient temperature rise margin allows for an increase in the relaxation coefficient, while daily temperature rise close to the limit requires a decrease in the coefficient, ultimately yielding the preset relaxation coefficient. Finally, the four indicators are quantified and scored separately, and the weighted summaries are made according to the weight of each indicator's impact on energy storage safety. Then, the results of past relaxation allocation simulations in the park are combined to make minor adjustments and corrections. Finally, the preset relaxation coefficient for each energy storage module is determined and used to calculate the expansion limit.

[0083] After the relaxation allocation is completed, the final allocated power of each energy storage module obtained after capacity expansion and adjustment is fixed, and this serves as the initial charging and discharging strategy for this scheduling window. Simultaneously, an alarm message containing an incomplete convergence flag is generated. This alarm message is archived and uploaded synchronously with the window's operational data. When the power calculation phase begins in the next scheduling window, the scheduling control terminal reads this flag and triggers a targeted correction of the power adjustment coefficient for that window. This rolling optimization approach mitigates the impact of the current allocation non-convergence.

[0084] As can be seen from the above, this embodiment constructs a relaxation factor by multiplying the adjustable margin and priority, integrating the differentiated allocation margin based on the power distribution of the dominant charging and discharging modes. During charging, capacity is preferentially expanded according to the upper limit of power; during discharging, allocation is preferentially distributed according to the lower limit. By appropriately relaxing the upper limit of module power through the relaxation coefficient, the remaining power to be allocated is effectively utilized. Simultaneously, unconverged flags are retained and adjusted in the next cycle, avoiding the inability to allocate dispatched power due to capacity constraints, thus improving the efficiency of energy storage resource utilization. Furthermore, closed-loop rolling correction continuously optimizes subsequent dispatch accuracy.

[0085] In one embodiment of this application, when the probability of a fault occurrence is greater than a preset first fault occurrence probability threshold and less than a preset second fault occurrence probability threshold, the cooperative control scheme is modified based on the fault type to obtain a target cooperative control scheme, including: When the probability of a fault occurrence is greater than a preset first fault occurrence probability threshold and less than a preset second fault occurrence probability threshold, the extreme type of the current operating scenario is identified. The extreme types include sudden power change scenarios and extreme electricity price fluctuation scenarios. Obtain the fault type output by the fault analysis model. When the fault type is an overcurrent fault or an overvoltage fault, match and correct the fault type with the extreme types to obtain the target cooperative control scheme, where: If the fault type is an overcurrent fault and the current scenario is a sudden change in grid power, then the power change rate in the coordinated control scheme will be limited to the preset maximum change rate, and the power waveform will be low-pass filtered. If the fault type is an overvoltage fault and the current scenario is an extreme electricity price fluctuation scenario, then the voltage regulation coefficient is calculated based on the electricity price deviation, and the upper limit of the charging power of the corresponding energy storage module is reduced based on the voltage regulation coefficient. If the fault type is not directly related to the current scenario, the collaborative control scheme is modified based on the corresponding correction strategy in the preset fault type-correction table.

[0086] In this embodiment, when the probability of a fault occurrence is greater than a preset first fault occurrence probability threshold but less than a preset second fault occurrence probability threshold, it indicates that each energy storage module has a moderate fault risk, but has not yet reached the danger level of an emergency shutdown alarm. Therefore, the targeted correction process of the collaborative control scheme is initiated. Simultaneously, real-time power grid operation data and time-of-use electricity price data from the smart industrial park are captured to identify the extreme types of the current operating scenario. These extreme types include sudden power grid power fluctuations and extreme electricity price volatility scenarios. First, fault type data synchronously output from the fault analysis model is retrieved to identify whether the fault type is an overcurrent fault or an overvoltage fault. This is used to define differentiated correction logic, matching the fault type with extreme types to obtain the target collaborative control scheme. Specifically, if the fault type is an overcurrent fault and the current scenario is a sudden change in grid power, it means that grid-side power disturbances will instantly cause a jump in energy storage charging and discharging power, leading to an impact overcurrent problem. Therefore, the power change rate in the collaborative control scheme needs to be limited to a preset maximum change rate, and the power waveform needs to be low-pass filtered to remove instantaneous power spikes and smooth the power regulation curve, suppressing the probability of overcurrent faults from the power change rhythm. If the fault type is an overvoltage fault and the current scenario is an extreme electricity price fluctuation scenario, large price fluctuations can easily drive energy storage to blindly charge at full power under high price incentives, leading to continuous high voltage accumulation in the cells and inducing overvoltage hazards. In this case, the electricity price deviation is calculated based on the difference between the real-time electricity price and the benchmark electricity price, and then the voltage regulation coefficient is calculated based on the electricity price deviation.

[0087] The preset maximum rate of change is set comprehensively based on the electrical tolerance parameters of the energy storage unit, the carrying capacity of the industrial park's power grid, and historical power surge fault data. For example, firstly, the allowable power ramp-up rate and the inverter's tolerance to instantaneous power surge indicators are extracted from the energy storage equipment's factory specifications to obtain the inherent limit rate of change at the hardware level. Then, samples of multiple rounds of power surge conditions in the smart industrial park's history are collected, and the critical power change value that induces overcurrent faults in the energy storage is statistically analyzed. Based on the short-term current surge tolerance capacity of the park's distribution lines, grid-side constraint boundaries are defined. A safety margin is reserved downwards based on the equipment hardware limits and the grid safety threshold. Simultaneously, the probability of overcurrent fault occurrence under different power change rates is verified, and a fixed value that meets both daily power regulation needs and avoids overcurrent problems caused by sudden surges is selected as the preset maximum rate of change. The reserved safety margin is a rate surplus space left downwards based on the two upper limit benchmark values: the energy storage hardware limit power change rate and the grid tolerance critical power change rate. The preset first fault occurrence probability threshold and second fault occurrence probability threshold are calibrated in segments based on the historical fault statistics of the energy storage module throughout its entire lifecycle and the requirements of hierarchical management. Specifically, the historical operation data of similar energy storage in smart industrial parks are first compiled, and equipment failure instances corresponding to different failure probabilities are statistically analyzed. Three probability intervals are defined as follows: normal operation without hidden dangers, moderate risk with emerging hidden dangers, and high risk of impending failure.

[0088] Specifically, firstly, a benchmark electricity price for the industrial park is pre-selected as the calculation benchmark. This benchmark price can be either the normal stable electricity price within the dispatch cycle or the base price set a day before. The difference between the real-time executed electricity price and the benchmark electricity price is calculated to obtain the price deviation value. Then, the price deviation is divided by the benchmark electricity price to obtain the price deviation degree. Subsequently, a correlation mapping relationship is established between the price deviation degree and the voltage regulation coefficient. The voltage regulation coefficient ranges from 0 to 1. The larger the price deviation degree, the more significant the deviation of the electricity price from the normal level, and the higher the risk of overvoltage induced by overcharging of energy storage. The corresponding voltage regulation coefficient decreases accordingly. The smaller the price deviation degree, the closer the voltage regulation coefficient is to 1. Based on the mapping rule and the price deviation degree, the voltage regulation coefficient is finally obtained. According to the voltage regulation coefficient, the upper limit of the charging power of the corresponding energy storage module is reduced. By constraining the full-load charging space, the overvoltage condition of the battery cell is suppressed, thereby achieving source control of fault risk. If the fault type is not directly related to the current scenario, it means that the fault cause is not an external disturbance caused by the two extreme scenarios, but is caused by internal factors such as the aging of the energy storage unit itself and the deterioration of components. In this case, the collaborative control scheme is modified based on the correction strategy corresponding to the preset fault type-correction table. The charging and discharging allocation logic and operating power limit are adjusted according to the various fault-specific rectification rules stored in the preset fault type-correction table. Finally, the collaborative control scheme is optimized and a target collaborative control scheme adapted to the fault risk is generated.

[0089] As can be seen from the above, this embodiment distinguishes between two extreme operating conditions—sudden changes in grid power and extreme electricity price fluctuations—for medium-risk fault ranges, and combines overcurrent and overvoltage fault types with differentiated optimization and control schemes. When sudden grid changes induce overcurrent, the power change rate is constrained and low-pass filtering is used to suppress instantaneous power surges; when electricity price fluctuations cause overvoltage, the voltage regulation coefficient is calculated based on the electricity price deviation, and the charging upper limit is lowered, reducing voltage exceedances from the source. When the fault and scenario do not match, a standardized optimization control strategy is implemented through a preset correction form. This approach can address potential fault hazards without directly triggering shutdown alarms, preserving the flexibility of energy storage dispatch and electricity price arbitrage profits, while proactively avoiding overcurrent and overvoltage faults, thus balancing system operational safety and the economic benefits of park energy storage dispatch.

[0090] Corresponding to the modular energy storage method in the above embodiments, Figure 2 This is a structural block diagram of a modular energy storage system provided according to an embodiment of this application. For ease of explanation, only the parts relevant to the embodiment of this application are shown. References Figure 2 The modular energy storage system 20 includes: a multi-source data acquisition module 21, a data feature analysis module 22, a source-load power prediction module 23, a preliminary charging and discharging strategy module 24, a collaborative scheme generation module 25, a fault prediction and calculation module 26, and a graded risk control and disposal module 27.

[0091] Among them, the multi-source data acquisition module 21 is used to collect the operating data of power generation equipment, power grid and power consumption equipment in the smart industrial park, as well as the basic operating parameters and extended parameters of each energy storage module; The data feature analysis module 22 is used to standardize and mine the operating data, extract the correlation features between power generation and power consumption, fuse the basic operating parameters and extended parameters, and determine the health status of each energy storage module. The source-load power prediction module 23 is used to input the associated features into the preset power generation and power consumption prediction model to obtain the predicted power generation and predicted power consumption. The initial charging and discharging strategy module 24 is used to determine the initial charging and discharging strategy of each energy storage module based on the predicted power generation and predicted electricity demand. The collaborative scheme generation module 25 is used to perform task allocation and power balancing of the initial charging and discharging strategy based on the health status of each energy storage module, and generate a collaborative control scheme between each energy storage module. The fault prediction and calculation module 26 is used to input the collaborative control scheme and health status into the preset fault analysis model to obtain the predicted fault types and fault occurrence probabilities that will occur when each energy storage module executes the collaborative control scheme. The graded risk control and handling module 27 is used to modify the collaborative control scheme based on the fault type when the probability of a fault occurrence is greater than the preset first fault occurrence probability threshold and less than the preset second fault occurrence probability threshold, to obtain a target collaborative control scheme, and to adjust the charging and discharging power of each energy storage module based on the target collaborative control scheme; when the probability of a fault occurrence is greater than the preset second fault occurrence probability threshold, a warning message is generated and output.

[0092] See Figure 3 , Figure 3 This is a schematic block diagram of an electronic device provided according to an embodiment of this application. Figure 3 The electronic device 300 in this embodiment may include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The processors 301, input devices 302, output devices 303, and memories 304 communicate with each other via a communication bus 305. The memories 304 store computer programs, including program instructions. The processors 301 execute the program instructions stored in the memories 304. Specifically, the processors 301 are configured to invoke the program instructions to perform the functions of the modules in the aforementioned device embodiments, for example... Figure 2 The functions of the multi-source data acquisition module 21, data feature analysis module 22, source-load power prediction module 23, preliminary charging and discharging strategy module 24, collaborative scheme generation module 25, fault prediction and calculation module 26, and graded risk control and disposal module 27 are shown.

[0093] It should be understood that, in the embodiments of this application, the processor 301 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0094] Input device 302 may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint orientation information), a microphone, etc., and output device 303 may include a display (LCD, etc.), a speaker, etc.

[0095] The memory 304 may include read-only memory and random access memory, and provides instructions and data to the processor 301. A portion of the memory 304 may also include non-volatile random access memory. For example, the memory 304 may also store device type information.

[0096] In specific implementations, the processor 301, input device 302, and output device 303 described in the embodiments of this application can execute the implementation methods described in any embodiment of the modular energy storage method provided in the embodiments of this application, or they can execute the implementation methods of the electronic devices described in the embodiments of this application, which will not be repeated here.

[0097] In another embodiment of this application, a computer-readable storage medium is provided. This computer-readable storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the methods described above. Alternatively, the computer program can instruct related hardware to complete the process. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0098] The computer-readable storage medium can be an internal storage unit of the electronic device in any of the foregoing embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device. Furthermore, the computer-readable storage medium can include both internal and external storage units of the electronic device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0099] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0100] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the electronic devices and units described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0101] In the several embodiments provided in this application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces or units, or it may be an electrical, mechanical, or other form of connection.

[0102] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0103] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0104] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A modular energy storage method, characterized in that, include: Collect operational data from power generation equipment, the main power grid, and power consumption equipment in the smart industrial park, as well as the basic and extended operational parameters of each energy storage module; The operational data is standardized and data-mined to extract the correlation features between power generation and power consumption; The basic operating parameters and extended parameters are fused together, and the health status of each energy storage module is determined. The associated features are input into a preset power generation and electricity consumption prediction model to obtain the predicted power generation and predicted electricity consumption. Based on the predicted power generation and the predicted electricity demand, the initial charging and discharging strategy for each energy storage module is determined. Based on the health status of each energy storage module, the initial charging and discharging strategy is assigned tasks and power balanced to generate a collaborative control scheme among the energy storage modules. The collaborative control scheme and the health status are input into a preset fault analysis model to obtain the predicted fault types and fault occurrence probabilities that will occur when each energy storage module executes the collaborative control scheme. When the probability of the fault occurrence is greater than a preset first fault occurrence probability threshold and less than a preset second fault occurrence probability threshold, the collaborative control scheme is modified based on the fault type to obtain a target collaborative control scheme, and the charging and discharging power of each energy storage module is adjusted based on the target collaborative control scheme; when the probability of the fault occurrence is greater than the preset second fault occurrence probability threshold, a warning message is generated and output.

2. The modular energy storage method according to claim 1, characterized in that, The standardization and data mining of the operational data to extract the correlation features between power generation and power consumption includes: The operational data is double-aligned based on timestamps and the spatial layout of the energy storage modules to construct a spatiotemporal data matrix; The spatiotemporal data matrix is ​​split into time series of different nodes based on the spatial dimension, and variational mode decomposition is performed on the time series of each node to obtain the intrinsic mode components of each node. Calculate the spatiotemporal mutual information entropy of power generation and power consumption mode components between different spatial nodes and within the same node, and construct a correlation strength function with spatial distance as the attenuation factor; Node pairs and their corresponding modal component labels that are greater than a preset node threshold in the correlation strength function are extracted as the correlation features between power generation and power consumption.

3. The modular energy storage method according to claim 2, characterized in that, The step of splitting the spatiotemporal data matrix into time series of different nodes based on spatial dimensions, and performing variational mode decomposition on the time series of each node to obtain the intrinsic mode components of each node includes: Traverse the spatial dimension index of the spatiotemporal data matrix, using the spatial coordinates of each energy storage module as node identifiers, and extract the data vectors of each node at all timestamps from the spatiotemporal data matrix to form the time series of each node. The initial modal decomposition number K and the second-order penalty factor α are calculated based on the signal length and sampling frequency of the time series at each node; where K represents the total number of intrinsic modal components to be obtained by decomposition. The convergence objective is to ensure that the center frequencies of each intrinsic mode component no longer overlap during the iteration process. The alternating direction multiplier method is used to perform variational mode decomposition on the time series of each node, and the spectral overlap between any two intrinsic mode components is calculated after each iteration. In the first iteration, the spectral overlap is not compared, and the next iteration begins. When the spectral overlap is greater than the spectral overlap calculated in the previous iteration, the value of k is increased by 1, and the center frequency of the newly added intrinsic mode component is initialized to the midpoint value of the maximum interval between any two adjacent center frequencies according to the current center frequency distribution of each intrinsic mode component, and the iterative decomposition continues. When the rate of change of the center frequency of the modal components in two consecutive iterations is less than the preset rate of change of the center frequency, the decomposition stops and the intrinsic modal components of each node are obtained.

4. The modular energy storage method according to claim 1, characterized in that, The determination of the initial charging and discharging strategy for each energy storage module based on the predicted power generation and the predicted electricity demand includes: The predicted power generation is aligned with the predicted electricity demand based on time series, and the power difference sequence at each time segment is calculated. Based on the alternating positive and negative characteristics of the power difference sequence, the time interval of continuous power differences in the same direction is divided into a window to obtain the scheduling window; Based on the power limit for interaction between electricity price information and the main power grid, calculate the cumulative power deviation from the start time to the current time of the scheduling window; The power difference sequence is smoothed using a moving average filter to generate the power adjustment coefficient for the scheduling window; Based on the health status and the current state of charge of each energy storage module, the response priority of each energy storage module is calculated. Based on the response priority and the power adjustment coefficient, the charging and discharging power of each energy storage module is allocated to obtain the charging and discharging power and charging and discharging time period of each energy storage module as the initial charging and discharging strategy.

5. A modular energy storage method according to claim 4, characterized in that, The step of allocating charging and discharging power to each energy storage module based on the response priority and the power adjustment coefficient to obtain the charging and discharging power and charging and discharging time period of each energy storage module, as an initial charging and discharging strategy, includes: Calculate the net total charging and discharging power within the scheduling window based on the power adjustment coefficient and the power difference within the scheduling window; The net total charging and discharging power is initially allocated based on the response priority of each energy storage module to obtain the initial allocated power for each energy storage module. Based on the current state of charge, maximum charge / discharge power limit, and health status of each energy storage module, an adjustable range for each energy storage module is constructed. If the initial allocated power does not exceed the adjustable range of the corresponding energy storage module, then the initial allocated power shall be taken as the final allocated power of the energy storage module. If the initial allocated power exceeds the adjustable range of the corresponding energy storage module, the boundary value of the adjustable range of the energy storage module is taken as the final allocated power, and the power difference of the excess part is calculated. The total amount of power difference exceeding the limit is summed up to obtain the total amount of power to be redistributed. Based on the response priority of the remaining energy storage modules that are within the adjustable range, the total power to be redistributed is then allocated a second time. If the allocated power of all energy storage modules falls within the corresponding adjustable range within the preset maximum number of iterations, then the final allocated power of each energy storage module and the corresponding charging and discharging period are used as the initial charging and discharging strategy.

6. A modular energy storage method according to claim 5, characterized in that, The adjustable range of each energy storage module is constructed based on its current state of charge, maximum charge / discharge power limit, and health status, including: Based on the maximum charge and discharge power limits of each energy storage module and the aforementioned health status, a steady-state rigid boundary is constructed. A transient elastic boundary is constructed based on the deviation between the state of charge of each energy storage module and the preset standard state of charge, as well as the decay of the healthy state. Obtain the current operating mode of each energy storage module, including charging mode, discharging mode and standby mode; Based on the current operating mode, determine the enabling state of the limiting transient elastic boundary to obtain the target enabling state; Based on the target enable state, determine the elastic buffer; The steady-state rigid boundary is superimposed on the elastic buffer to obtain the adjustable range of each energy storage module; wherein, the power allocation within the elastic buffer requires triggering an additional security verification mechanism.

7. A modular energy storage method according to claim 6, characterized in that, Power allocation within the elastic buffer requires triggering additional security verification mechanisms, including: When the power allocated to any energy storage module falls into the elastic buffer within the adjustable range, the energy storage module is marked as an elastic call module. Obtain the current temperature, temperature rise rate, and health status decay rate of the elastic call module; Calculate the first safety factor based on the ratio of the current temperature to the preset temperature safety threshold; The second safety factor is calculated based on the ratio of the temperature rise rate to the preset maximum allowable temperature rise rate. The third safety factor is calculated based on the ratio of the health status decay rate to the preset decay rate alarm threshold. The first security factor, the second security factor, and the third security factor are weighted and summed to obtain the comprehensive security risk index of the elastic call module; If the comprehensive safety risk index is greater than the preset safety risk index threshold, the elastic power allocation of the elastic call module will be rejected, the final allocated power of the module will be forcibly limited to its steady-state rigid boundary value, and the power gap will be redistributed to other energy storage modules within the steady-state rigid boundary. If the comprehensive security risk index is less than or equal to the preset security risk index threshold, then the elastic call module is allowed to execute based on the allocated power in the elastic buffer.

8. A modular energy storage method according to claim 5, characterized in that, Also includes: If the preset maximum number of iterations is reached and there is still an unconverged power difference exceeding the adjustable range, then the product of the adjustable range margin and response priority of each energy storage module in the current iteration round is obtained to construct a relaxation factor vector. Based on the sign direction of the unconverged power difference, the charging / discharging dominant mode of the scheduling window is determined; When the scheduling window is in charging-dominated mode, the energy storage modules are sorted from high to low according to the upper limit of the adjustable range. The power difference of the excess portion is redistributed to the energy storage module with the highest upper limit of the adjustable range in decreasing relaxation step size until the power difference is zero or all modules reach their expansion limit. The expansion limit is the original upper limit of the adjustable range multiplied by a preset relaxation coefficient. When the scheduling window is in the discharge-dominant mode, the energy storage modules are sorted from low to high according to the lower limit of the adjustable range, and the excess power difference is allocated to the energy storage module with the lowest lower limit of the adjustable range in turn. The final allocated power of each energy storage module after relaxation is used as the initial charging and discharging strategy, and an alarm message containing an incomplete convergence flag is generated to trigger the power adjustment coefficient correction in the next scheduling window.

9. A modular energy storage method according to claim 4, characterized in that, When the probability of a fault occurrence is greater than a preset first fault occurrence probability threshold and less than a preset second fault occurrence probability threshold, the cooperative control scheme is modified based on the fault type to obtain a target cooperative control scheme, including: When the probability of the fault occurrence is greater than a preset first fault occurrence probability threshold and less than a preset second fault occurrence probability threshold, the extreme type of the current operating scenario is identified. The extreme type includes a sudden change in grid power and an extreme fluctuation in electricity price. Obtain the fault type output by the fault analysis model. When the fault type is an overcurrent fault or an overvoltage fault, match and correct the fault type with the extreme type to obtain the target cooperative control scheme, wherein: If the fault type is an overcurrent fault and the current scenario is a sudden change in grid power, then the power change rate in the cooperative control scheme is limited to within the preset maximum change rate, and the power waveform is subjected to low-pass filtering. If the fault type is an overvoltage fault and the current scenario is an extreme electricity price fluctuation scenario, then the voltage regulation coefficient is calculated based on the electricity price deviation, and the upper limit of the charging power of the corresponding energy storage module is reduced based on the voltage regulation coefficient. If the fault type is not directly related to the current scenario, the collaborative control scheme is modified based on the corresponding correction strategy in the preset fault type-correction table.

10. A modular energy storage system, characterized in that, include: The multi-source data acquisition module is used to collect operating data of power generation equipment, power grid and power consumption equipment in the smart industrial park, as well as the basic operating parameters and extended parameters of each energy storage module. The data feature analysis module is used to standardize and mine the operational data to extract the correlation features between power generation and power consumption. The basic operating parameters and extended parameters are fused together, and the health status of each energy storage module is determined. The source-load power prediction module is used to input the associated features into a preset power generation and consumption prediction model to obtain the predicted power generation and predicted electricity demand. The initial charge and discharge strategy module is used to determine the initial charge and discharge strategy of each energy storage module based on the predicted power generation and the predicted power demand. A collaborative scheme generation module is used to perform task allocation and power balancing on the initial charging and discharging strategy based on the health status of each energy storage module, and generate a collaborative control scheme between each energy storage module. The fault prediction and calculation module is used to input the collaborative control scheme and the health status into a preset fault analysis model to obtain the predicted fault types and fault occurrence probabilities that will occur when each energy storage module executes the collaborative control scheme. The tiered risk control and handling module is used to modify the collaborative control scheme based on the fault type when the probability of the fault occurrence is greater than a preset first fault occurrence probability threshold and less than a preset second fault occurrence probability threshold, to obtain a target collaborative control scheme, and to adjust the charging and discharging power of each energy storage module based on the target collaborative control scheme; when the probability of the fault occurrence is greater than the preset second fault occurrence probability threshold, a warning message is generated and output.

Images