A power system dynamic adaptive battery energy storage energy management system

Abstract

The application discloses a power system dynamic self-adaptive battery energy storage energy management system and relates to the technical field of battery energy storage management. The system realizes fine capacity configuration by introducing a dynamic prediction and shortest path optimization algorithm, effectively improves the response capability of the system to extreme working conditions in combination with a multi-scene driven adaptive scheduling strategy, extracts strategy execution lag rate characteristics by using a feedback mechanism, evaluates operation deviation in real time and triggers rolling optimization, realizes the collaborative closed loop of the energy storage system from static planning to dynamic control, and significantly enhances the flexibility, stability and economy of power grid operation.

Description

Technical Field

[0001] This invention relates to the field of battery energy storage management technology, and more specifically to a dynamic adaptive battery energy storage management system for power systems. Background Technology

[0002] Dynamic adaptive battery energy storage management for power systems refers to the process of dynamically adjusting the charging and discharging strategies of energy storage devices based on real-time load demand, grid conditions, and fluctuations in renewable energy during power system operation. This achieves efficient management and optimized allocation of electrical energy. This management method possesses adaptive capabilities, automatically optimizing control strategies according to changes in the external environment and system operating conditions, improving grid stability, reliability, and energy utilization efficiency. It is particularly suitable for modern power systems with a large amount of renewable energy integrated into them.

[0003] The existing technology has the following shortcomings:

[0004] In the dynamic planning method for active distribution network energy storage systems, the upper-level planning model mainly formulates the installed capacity strategy based on static annual average or typical daily data, while the lower-level operation strategy needs to cope with real-time, variable, and even extreme grid operating environments. However, when unforeseen events such as sudden weather, drastic fluctuations in user load, or equipment failures occur, there may be a significant deviation between the lower-level scheduling strategy and the upper-level planning objectives, leading to energy storage system failure, increased peak-to-valley load differences, and even voltage or frequency instability, seriously affecting the system's economy and reliability. Summary of the Invention

[0005] The purpose of this invention is to provide a dynamic adaptive battery energy storage management system for power systems to address the shortcomings of the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a dynamic adaptive battery energy storage management system for power systems, comprising a data acquisition module, a capacity planning module, a scenario generation module, a scheduling control module, and a feedback coupling module;

[0007] The data acquisition module is used to acquire power system topology information, historical load data, distributed energy output data, electricity price information, and weather forecast data.

[0008] The capacity planning module uses a dynamic prediction method based on the acquired data, combined with the shortest path algorithm, to optimize the configuration of the installed capacity of energy storage systems and distributed photovoltaics.

[0009] The scenario generation module is used to construct several running scenarios with different running states, and to train the scheduling strategy model based on the running scenario group;

[0010] The scheduling and control module, coupled with the capacity planning module, is used to optimize the charging and discharging behavior of the energy storage system in various operating scenarios in real time, thereby reducing the load peak-valley difference.

[0011] The feedback coupling module is used to collect the system's operating status and strategy execution results in real time, and feed back the severity of the operating deviation to the capacity planning module to achieve rolling optimization and dynamic adjustment.

[0012] Preferably, the capacity planning module uses a time series prediction algorithm to predict the load and distributed energy output for the next 7 to 30 days, constructs a multi-scenario input dataset, uses the shortest path algorithm to evaluate the energy transmission loss of each node in the power distribution network, and prioritizes the allocation of energy storage and photovoltaic capacity for nodes with short energy loss paths.

[0013] Preferably, the shortest path algorithm is used to evaluate the energy transmission loss of each node in the distribution network, which includes: modeling the distribution network as a directed graph structure, with nodes as buses and edges as lines; calculating the power flow path length or impedance loss for each node; using the shortest path algorithm to find the path with the minimum energy loss for different installation locations; using the path loss as part of the objective function to guide the system to prioritize the configuration of nodes with shorter loss paths, and forming a preliminary configuration suggestion for photovoltaic and energy storage deployment that is simultaneously optimal in terms of geographical and electrical location.

[0014] Preferably, the scheduling control module uses a model predictive control algorithm to optimize the dynamic charging and discharging control of the energy storage system under different time scales and operating conditions, specifically:

[0015] Before the start of each control cycle, the current grid operating status is collected, including load level, energy storage system state of charge, photovoltaic output, electricity price, and ambient temperature data. Using the short-term future prediction results provided by the scenario generation module, the trend of the grid operating status in the subsequent several time periods is predicted. A model predictive control optimization model is constructed, with the goal of minimizing electricity purchase cost, load peak-valley difference, and SOC offset. A comprehensive objective function containing multiple weighting factors is established. The control variables are the charging power and discharging power of the energy storage system in each time step. The constraints include the power range of the energy storage system, upper and lower limits of SOC, power supply and demand balance, and charging and discharging mutual exclusion constraints. The solver is used to perform real-time calculations on the optimization model to generate the optimal charging and discharging strategy for each time period within the future prediction time window. The control time window is rolled forward by one step, and the grid operating status and prediction data are collected again, and the optimization solution process is repeated to form a rolling optimization control mechanism.

[0016] Preferably, the feedback coupling module is used to collect the system operating status and strategy execution results in real time, including: extracting the frequent control reversal rate feature from the real-time collected system operating status data. The extraction method is: in each rolling control cycle, record the number of control direction changes and divide by the total number of time steps to obtain the frequent control reversal rate, expressed as: ; sx is the number of control direction changes, TQ is the total number of time steps, and CRF is the frequent control reversal rate.

[0017] Preferably, the strategy execution lag rate feature is extracted from the real-time collected strategy execution result data. The calculation method is as follows: Let the time series be discrete time points t=1,2,...,T, the control command sequence be u(t), and the actual response sequence be y(t). Set the sliding window width w, and for each time point t, search backward within the window to determine when the response sequence y(t+τ) first reaches the value of the command u(t), and set the lag step size. The average lag step size is obtained by averaging all lag step sizes; the formula for calculating the policy execution lag rate is: Where K is the indicator function, and is the defined hysteresis threshold step size. The strategy execution lag rate is represented by T, where T is the total number of time points.

[0018] Preferably, the frequent control reversal rate feature and the strategy execution lag rate feature are normalized and transformed to express the deviation intensity between 0 and 1, and the deviation severity analysis value is constructed by combining the weighting factors α and β.

[0019] Preferably, the obtained deviation severity analysis value is compared with the gradient standard threshold, which includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold. The deviation severity analysis value is compared with the first standard threshold and the second standard threshold respectively.

[0020] If the severity analysis value of the deviation is greater than the second standard threshold, it is a serious deviation, and the capacity planning module will be immediately restarted for rolling optimization; if the severity analysis value of the deviation is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it is a minor deviation, and the capacity configuration will be periodically corrected; if the severity analysis value of the deviation is less than the first standard threshold, the operation is normal and no adjustment is required.

[0021] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0022] 1. This invention constructs a closed-loop, self-adjusting energy management system by integrating five modules: data acquisition, capacity planning, scenario generation, scheduling control, and feedback coupling. The system performs dynamic prediction based on multi-source data and employs a shortest path algorithm to evaluate the energy consumption path of the distribution network, achieving efficient and optimized allocation of energy storage and photovoltaic installed capacity. By constructing multiple operating scenarios and training scheduling strategy models, the scheduling control module can achieve dynamic charging and discharging optimization of the energy storage system under different operating conditions, balancing grid economy and operational stability. Simultaneously, rolling optimization is achieved using model predictive control algorithms to ensure the timeliness and adaptability of the strategy.

[0023] 2. This invention further extracts two deep operational features—frequent control reversal rate and strategy execution lag rate—to comprehensively analyze the degree of strategy execution deviation. Using the severity of the deviation as a criterion, it triggers periodic adjustments or restarts to optimize capacity configuration, achieving dynamic linkage from system planning to real-time control. This system effectively overcomes the problems of disconnect between upper and lower level strategies, response lag, and insufficient ability to cope with extreme operating conditions in traditional systems, demonstrating significant technical advantages in improving the flexibility, intelligence, and sustainability of power systems. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0025] Figure 1 This is a system module diagram of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] For examples, please refer to Figure 1 As shown in this embodiment, a dynamic adaptive battery energy storage management system for power systems includes a data acquisition module, a capacity planning module, a scenario generation module, a scheduling and control module, and a feedback coupling module.

[0028] The data acquisition module is used to acquire power system topology information, historical load data, distributed energy output data, electricity price information, and weather forecast data.

[0029] The capacity planning module uses a dynamic prediction method based on the acquired data, combined with the shortest path algorithm, to optimize the configuration of the installed capacity of energy storage systems and distributed photovoltaics.

[0030] The scenario generation module is used to construct several running scenarios with different running states, and to train the scheduling strategy model based on the running scenario group;

[0031] The scheduling and control module, coupled with the capacity planning module, is used to optimize the charging and discharging behavior of the energy storage system in various operating scenarios in real time, thereby reducing the load peak-valley difference.

[0032] The feedback coupling module is used to collect the system's operating status and strategy execution results in real time, and feed back the severity of the operating deviation to the capacity planning module to achieve rolling optimization and dynamic adjustment.

[0033] The data acquisition module is a fundamental component of the dynamic adaptive battery energy storage management system for power systems, and is responsible for acquiring, cleaning, preprocessing and standardizing various types of data required for system operation.

[0034] The power system topology information collected includes: node information (such as bus number, load node, generation node, energy storage node), line connection relationship (topology matrix), transformer parameters, switch status, power flow direction, and real-time topology status (such as fault status, isolation status).

[0035] Data acquisition method: Access the dispatching system or power distribution automation system (such as SCADA, EMS); use the IEC 61850 or DL / T 860 communication standard protocol to acquire structured data.

[0036] Historical load data collection includes: historical load curves by node or region (usually at a granularity of 15 minutes / hour), daily load curves, seasonal fluctuation characteristics, records of abnormal electricity consumption, and load characteristics on special dates (holidays, high-temperature days, etc.).

[0037] Data collection method: Access the electricity information collection system (AMI) or the distribution network load monitoring platform; collect smart meter data and classify the user-side data (residential, industrial, commercial).

[0038] Data collection content for distributed energy output includes: actual output power of distributed generation units such as photovoltaic and wind power, power generation curves (collected by time period), and parameters such as photovoltaic panel module efficiency, azimuth angle, and tilt angle (used for modeling).

[0039] Historical and real-time output comparison values ​​(for deviation analysis) acquisition method: Collect distributed energy data through inverter interface or energy gateway, connect to local energy management system (EMS) or microgrid controller, and use communication protocols such as Modbus, OPC UA to interact with equipment.

[0040] Electricity price information collected includes: real-time electricity price, time-of-use electricity price, peak-valley electricity price, regulation service price, day-ahead market price, electricity spot price, and ancillary service market price.

[0041] Data acquisition method: Access the power grid dispatching system, power trading platform API, or obtain future electricity price trends as input parameters through simulation algorithms.

[0042] Meteorological forecast data collection includes: real-time meteorological data such as solar radiation intensity, temperature, wind speed, and cloud cover; hourly meteorological forecast data for the next 24 hours to 7 days; and historical meteorological data used for modeling (such as the analysis of the impact of weather on load / photovoltaic output).

[0043] Data collection method: Access to commercial meteorological service platforms, satellite data or local environmental sensors to obtain refined meteorological information (such as local cloud dynamics at photovoltaic sites).

[0044] The core tasks of data cleaning and standardization are: timestamp alignment, data interpolation completion, outlier removal, and normalization. Multi-source heterogeneous data formats are uniformly converted into system-recognizable standard formats (such as CSV, JSON, TSDB). Data cleaning algorithms (sliding window, Z-score anomaly detection, KNN interpolation, etc.) are used. It can be combined with databases (such as InfluxDB, TimescaleDB) or real-time data buses (such as Kafka, MQTT).

[0045] Multi-source data from the data acquisition module is integrated to construct an input matrix for modeling. Electricity and power forecasting modeling enables medium- to long-term predictions of future electricity load and renewable energy output, providing a dynamic basis for capacity optimization.

[0046] Load forecasting: Use time series forecasting algorithms (such as ARIMA, LSTM, GRU) to forecast user-side load at 7-day, 30-day, and seasonal levels;

[0047] Photovoltaic output prediction: Based on historical radiation data, module characteristics and weather forecasts, use radiation-output regression models or neural network-based PV prediction models;

[0048] The prediction results are used to construct multi-scenario output sequences according to three types of fluctuation amplitude: high, medium and low, in order to improve the robustness of the system.

[0049] Establish a mathematical model for the capacity optimization problem, clarify the constraints and optimization objectives, and define the modeling elements:

[0050] Decision variables: Energy storage system capacity (kWh), maximum charge / discharge power (kW); Photovoltaic system installed capacity (kW);

[0051] Objective function: Minimize the total annual cost of the system, including investment, operation and maintenance, electricity purchase cost, and energy loss;

[0052] Constraints: Node voltage constraint (±10%), system energy balance constraint (generation-consumption+storage=0), energy storage system SOC constraint (cannot be overcharged / over-discharged), investment budget constraint (total system investment must not exceed the budget limit), and physical limitations on node installed capacity (roof area, load level, etc.).

[0053] Network loss assessment using shortest path algorithms:

[0054] The distribution network is modeled as a directed graph structure (nodes are buses, and edges are lines).

[0055] Calculate the power flow path length or impedance loss for each node;

[0056] Use shortest path algorithms (such as Dijkstra or Floyd) to find the path with minimum energy loss for different installation locations;

[0057] By incorporating path loss as part of the objective function (or adding weights), the system is guided to prioritize the allocation of nodes with "shorter loss paths".

[0058] Preliminary recommendations for optimal photovoltaic and energy storage deployment locations based on both geographical and electrical factors are proposed.

[0059] Construct a comprehensive objective function that considers multiple factors to achieve a multi-objective trade-off:

[0060] Cost model: Annual fixed investment discounted cost (CAPEX), annual operation and maintenance cost (OPEX), electricity purchase cost (including the impact of time-of-use pricing), and transmission loss cost.

[0061] Efficiency model: Energy saving and emission reduction conversion (such as calculating carbon emission reduction benefits based on carbon emission factors), saving peak electricity purchase costs.

[0062] Utilize intelligent optimization algorithms to obtain the optimal installed capacity configuration. Select the optimization algorithm: if the model is linearly differentiable, use linear programming or mixed-integer linear programming (MILP); if it is nonlinear, use heuristic algorithms such as genetic algorithm (GA), particle swarm optimization (PSO), differential evolution (DE), etc.

[0063] Multi-scenario analysis: Independently optimize and summarize statistics under different prediction scenarios (such as sunny days, peak days, extreme weather); extract robust solutions (the median solution that satisfies most scenarios) or adopt multi-objective robust optimization strategies;

[0064] Results output: node-level energy storage capacity, photovoltaic capacity; return on investment cycle analysis (ROI, LCOE); baseline comparison of overall system cost savings and carbon emissions.

[0065] The optimization results are passed to the lower-level scheduling module, simulation module, and user interface. A configuration table (including the capacity of each node and device parameters) is generated; a model report is output (including prediction model error analysis, sensitivity analysis, and scenario comparison analysis); and a data interface (API or database) is provided for the real-time scheduling system to call capacity boundary conditions.

[0066] Key variables such as load, photovoltaic output, electricity price, and weather are extracted from long-term operational data to serve as the basis for scenario generation. Collected variables include: Load: daily curve, peak time, load fluctuation amplitude; Renewable Energy: daily output curve of photovoltaic / wind power; Electricity Price: time-of-use electricity price trend; Weather: temperature, radiation, wind speed, etc.

[0067] Data distribution analysis to extract features (such as average value, volatility, peak-to-valley ratio); clustering by day / hour (such as K-means or DBSCAN) to extract typical operating patterns (such as "sunny days with high load" and "cloudy days with low load weekends").

[0068] Construct a set of representative operating scenarios covering possible future operating states. Typical scenario generation (based on cluster centers): Each cluster result is considered a typical scenario, representing a typical combination of load, output, electricity price, and weather conditions; Random disturbance scenario generation (considering uncertainty): Based on the typical scenarios, disturbance variables (such as ±10% PV error, ±15% load deviation) are introduced; large-sample simulation scenarios are generated using Monte Carlo methods or Latin hypercube sampling (LHS); each disturbance scenario has a probability weight.

[0069] Extreme scenario construction (extreme value theory): Select historical extreme load days and meteorological disaster days; generate extreme combination scenarios of "peak load + low photovoltaic + high electricity price"; used to test the scheduling robustness of the system under emergency conditions.

[0070] Scene compression algorithms (such as K-means merging and K-medoids clustering) are used to reduce redundant scenes; probability weights are assigned to each scene (based on historical frequency, simulated probability, etc.); several scenes (such as 10 to 50) and their parameter combinations and weights are output for training and scheduling simulation.

[0071] Each generated scenario is used as a set of input samples, including: initial SOC, electricity price curve, photovoltaic forecast, load forecast, and system state; the following algorithms can be used to learn the scheduling strategy: linear / nonlinear optimization algorithms (for rule-based schedulers); reinforcement learning (such as DQN, PPO); model predictive control (MPC); after each round of training, the performance indicators (cost, stability, response time) of the strategy in all scenarios are recorded and used for scheduler self-optimization.

[0072] The scheduling and control module is based on the model predictive control (MPC) algorithm. It combines the installed capacity parameters provided by the capacity planning module and the operating scenarios output by the scenario generation module to achieve dynamic charging and discharging optimization control of the energy storage system under different time scales and operating conditions.

[0073] Before the start of each control cycle, the system first collects data on the current grid operating status, including load level, state of charge (SOC) of the energy storage system, photovoltaic output, electricity price, and ambient temperature. Simultaneously, using short-term future predictions provided by the scenario generation module, it forecasts the trends of these variables over the next several time periods. The prediction period is typically 12 to 24 hours, and the control granularity can be set from 15 minutes to 1 hour.

[0074] Based on the above inputs, the system constructs a model predictive control optimization model. This model aims to minimize electricity purchase cost, load peak-to-valley difference, and SOC offset, establishing a comprehensive objective function with multiple weighting factors.

[0075] The control variables are the charging power and discharging power of the energy storage system at each time step. Constraints include the power range of the energy storage system, the upper and lower limits of SOC, power supply and demand balance, and charging and discharging mutual exclusion constraints, to ensure that the control strategy is feasible within the physical and operational range.

[0076] The system uses a solver to perform real-time calculations on the above optimization model, generating the optimal charging and discharging strategy for each time period within the future prediction time window. However, in actual control, the system only executes the first set of control actions at the current time step to ensure that the strategy is always based on the latest prediction data.

[0077] Subsequently, the control time window is rolled forward by one step, the state and prediction data are re-acquired, and the optimization solution process is repeated to form a rolling optimization control mechanism.

[0078] Through the aforementioned control strategies, the dispatch control module can effectively reduce peak loads, increase photovoltaic self-consumption rate, reduce dependence on the grid, and optimize overall electricity costs while ensuring stable system operation. Compared with traditional static dispatch methods, this scheme has significant advantages such as real-time optimization, continuous rolling updates, intelligent response to deviations, and adaptive parameter adjustment, and is suitable for energy storage dispatch control in various types of active distribution networks, microgrids, and smart energy systems.

[0079] The feedback coupling module is used to collect system operating status and strategy execution results in real time, and feeds back the severity of operating deviations to the capacity planning module, including:

[0080] The frequent control reversal rate feature is extracted from real-time system operation status data. This feature represents the frequency of "charge-discharge-charge" or "discharge-charge-discharge" behavior switching within a short period. It reflects potential over-control issues in the current control strategy when dealing with uncertainty, which can lead to reduced lifespan and decreased energy efficiency of energy storage equipment. The extraction method is as follows: in each rolling control cycle, record the number of control direction changes and divide by the total number of time steps to obtain the frequent control reversal rate, expressed as: ; sx represents the number of control direction changes, TQ represents the total number of time steps, and CRF represents the frequent control reversal rate. Threshold setting example: Normal: <10%, Abnormal: 10% ~ 20%, Severe: >20%.

[0081] Extracting policy execution lag rate characteristics from real-time collected policy execution result data, that is, calculating the average delay between scheduling instructions and actual responses, in units of time steps or percentages; can reflect the physical lag or communication delay problems of the system when responding to sudden events or policy changes.

[0082] The calculation method is as follows: Let the time series be discrete time points t=1,2,...,T, control command sequence: u(t), actual response sequence: y(t); set the sliding window width w, such as w=3~5 time steps.

[0083] For each time point t, search backward within the window to determine when the response sequence y(t+τ) first reaches or approaches the value of the instruction u(t). Lag step size. Defined as: ; For a moment The actual response power value of the device, ϵ is the allowable response error tolerance (e.g., within 1kW). If τ exceeds the window length, it is considered a timeout response or no response. The average lag step size is obtained by averaging all lag step sizes; the formula for calculating the strategy execution lag rate is: Where K is an indicator function, The defined hysteresis threshold step size, The strategy execution lag rate is represented by T, where T is the total number of time points.

[0084] The frequent control reversal rate and strategy execution lag rate features are normalized to represent the bias intensity between 0 and 1. A bias severity analysis value D is then constructed by combining weighting factors α and β, with the expression: Where α=0.6 (strategy lag has a greater impact on system stability) β=0.4.

[0085] The obtained deviation severity analysis value is compared with the gradient standard threshold, which includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold. The deviation severity analysis value is compared with the first standard threshold and the second standard threshold respectively.

[0086] If the severity analysis value of the deviation is greater than the second standard threshold, it is considered a serious deviation and the capacity planning module will be immediately restarted for rolling optimization.

[0087] If the severity analysis value of the deviation is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it is considered a minor deviation, and it is recommended to periodically adjust the capacity configuration.

[0088] If the severity of the deviation analysis value is less than the first standard threshold, the system is operating normally and no adjustment is required.

[0089] Coupling mechanism with the capacity planning module: Feedback data is transmitted to the capacity planning engine in JSON format or via a database interface; this includes fields such as current system node, timestamp, SELR, CRF, deviation level, and running labels (e.g., "over-control" or "hysteresis"). Minor deviations require periodic correction (e.g., weekly re-optimization); severe deviations require immediate restart of the upper-level optimization model.

[0090] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0091] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0092] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A dynamic adaptive battery energy storage management system for power systems, characterized in that: It includes a data acquisition module, a capacity planning module, a scenario generation module, a scheduling and control module, and a feedback coupling module; The data acquisition module is used to acquire power system topology information, historical load data, distributed energy output data, electricity price information, and weather forecast data. The capacity planning module uses a dynamic prediction method based on the acquired data, combined with the shortest path algorithm, to optimize the configuration of the installed capacity of energy storage systems and distributed photovoltaics. The capacity planning module uses a time series prediction algorithm to predict the load and distributed energy output for the next 7 to 30 days, constructs a multi-scenario input dataset, uses the shortest path algorithm to evaluate the energy transmission loss of each node in the distribution network, and prioritizes the energy storage and photovoltaic capacity of nodes with short energy loss paths. The shortest path algorithm is used to evaluate the energy transmission loss of each node in the distribution network. This includes: modeling the distribution network as a directed graph structure, with nodes as buses and edges as lines; calculating the power flow path length or impedance loss for each node; using the shortest path algorithm to find the path with the minimum energy loss for different installation locations; and using the path loss as part of the objective function to guide the system to prioritize the configuration of nodes with shorter loss paths, and to form a preliminary configuration suggestion for photovoltaic and energy storage deployment that is simultaneously optimal in terms of geographical and electrical location. The scenario generation module is used to construct several running scenarios with different running states, and to train the scheduling strategy model based on the running scenario group; The scheduling and control module, coupled with the capacity planning module, is used to optimize the charging and discharging behavior of the energy storage system in various operating scenarios in real time, thereby reducing the load peak-valley difference. The feedback coupling module is used to collect the system's operating status and strategy execution results in real time, and feed back the severity of the operating deviation to the capacity planning module to achieve rolling optimization and dynamic adjustment.

2. The power system dynamic adaptive battery energy storage management system according to claim 1, characterized in that: The scheduling and control module, based on model predictive control algorithms, optimizes the dynamic charging and discharging control of the energy storage system under different time scales and operating conditions. Specifically: Before the start of each control cycle, the current grid operating status is collected, including load level, energy storage system state of charge, photovoltaic output, electricity price and ambient temperature data. Using the short-term future prediction results provided by the scenario generation module, the trend of the grid operating status in the subsequent several periods is predicted. A predictive control optimization model is constructed with the objectives of minimizing electricity purchase cost, load peak-valley difference, and SOC offset. A comprehensive objective function with multiple weighting factors is established. The control variables are the charging power and discharging power of the energy storage system in each time step. The constraints include the power range of the energy storage system, upper and lower limits of SOC, power supply and demand balance, and charging and discharging mutual exclusion constraints. The solver is used to perform real-time calculations on the optimization model to generate the optimal charging and discharging strategy for each time period within the future prediction time window. The control time window is rolled forward by one step, and the grid operating status and prediction data are re-collected and the optimization solution process is repeated to form a rolling optimization control mechanism.

3. The power system dynamic adaptive battery energy storage management system according to claim 2, characterized in that: The feedback coupling module is used to collect system operating status and strategy execution results in real time. This includes extracting frequent control reversal rate features from the real-time system operating status data. The extraction method is as follows: in each rolling control cycle, record the number of control direction changes and divide by the total number of time steps to obtain the frequent control reversal rate, expressed as: ; sx is the number of control direction changes, TQ is the total number of time steps, and CRF is the frequent control reversal rate.

4. The power system dynamic adaptive battery energy storage management system according to claim 3, characterized in that: The strategy execution lag rate feature is extracted from the real-time collected strategy execution result data. The calculation method is as follows: Let the time series be discrete time points t=1,2,...,T, the control command sequence be u(t), and the actual response sequence be y(t). Set the sliding window width w, and for each time point t, search backward within the window to determine when the response sequence y(t+τ) first reaches the value of the command u(t), and set the lag step size. The average lag step size is obtained by averaging all lag step sizes; the formula for calculating the policy execution lag rate is: Where K is an indicator function, The defined hysteresis threshold step size, The strategy execution lag rate is represented by T, where T is the total number of time points.

5. A dynamic adaptive battery energy storage management system for power systems according to claim 4, characterized in that: The frequent control reversal rate and strategy execution lag rate characteristics are normalized and transformed to express the bias intensity between 0 and 1. The bias severity analysis value is constructed by combining the weighting factors α and β.

6. The power system dynamic adaptive battery energy storage management system according to claim 5, characterized in that: The obtained deviation severity analysis value is compared with the gradient standard threshold, which includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold. The deviation severity analysis value is compared with the first standard threshold and the second standard threshold respectively. If the severity analysis value of the deviation is greater than the second standard threshold, it is considered a serious deviation and the capacity planning module will be immediately restarted for rolling optimization. If the severity analysis value of the deviation is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it is considered a minor deviation, and periodic adjustments to the capacity configuration are performed. If the severity of the deviation analysis value is less than the first standard threshold, the system is operating normally and no adjustment is required.

Images