Methods, systems, and computer-readable media for constructing digital control models for energy storage power station clusters
By constructing a five-level progressive digital model, the problem of poor dynamic adaptability in energy storage regulation modeling technology is solved, realizing the high reliability of energy storage resources to be transformed into grid support resources, improving the power supply accuracy of the grid and the resource support capability for emergency conditions, and possessing high precision and rapid response capabilities.
Patent Information
- Application Number
- CN202611244416.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-17
- Publication Date
- 2026-11-17
AI Technical Summary
Existing energy storage regulation modeling technologies cannot adapt to the regulation needs of large-scale energy storage clusters and complex grid operating conditions. They suffer from poor dynamic adaptability, insufficient regulation accuracy, and an imbalance between safety and economy, making it difficult to transform energy storage resources into observable, measurable, and controllable grid support resources.
A progressively advanced five-level digital model is constructed, including a basic rated operating condition model, a historical operation statistics credibility model, a real-time status perception model, an artificial intelligence fault prediction and adaptive optimization model, and a three-dimensional game control model. Machine learning and multi-objective game optimization mechanisms are integrated to achieve full-dimensional digital modeling.
It realizes the high reliability of energy storage resources to be transformed into grid support resources, improves the power supply accuracy of the grid, the reliability of emergency operating conditions resource support, and the economic efficiency of power spot operating conditions. It solves the problems of single level and poor dynamic adaptability in existing technologies, and has high precision, fast response and full-scenario adaptability capabilities.
Smart Images

Figure CN123121748A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of energy storage regulation technology, specifically a method, system, and computer-readable medium for constructing a digital regulation model for energy storage power station clusters. Background Technology
[0002] With the accelerated construction of new power systems, large-scale energy storage power stations with capacities of tens of millions of kilowatts have been centrally connected to the grid, becoming a core flexible resource for grid peak shaving, frequency regulation, voltage regulation, and emergency backup. Current energy storage control modeling technology has significant shortcomings, making it difficult to adapt to the control requirements of large-scale energy storage clusters and the complex operating conditions of the power grid. First, most existing energy storage models are static models based on a single rated operating condition, relying solely on the rated parameters of the equipment for control calculations, without taking into account historical operating patterns, the state of health (SOH) of the equipment's batteries, real-time operating conditions, and grid constraints, resulting in poor dynamic adaptability of the models. Second, the lack of a hierarchical and progressive multi-level modeling system makes it impossible to achieve full-dimensional coverage of energy storage resources from basic parameters to intelligent prediction and safe and economic coordinated regulation, and the hierarchical and regional regulation accuracy of multi-million kilowatt clusters is insufficient. Third, conventional models can only realize the basic regulation function of energy storage resources, without integrating multi-dimensional game mechanisms such as fault prediction, grid security assessment, and economic regulation of electricity spot market. Under extreme conditions such as grid emergency faults, sudden load changes, and large fluctuations in new energy sources, the regulation response delay is generally greater than 1 second, resulting in problems such as unobservable and unmeasurable regulation resources, insufficient regulation accuracy, and an imbalance between safety and economy. Fourth, existing modeling methods cannot transform passive, flexible, and adjustable energy storage resources into active, controllable, reliable, and precise grid support resources, which seriously restricts the improvement of grid power supply reliability, resource utilization efficiency, and emergency response capabilities.
[0003] Existing technologies for energy storage digital modeling often focus on single-device mechanism modeling, simple data fitting, or local state monitoring. These models suffer from limited dimensions, ambiguous hierarchies, and fragmented functions, failing to form a five-level progressive closed-loop modeling system covering static rated parameters, historical operating statistics, real-time state perception, intelligent fault prediction, and safety-economic game-theoretic control. This approach cannot address the challenges of controlling large-scale, clustered, multi-condition, and multi-constraint energy storage power station clusters with tens of millions of kilowatts of capacity, and struggles to simultaneously meet the multiple demands of grid safety and stability, rapid emergency support, and economic operation in the electricity spot market. Therefore, there is an urgent need to develop a five-level progressive digital model construction method for large-scale energy storage power station control that is hierarchically clear, comprehensive in dimensions, dynamically iterative, and combines safety and economic considerations. Summary of the Invention
[0004] In view of this, the present invention addresses the shortcomings of the prior art by providing a method, system, and computer-readable medium for constructing a digital control model for energy storage power station clusters.
[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for constructing a digital control model for energy storage power station clusters, characterized by comprising the following steps: Collect the rated operating parameters of each energy storage unit in the energy storage power station group and construct a primary basic rated operating condition model; Collect historical operation data of the energy storage power station group throughout its entire lifecycle. Using daily, weekly, monthly, quarterly, and annual statistical periods, statistically analyze the six-dimensional extreme value data of the operation indicators of each energy storage unit to evaluate the reliability of extreme value operation. Statistically analyze the effective regulation times of each energy storage unit participating in grid regulation, calculate the average regulation value index, and evaluate the reliability of routine regulation. Evaluate the reliability of the fit between the operating parameters of each energy storage unit and the rated operating condition parameters. Construct a two-level historical operation statistical reliability model using a discrete reliability evaluation mechanism. Based on the rated operating parameters of the first-level basic rated operating condition model and the high-reliability historical operating data of the second-level historical operating statistical reliability model, the on-site monitoring data of the energy storage power station group and the power grid operation data are accessed in real time. Battery health status parameters, equipment operating status, and battery real-time state of charge are collected simultaneously. Combined with the real-time load of the power grid, the output of new energy sources, the voltage of the access bus, and the power flow data of the line, the available adjustable capacity, power adjustment range and response capability of each energy storage unit are updated in real time to construct a third-level real-time state perception model. Based on the full-dimensional real-time data of the three-level real-time state perception model, and by integrating machine learning, time series prediction, and anomaly recognition artificial intelligence algorithms, a four-level artificial intelligence fault prediction and adaptive optimization model is constructed. Based on the aforementioned Level 1 basic rated operating condition model, Level 2 historical operation statistical credibility model, Level 3 real-time status perception model, and Level 4 artificial intelligence fault prediction and adaptive optimization model, and integrating three control modes—grid safety assessment, grid emergency safety regulation, and electricity spot market economic regulation—a five-level three-dimensional game control model of grid safety, emergency regulation, and spot market economic regulation is constructed using a multi-objective game optimization mechanism to conduct full-scenario optimization and control of energy storage power station clusters.
[0006] Furthermore, the rated operating parameters include rated capacity, rated charge / discharge power, rated charge / discharge voltage, rated operating current, rated operating efficiency, rated response delay, and maximum allowable charge / discharge rate; the operating indicators include charge / discharge power, adjustment range, response speed, and continuous operating time; the six-dimensional extreme value data include the maximum value, the second largest value, the second second largest value, the minimum value, the second smallest value, and the second second smallest value.
[0007] Furthermore, the construction of the primary basic rated operating condition model includes: Construct the rated parameter matrix of the energy storage unit, and define the rated operating condition parameter vector of energy storage unit i:
[0008] In the formula: The rated charge and discharge power of unit i; Rated capacity; This is the rated charge / discharge voltage; Rated operating current; Rated operating efficiency; This is the rated response delay; This refers to the rated maximum permissible charge / discharge rate. The rated baseline matrix for energy storage clusters is as follows: .
[0009] Furthermore, a two-level historical operational statistical credibility model is constructed using a discrete credibility evaluation mechanism, including: Constructing scoring functions for three types of discrete evaluation indicators These correspond to the reliability of extreme value operation, the reliability of normal adjustment, and the reliability of operating condition adaptation, respectively. The overall reliability score is as follows:
[0010] in accordance with The confidence level is divided into three levels based on the threshold: high confidence (8≤Q≤10), medium confidence (4≤Q<8), and low confidence (0≤Q<4).
[0011] Furthermore, a three-level real-time state awareness model is constructed, including: Battery health state quantification model
[0012] In the formula: This represents the current actual available capacity. For rated capacity, Characterizes the degree of battery aging and degradation; Real-time effective adjustable power calculation model
[0013] In the formula: The unit i has adjustable power available in real time; The constraint coefficient is adjusted effectively for SOC.
[0014] Furthermore, a four-level artificial intelligence fault prediction and adaptive optimization model is constructed, including: Using a long short-term memory neural network artificial intelligence algorithm, a fault prediction model is trained with battery health state degradation trend, battery state of charge fluctuation law, equipment operating temperature, electrical parameter deviation, and historical fault sample data as training datasets. The model outputs fault probability, fault impact range, fault prediction time and adjustable resource loss after fault. An adaptive optimization and control module is built to correct the energy storage control strategy in advance based on the fault prediction results. High-risk energy storage units are operated at reduced capacity and controlled with fault tolerance in advance, while healthy and high-quality energy storage units are given priority in the allocation of control tasks.
[0015] Furthermore, a five-level three-dimensional game-theoretic control model is constructed, including: Construct a multi-objective game objective function that balances the optimality of network infrastructure security, emergency support, and spot market economics. The overall objective function is:
[0016] Constraints:
[0017] In the formula: The target for safe loss of the space frame; The target is to compensate for the emergency power deficit; The economic benefit target for electricity spot trading; These are dynamic weighting coefficients; Weighted dynamic game logic: Power grid emergency conditions: Prioritize ensuring the safety of emergency power supply; Standard spot market operating conditions: Balanced assignment maximizes economic benefits under security constraints.
[0018] This invention also provides a system for constructing a digital control model for energy storage power station clusters, comprising: The first-level basic rated operating condition model construction module is configured to collect the rated operating condition parameters of each energy storage unit in the energy storage power station group and construct the first-level basic rated operating condition model. The module for constructing a secondary historical operation statistical credibility model is configured to collect full-cycle historical operation data of energy storage power station groups, with daily, weekly, monthly, quarterly, and annual statistical periods. It calculates the six-dimensional extreme value data of the operation indicators of each energy storage unit and evaluates the credibility of extreme value operation; it also calculates the effective regulation times of each energy storage unit participating in grid regulation, calculates the average regulation value index, and evaluates the credibility of routine regulation; it evaluates the compatibility credibility of the operating parameters of each energy storage unit with the rated operating condition parameters; and it constructs a secondary historical operation statistical credibility model using a discrete credibility evaluation mechanism. The three-level real-time state perception model construction module is configured to rely on the rated operating parameters of the first-level basic rated operating condition model and the high-reliability historical operating data of the second-level historical operating statistical reliability model, and to access the on-site monitoring data of the energy storage power station group and the power grid operation data in real time. It synchronously collects battery health status parameters, equipment operating status, and battery real-time state of charge. Combined with the real-time load of the power grid, the output of new energy sources, the voltage of the access bus, and the power flow data of the line, it updates the available adjustable capacity, power adjustment range and response capability of each energy storage unit in real time to construct a three-level real-time state perception model. The Level 4 AI Fault Prediction and Adaptive Optimization Model Construction Module is configured to construct a Level 4 AI Fault Prediction and Adaptive Optimization Model by integrating machine learning, time series prediction, and anomaly recognition AI algorithms based on the full-dimensional real-time data of the Level 3 Real-time State Awareness Model. The five-level three-dimensional game-theoretic control model construction module is configured to integrate three control modes: grid safety assessment, grid emergency safety regulation, and electricity spot market economic regulation. It adopts a multi-objective game optimization mechanism to construct a five-level three-dimensional game-theoretic control model for full-scenario optimization and control of energy storage power station groups. This is based on the first-level basic rated operating condition model, the second-level historical operation statistical credibility model, the third-level real-time status perception model, and the fourth-level artificial intelligence fault prediction and adaptive optimization model.
[0019] The present invention also provides a computer device, including one or more processors, wherein the processors store a program, and when the program is executed by the processors, the method for constructing a digital control model for an energy storage power station group as described above is implemented.
[0020] The present invention also provides a computer-readable medium storing a computer program, which, when executed, implements the method for constructing a digital control model for an energy storage power station group as described above.
[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention constructs a five-level digital model that is progressively enhanced and empowered at each level. This model enables full-dimensional digital modeling of large-scale energy storage resources, from the solidification of basic parameters, mining of historical patterns, accurate perception of real-time status, intelligent fault prediction to safe and economic game-theoretic regulation. It transforms the passive, adjustable, and flexible resources of energy storage into observable, measurable, adjustable, and controllable highly reliable grid support resources. It precisely adapts to the hierarchical and zoned regulation needs of multi-megawatt-level energy storage power station clusters, significantly improving the power supply accuracy under normal grid operating conditions, the reliability of resource support under emergency operating conditions, and the economic efficiency of electricity spot market operation. It solves the problems of existing energy storage regulation modeling technologies, such as single-level modeling, poor dynamic adaptability, lack of game-theoretic regulation mechanism, and insufficient emergency support capability.
[0022] This invention integrates six-dimensional extreme value statistics with a discrete reliability quantification evaluation mechanism, significantly improving the accuracy of routine regulation. It pioneers a six-dimensional boundary characterization method based on maximum / second-largest / second-second-largest and minimum / second-smallest values, combined with a 0-10 discrete quantification reliability rating system. This accurately quantifies the deviation between the actual operating capacity of energy storage and its theoretical rated value, avoiding the shortcomings of traditional models that rely solely on rated parameters for blind regulation. This improves the accuracy of steady-state regulation and cluster hierarchical and partitioned matching, adapting to the needs of large-scale, refined energy storage scheduling.
[0023] This invention constructs an LSTM-based time-series AI fault prediction and adaptive weight optimization quantification mechanism, achieving proactive risk prevention and significantly improving regulation stability. The invention's four-level model can predict faults 12–48 hours in advance; through a dynamic weight correction formula, it achieves fault tolerance reduction for high-risk units and prioritizes output for healthy units, reducing the probability of regulation failure caused by energy storage cluster faults, significantly improving the grid's ability to withstand disturbances, and filling the technological gap in proactive risk control of large-scale energy storage clusters.
[0024] This invention employs a three-dimensional game theory multi-objective quantitative control model encompassing safety, emergency response, and economics, making it adaptable to the control of multi-megawatt-level energy storage clusters across all scenarios, demonstrating significant technological barriers. It achieves synergistic optimization of safety and economics under conventional spot market conditions, adapting to multiple scenarios including cluster control, grid emergency support, and the market-based operation of electricity spot markets, exhibiting strong differentiation and innovation. Attached Figure Description
[0025] Figure 1 This is a flowchart of the method according to Embodiment 1 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.
[0027] Example 1 like Figure 1 As shown, a method for constructing a digital control model for an energy storage power station cluster is characterized by the following steps: S1. Collect the rated operating parameters of each energy storage unit in the energy storage power station group and construct a primary basic rated operating condition model. S2. Collect historical operation data of the energy storage power station group throughout its entire lifecycle. Using daily, weekly, monthly, quarterly, and annual statistical periods, collect six-dimensional extreme value data of the operation indicators of each energy storage unit to evaluate the reliability of extreme value operation; count the effective regulation times of each energy storage unit participating in grid regulation, calculate the average regulation value index, and evaluate the reliability of routine regulation; evaluate the reliability of the fit between the operation parameters of each energy storage unit and the rated operating condition parameters; and construct a two-level historical operation statistical reliability model using a discrete reliability evaluation mechanism. S3. Based on the rated operating parameters of the first-level basic rated operating condition model and the high-reliability historical operating data of the second-level historical operating statistical reliability model, the on-site monitoring data of the energy storage power station group and the power grid operation data are accessed in real time. Battery health status parameters, equipment operating status, and battery real-time state of charge are collected simultaneously. Combined with real-time power grid load, new energy output, access bus voltage, and line power flow data, the available adjustable capacity, power adjustment range, and response capability of each energy storage unit are updated in real time to construct a third-level real-time state perception model. S4. Based on the full-dimensional real-time data of the three-level real-time state perception model, integrate machine learning, time series prediction, and anomaly recognition artificial intelligence algorithms to construct a four-level artificial intelligence fault prediction and adaptive optimization model. S5. Based on the first-level basic rated operating condition model, the second-level historical operation statistical credibility model, the third-level real-time status perception model, and the fourth-level artificial intelligence fault prediction and adaptive optimization model, and integrating three control modes—grid network safety assessment, grid emergency safety regulation, and electricity spot market economic regulation—a five-level three-dimensional game control model is constructed using a multi-objective game optimization mechanism to carry out full-scenario optimization control of energy storage power station groups.
[0028] Specifically, in step S1, the rated operating parameters include rated capacity, rated charge / discharge power, rated charge / discharge voltage, rated operating current, rated operating efficiency, rated response delay, and maximum allowable charge / discharge rate; in step S2, the operating indicators include charge / discharge power, adjustment range, response speed, and continuous operating time; the six-dimensional extreme value data include the maximum value, the second largest value, the second second largest value, the minimum value, the second smallest value, and the second second smallest value.
[0029] In step S1, a first-level basic rated operating condition model is constructed, including: Construct the rated parameter matrix of the energy storage unit, and define the rated operating condition parameter vector of energy storage unit i:
[0030] In the formula: The rated charge and discharge power of unit i; Rated capacity; This is the rated charge / discharge voltage; Rated operating current; Rated operating efficiency; This is the rated response delay; This refers to the rated maximum permissible charge / discharge rate. The rated baseline matrix for energy storage clusters is as follows: .
[0031] In step S2, taking charge and discharge power as an example, the six-dimensional extreme values are statistically analyzed: Power sequence of unit i within statistical period T Extract the six-dimensional extreme value set:
[0032] In the formula: These are the maximum value, the second largest value, and the second second largest value, respectively. These represent the minimum, second minimum, and second-second minimum values, accurately depicting the actual operating boundaries of the equipment.
[0033] Average regulation power within the cycle:
[0034] In the formula: N represents the number of effective adjustments within the period. Let be the power adjusted in the kth iteration.
[0035] In step S2, a two-level historical operational statistical credibility model is constructed using a discrete credibility evaluation mechanism, including: Constructing scoring functions for three types of discrete evaluation indicators These correspond to the reliability of extreme value operation, the reliability of normal adjustment, and the reliability of operating condition adaptation, respectively. The overall reliability score is as follows:
[0036] in accordance with The confidence level is divided into three levels based on thresholds: high confidence (8≤Q≤10), medium confidence (4≤Q<8), and low confidence (0≤Q<4). This achieves discrete quantitative rating of energy storage operation capabilities, overcoming the technical deficiency of traditional models that lack quantitative confidence.
[0037] Specifically, in step S3, the battery health status parameters include battery degradation rate, internal resistance change, capacity loss value, aging degree, and battery cluster consistency deviation; the equipment operating status is divided into four categories: normal, warning, fault, and maintenance.
[0038] In step S3, a three-level real-time state awareness model is constructed, including: (1) Quantitative model of battery health status
[0039] In the formula: This represents the current actual available capacity. For rated capacity, Characterizes the degree of battery aging and degradation; (2) Real-time effective adjustable power calculation model
[0040] In the formula: The unit i has adjustable power available in real time; This formula provides an effective constraint coefficient for SOC adjustment. It dynamically adjusts the value based on the state of charge to avoid overcharging and over-discharging. The formula achieves three-dimensional dynamic correction from rated parameters to healthy state to real-time operating conditions, accurately outputting the baseline of adjustable resources in real time.
[0041] In step S4, a four-level artificial intelligence fault prediction and adaptive optimization model is constructed, including: Using a long short-term memory neural network artificial intelligence algorithm, a fault prediction model is trained with battery health state degradation trend, battery state of charge fluctuation law, equipment operating temperature, electrical parameter deviation, and historical fault sample data as training datasets. The model outputs fault probability, fault impact range, fault prediction time and adjustable resource loss after fault. An adaptive optimization and control module is built to correct the energy storage control strategy in advance based on the fault prediction results. High-risk energy storage units are operated at reduced capacity and controlled with fault tolerance in advance, while healthy and high-quality energy storage units are given priority in the allocation of control tasks.
[0042] (1) LSTM timing fault prediction model Historical time-series state dataset Using the input, construct a fault probability prediction function:
[0043] In the formula: Given the potential failure probability of unit i, output the quantified value of failure risk.
[0044] (2) Adaptive adjustment weight correction formula
[0045]
[0046] In the formula: To adjust the weighting coefficients; This is the risk penalty coefficient; To ultimately issue control commands for power, adaptive optimization is achieved, enabling high-risk units to automatically reduce their output while healthy units are given priority in output.
[0047] Specifically, in step S5, the following dimensions are considered: Grid safety assessment: Real-time analysis of power flow constraints, voltage stability thresholds, and grid topology carrying capacity of the main and branch lines of the power grid; verification of the impact of energy storage charging and discharging on the power grid; and delineation of safety control boundaries. Emergency safety regulation: For emergency states such as grid failures, sudden load loss, and new energy disconnection, priority is given to ensuring grid safety and stability, maximizing the use of energy storage emergency regulation resources to support the rapid recovery of the grid to steady-state operation. Spot market economic regulation: Combining electricity spot market price fluctuations, peak-valley price differences, and trading rules, the timing and power of energy storage charging and discharging are optimized to reduce grid operating costs and improve the economic benefits of energy storage resource utilization.
[0048] In step S5, a five-level three-dimensional game regulation model is constructed, including: Construct a multi-objective game objective function that balances the optimality of network infrastructure security, emergency support, and spot market economics. The overall objective function is:
[0049] Constraints:
[0050] In the formula: The target for safe loss of the space frame; The target is to compensate for the emergency power deficit; The economic benefit target for electricity spot trading; These are dynamic weighting coefficients; Weighted dynamic game logic: Power grid emergency conditions: Prioritize ensuring the safety of emergency power supply; Standard spot market operating conditions: Balanced assignment maximizes economic benefits under security constraints.
[0051] Real-time verification of power flow constraints, voltage stability thresholds, and grid topology carrying capacity of the main and branch lines of the power grid is performed to define the safety control boundaries of energy storage. For emergency conditions such as grid faults, sudden load loss, and new energy disconnection, an emergency safety regulation priority strategy is initiated to maximize the use of energy storage emergency regulation resources to support the steady-state recovery of the power grid. For normal operating conditions, energy storage charging and discharging strategies are optimized based on electricity spot market price fluctuations, peak-valley price differences, and trading rules. A multi-objective game theory algorithm is used to balance the relationship between grid safety constraints, emergency guarantee priorities, and spot market economic benefits, achieving precise and controllable regulation of multi-megawatt-level energy storage power station clusters across different layers, zones, scenarios, and priorities. By employing a multi-objective game theory algorithm, the relationship between grid security constraints, emergency response priorities, and spot economic benefits is balanced. Under normal operating conditions, safety and economy are optimized, while under emergency conditions, priority is given to ensuring grid security and stability. This enables precise and controllable regulation of large-scale energy storage resources by layer, zone, scenario, and priority.
[0052] Application examples: A provincial-level energy storage power station cluster with a capacity of tens of millions of kilowatts was selected as the implementation target. The energy storage resources in the entire region were divided into four levels of management: provincial control layer, regional layer, substation layer, and energy storage unit layer. The five-level digital control model construction method of this invention was fully implemented. First-level basic rated operating condition model construction: divide energy storage units / substations / areas according to grid zoning and grid topology, batch input all factory parameters such as rated capacity, charging and discharging power, voltage, current, response delay, and charging and discharging rate, unify data accuracy thresholds and verification rules, and generate a global static rated parameter matrix as the benchmark for all upper-level models.
[0053] Building a Level 2 Historical Operation Statistical Reliability Model: Retrieve the daily / monthly / quarterly charge / discharge historical curves of the entire energy storage unit over the past 12 months, extract six-dimensional extreme value data, and count the total number of adjustments and the number of effective adjustments in each cycle; independently weight and score three types of indicators—extreme value operation, routine adjustment, and operating condition adaptation—from 0 to 10 points, and divide them into three levels of high / medium / low reliability ledgers to correct the deviation of the pure theoretical rated parameters of the Level 1 model.
[0054] A three-level real-time status awareness model is established: Millisecond-level access to full-site BMS monitoring data and real-time power grid dispatch data; relying on a matching voltage relaxation integral area (SOH) fast assessment algorithm, second-level output of all battery health status; synchronously collecting real-time SOC, equipment operating status, bus voltage, and line power flow. The real-time adjustable power of each unit is dynamically updated, and the static deviation of the secondary historical statistics is corrected.
[0055] The fourth-level AI fault prediction and adaptive optimization model is built: an LSTM time series prediction model is trained using a dataset of 3 years of historical faults, SOH decay time series, SOC fluctuations, temperature, and electrical deviations of the entire station. It can predict battery thermal runaway, power module failure, and line contact faults 36 hours in advance. The output correction coefficient is dynamically generated according to the fault probability. Energy storage units with high SOH and low fault risk take priority in undertaking grid frequency regulation and peak shaving tasks, while fault risk units automatically operate at reduced capacity.
[0056] A five-level three-dimensional game-theoretic control model is established: real-time verification of power flow and bus voltage safety boundaries across the entire network; dynamic weight allocation distinguishes between two operating conditions. ① Emergency conditions such as grid failure, large-scale disconnection of new energy sources from the grid, and sudden load loss: Emergency safety has the highest priority, maximizing the use of energy storage resources to quickly smooth out the power shortage in the grid and restore the system to steady state; ② Conventional electricity spot market conditions: Balance the weights of grid security and economic benefits, optimize the charging and discharging time and power of energy storage by combining peak and valley electricity price differences, and improve the overall profitability of power plants.
[0057] Actual engineering measurement results: The effective regulation and utilization rate of energy storage has increased from 85% to 93.1%. The overall assessment error of energy storage resource status is 4.2%, which is 86.5% lower than that of traditional static models. The fault prediction accuracy rate is 96.3%, and it can reliably identify potential equipment hazards 12 to 48 hours in advance; The power grid emergency control response can be achieved in milliseconds, while the response delay of traditional solutions is greater than 1 second. The parallel calculation of tens of thousands of energy storage units across the entire region takes only tens of seconds, without the need for high-performance servers, and can be directly deployed on existing energy storage scheduling platforms; The full-site battery SOH parallel assessment relies on existing BMS data collection, requires no hardware modification, and can complete a health survey of tens of thousands of batteries in minutes.
[0058] Verified through actual engineering projects, the modeling method of this invention can achieve observable and measurable resources for multi-megawatt-level energy storage power station clusters. Compared with traditional energy storage control models, the five-level progressive quantification model of this invention has core advantages such as ultra-high computational efficiency, extremely fast verification response, high-precision prediction and judgment, and super-strong cluster aggregation control in the scenario of hierarchical and regional control of large-scale energy storage clusters. At the same time, it is equipped with a high-speed communication system device to comprehensively improve the overall control performance of large-scale energy storage clusters. The specific performance advantages are as follows: 1. Ultra-high-speed computing and state verification capabilities This invention incorporates a hierarchical, decoupled quantization calculation logic at each level of the model. Each level of the mathematical model operates independently, and hierarchical data is reused, avoiding the repetitive and redundant calculations inherent in traditional integrated modeling. This enables parallel batch calculations of the status of energy storage unit clusters at the tens of millions of kilowatts and tens of thousands of units. For core parameters such as the overall energy storage SOH health status, SOC state of charge, real-time adjustable power, operational reliability level, and grid safety margin, it can quickly and accurately verify the adjustable capabilities of hierarchical and zoned energy storage resources. This fully meets the high-frequency, high-precision, and high-efficiency dynamic verification requirements of large-scale energy storage clusters, solving the pain points of traditional models such as high calculation latency, delayed verification, and inability to adapt to real-time verification of large-scale clusters.
[0059] 2. Rapid operating condition response and command execution capability This invention's model is equipped with a high-speed communication system, employing a dual-link transmission architecture of 5G private network + power-dedicated optical fiber to achieve high-speed data interconnection at four levels: control center, regional cluster, energy storage substation, and individual unit. Combined with a five-level model-based pre-judgment and dynamic pre-correction control mechanism, optimal control commands can be directly output without secondary verification of on-site data. Compared to the second-level response of traditional energy storage control systems, this solution significantly reduces overall control response latency. In emergency situations such as grid faults, sudden load changes, and new energy fluctuations, it can achieve millisecond-level rapid response of energy storage resources, greatly improving the grid's emergency support speed and adaptability to different operating conditions.
[0060] 3. Highly accurate fault prediction and condition assessment capabilities This invention's four-level AI fault prediction model, trained on a multi-dimensional time-series dataset and combined with a real-time state quantification correction mechanism, can predict potential faults in energy storage units 12-48 hours in advance. Simultaneously, relying on a two-level six-dimensional extreme value statistics and discrete reliable quantification system, it accurately distinguishes between the theoretical rated capacity and the actual available adjustable capacity of energy storage, solving the problems of large deviations in state assessment, delayed risk prediction, and inaccurate calculation of adjustable resources in traditional models. This provides high-precision data support for the precise hierarchical and zoned control of large-scale energy storage.
[0061] 4. Extremely strong large-scale cluster aggregation and control capabilities Addressing the challenges of decentralized deployment, regional grid connection, and massive number of units in multi-megawatt energy storage power station clusters, this invention employs a hierarchical aggregation, regional optimization, and global collaborative model operation logic. It supports four-level resource aggregation: unit, substation, region, and global. This allows for the dynamic aggregation of energy storage resources with different health states, reliability levels, and grid connection areas, forming a standardized, uniformly dispatchable flexible resource cluster. The model can differentiate the allocation of regulation tasks to each region and substation based on grid load pressure, grid constraints, emergency needs, and spot electricity price differences. This enables refined division of labor and optimized aggregation scheduling within large-scale energy storage clusters, effectively solving the problems of weak cluster aggregation capabilities, unbalanced regional regulation, and uneven resource utilization inherent in traditional models.
[0062] 5. Overall control and comprehensive performance improvement effect With the support of high-speed communication devices and the coordinated operation of high-precision and high-efficiency digital models, this invention can achieve precise and controllable regulation of multi-megawatt-level energy storage power station clusters in a hierarchical and zoned manner. Compared with traditional regulation schemes, the overall performance has achieved a comprehensive quantitative leap: the effective utilization rate of energy storage resources has increased from 85% to over 92%, is adjustable, has a fault prediction accuracy rate of ≥95%, and has strong engineering adaptability and technological advancement, demonstrating clear quantitative technological progress and substantial innovation.
[0063] Example 2 This invention provides a system for constructing a digital control model for an energy storage power station cluster, comprising: The first-level basic rated operating condition model construction module is configured to collect the rated operating condition parameters of each energy storage unit in the energy storage power station group and construct the first-level basic rated operating condition model. The module for constructing a secondary historical operation statistical credibility model is configured to collect full-cycle historical operation data of energy storage power station groups, with daily, weekly, monthly, quarterly, and annual statistical periods. It calculates the six-dimensional extreme value data of the operation indicators of each energy storage unit and evaluates the credibility of extreme value operation; it also calculates the effective regulation times of each energy storage unit participating in grid regulation, calculates the average regulation value index, and evaluates the credibility of routine regulation; it evaluates the compatibility credibility of the operating parameters of each energy storage unit with the rated operating condition parameters; and it constructs a secondary historical operation statistical credibility model using a discrete credibility evaluation mechanism. The three-level real-time state perception model construction module is configured to rely on the rated operating parameters of the first-level basic rated operating condition model and the high-reliability historical operating data of the second-level historical operating statistical reliability model, and to access the on-site monitoring data of the energy storage power station group and the power grid operation data in real time. It synchronously collects battery health status parameters, equipment operating status, and battery real-time state of charge. Combined with the real-time load of the power grid, the output of new energy sources, the voltage of the access bus, and the power flow data of the line, it updates the available adjustable capacity, power adjustment range and response capability of each energy storage unit in real time to construct a three-level real-time state perception model. The Level 4 AI Fault Prediction and Adaptive Optimization Model Construction Module is configured to construct a Level 4 AI Fault Prediction and Adaptive Optimization Model by integrating machine learning, time series prediction, and anomaly recognition AI algorithms based on the full-dimensional real-time data of the Level 3 Real-time State Awareness Model. The five-level three-dimensional game-theoretic control model construction module is configured to be based on the first-level basic rated operating condition model, the second-level historical operation statistical credibility model, the third-level real-time status perception model, and the fourth-level artificial intelligence fault prediction and adaptive optimization model. It integrates three control modes: power grid safety assessment, power grid emergency safety regulation, and electricity spot market economic regulation. It adopts a multi-objective game optimization mechanism to construct a five-level three-dimensional game-theoretic control model for power grid safety, emergency regulation, and spot market economic regulation, and conducts full-scenario optimization and control of energy storage power station groups.
[0064] Example 3 This invention provides a computer device including one or more processors, each processor storing a program. When the program is executed by the processor, it implements the method for constructing a digital control model for an energy storage power station group as described in Embodiment 1.
[0065] Example 4 This invention provides a computer-readable medium storing a computer program, which, when executed, implements the method for constructing a digital control model for an energy storage power station group as described in Embodiment 1.
[0066] The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can still make modifications or equivalent substitutions to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention are within the protection scope of the claims of the present invention.
Claims
1. A method for constructing a digital control model for an energy storage power station cluster, characterized in that, Includes the following steps: Collect the rated operating parameters of each energy storage unit in the energy storage power station group and construct a primary basic rated operating condition model; Collect historical operation data of the entire life cycle of the energy storage power station group, and use daily, weekly, monthly, quarterly and annual statistical periods to collect six-dimensional extreme value data of the operation indicators of each energy storage unit and evaluate the reliability of extreme value operation. The effective regulation times of each energy storage unit in grid regulation are counted, the average regulation value index is calculated, and the reliability of routine regulation is evaluated; the compatibility reliability of the operating parameters of each energy storage unit with the rated operating condition parameters is evaluated; and a two-level historical operation statistical reliability model is constructed using a discrete reliability evaluation mechanism. Based on the rated operating parameters of the first-level basic rated operating condition model and the high-reliability historical operating data of the second-level historical operating statistical reliability model, the on-site monitoring data of the energy storage power station group and the power grid operation data are accessed in real time. Battery health status parameters, equipment operating status, and battery real-time state of charge are collected simultaneously. Combined with the real-time load of the power grid, the output of new energy sources, the voltage of the access bus, and the power flow data of the line, the available adjustable capacity, power adjustment range and response capability of each energy storage unit are updated in real time to construct a third-level real-time state perception model. Based on the full-dimensional real-time data of the three-level real-time state perception model, and by integrating machine learning, time series prediction, and anomaly recognition artificial intelligence algorithms, a four-level artificial intelligence fault prediction and adaptive optimization model is constructed. Based on the aforementioned first-level basic rated operating condition model, second-level historical operation statistical credibility model, third-level real-time status perception model, and fourth-level artificial intelligence fault prediction and adaptive optimization model, and integrating three control modes—grid network safety assessment, grid emergency safety regulation, and electricity spot market economic regulation—a five-level three-dimensional game control model is constructed using a multi-objective game optimization mechanism to conduct full-scenario optimization control of energy storage power station clusters.
2. The method for constructing a digital control model for energy storage power station clusters according to claim 1, characterized in that: The rated operating parameters include rated capacity, rated charge / discharge power, rated charge / discharge voltage, rated operating current, rated operating efficiency, rated response delay, and maximum allowable charge / discharge rate; the operating indicators include charge / discharge power, adjustment range, response speed, and continuous operating time; the six-dimensional extreme value data include the maximum value, the second largest value, the second second largest value, the minimum value, the second smallest value, and the second second smallest value.
3. The method for constructing a digital control model for an energy storage power station cluster according to claim 2, characterized in that: The construction of the primary basic rated operating condition model includes: Construct the rated parameter matrix of the energy storage unit, and define the rated operating condition parameter vector of energy storage unit i: In the formula: The rated charge and discharge power of unit i; Rated capacity; This is the rated charge / discharge voltage; Rated operating current; Rated operating efficiency; This is the rated response delay; This refers to the rated maximum permissible charge / discharge rate. The rated baseline matrix for energy storage clusters is as follows: 。 4. The method for constructing a digital control model for an energy storage power station cluster according to claim 2, characterized in that: A two-level historical operational statistical credibility model is constructed using a discrete credibility evaluation mechanism, including: Constructing scoring functions for three types of discrete evaluation indicators These correspond to the reliability of extreme value operation, the reliability of normal adjustment, and the reliability of operating condition adaptation, respectively. The overall reliability score is as follows: in accordance with The confidence level is divided into three levels based on the threshold: high confidence (8≤Q≤10), medium confidence (4≤Q<8), and low confidence (0≤Q<4).
5. The method for constructing a digital control model for an energy storage power station cluster according to claim 1, characterized in that: Construct a three-level real-time state awareness model, including: Battery health state quantification model In the formula: This represents the current actual available capacity. For rated capacity, Characterizes the degree of battery aging and degradation; Real-time effective adjustable power calculation model In the formula: The unit i has adjustable power available in real time; The constraint coefficient is adjusted effectively for SOC.
6. The method for constructing a digital control model for an energy storage power station cluster according to claim 1, characterized in that: Construct a four-level artificial intelligence fault prediction and adaptive optimization model, including: Using a long short-term memory neural network artificial intelligence algorithm, a fault prediction model is trained with battery health state degradation trend, battery state of charge fluctuation law, equipment operating temperature, electrical parameter deviation, and historical fault sample data as training datasets. The model outputs fault probability, fault impact range, fault prediction time and adjustable resource loss after fault. An adaptive optimization and control module is built to correct the energy storage control strategy in advance based on the fault prediction results. High-risk energy storage units are operated at reduced capacity and controlled with fault tolerance in advance, while healthy and high-quality energy storage units are given priority in the allocation of control tasks.
7. The method for constructing a digital control model for an energy storage power station cluster according to claim 1, characterized in that: Construct a five-level three-dimensional game-theoretic control model, including: Construct a multi-objective game objective function that balances the optimality of network infrastructure security, emergency support, and spot market economics. The overall objective function is: Constraints: In the formula: The target for safe loss of the space frame; The target is to compensate for the emergency power deficit; The economic benefit target for electricity spot trading; These are dynamic weighting coefficients; Weighted dynamic game logic: Power grid emergency conditions: Prioritize ensuring the safety of emergency power supply; Standard spot market operating conditions: Balanced assignment maximizes economic benefits under security constraints.
8. A system for constructing a digital control model for an energy storage power station cluster, characterized in that, include: The first-level basic rated operating condition model construction module is configured to collect the rated operating condition parameters of each energy storage unit in the energy storage power station group and construct the first-level basic rated operating condition model. The Level 2 Historical Operation Statistical Credibility Model Construction Module is configured to collect historical operation data of the entire life cycle of the energy storage power station group, and to collect six-dimensional extreme value data of the operation indicators of each energy storage unit in the form of daily, weekly, monthly, quarterly, and annual statistical periods, and to evaluate the credibility of extreme value operation. The number of effective regulation times of each energy storage unit participating in grid regulation is counted, the average regulation value index is calculated, and the reliability of conventional regulation is evaluated; the reliability of the fit between the operating parameters of each energy storage unit and the rated operating condition parameters is evaluated. A two-level historical operational statistical credibility model is constructed using a discrete credibility evaluation mechanism; The three-level real-time state perception model construction module is configured to rely on the rated operating parameters of the first-level basic rated operating condition model and the high-reliability historical operating data of the second-level historical operating statistical reliability model, and to access the on-site monitoring data of the energy storage power station group and the power grid operation data in real time. It synchronously collects battery health status parameters, equipment operating status, and battery real-time state of charge. Combined with the real-time load of the power grid, the output of new energy sources, the voltage of the access bus, and the power flow data of the line, it updates the available adjustable capacity, power adjustment range and response capability of each energy storage unit in real time to construct a three-level real-time state perception model. The Level 4 AI Fault Prediction and Adaptive Optimization Model Construction Module is configured to construct a Level 4 AI Fault Prediction and Adaptive Optimization Model by integrating machine learning, time series prediction, and anomaly recognition AI algorithms based on the full-dimensional real-time data of the Level 3 Real-time State Awareness Model. The five-level three-dimensional game-theoretic control model construction module is configured to integrate three control modes: grid safety assessment, grid emergency safety regulation, and electricity spot market economic regulation. It adopts a multi-objective game optimization mechanism to construct a five-level three-dimensional game-theoretic control model for full-scenario optimization and control of energy storage power station groups. This is based on the first-level basic rated operating condition model, the second-level historical operation statistical credibility model, the third-level real-time status perception model, and the fourth-level artificial intelligence fault prediction and adaptive optimization model.
9. A computer device comprising one or more processors, said processor storing a program, characterized in that, When the program is executed by the processor, it implements the method for constructing a digital control model for an energy storage power station group as described in any one of claims 1 to 7.
10. A computer-readable medium storing a computer program, characterized in that, When the computer program is executed, it implements the method for constructing a digital control model for energy storage power station clusters as described in any one of claims 1 to 7.
