Economic evaluation method and system for distributed energy storage of power distribution network

Abstract

The invention provides a power distribution network distributed energy storage economy evaluation method and system. According to the method, key features are extracted by combining a sliding window technology with an optimal approximate matrix multiplication; establishing an investment cost sub-model, an operation and maintenance cost sub-model and an income sub-model to form a comprehensive evaluation model; dynamic optimization is realized through an incremental graph structure and an approximate hopping neighborhood algorithm; and analyzing and processing a multi-scene evaluation result by adopting nearest neighbor scene aggregation. According to the method, the large-scale data processing efficiency is improved, the dynamic adaptive capacity of the evaluation model is enhanced, accurate evaluation of the economical efficiency of the energy storage system under multiple scenes is realized, and a reliable basis is provided for planning and decision making of the energy storage system of the power distribution network.

Description

Technical Field

[0001] This application relates to the field of power system economic assessment technology, and in particular to a method and system for assessing the economic viability of distributed energy storage in distribution networks. Background Technology

[0002] Economic evaluation of distributed energy storage systems in power distribution networks is a crucial step in power system planning and operation, involving a comprehensive analysis of multiple dimensions such as energy storage equipment investment, operation and maintenance costs, and grid benefits. Accurately assessing the economics of energy storage systems is essential for optimizing resource allocation and improving return on investment.

[0003] Currently, common evaluation methods mainly include traditional economic evaluation methods such as the net present value (NPV) method and the payback period method. These methods are usually based on simple cost-benefit analysis models and use fixed parameters for calculation, such as fixed charge-discharge efficiency and equipment lifespan. These methods are simple to operate, but they often cannot adapt to the dynamic changes of energy storage systems in actual operation, and the evaluation results may deviate from the actual situation.

[0004] More advanced evaluation technologies employ a data-driven approach, establishing an energy storage system operation database and combining it with multi-dimensional data such as historical load and electricity prices to construct an economic benefit evaluation model. This technology optimizes the charging and discharging strategies of the energy storage system and calculates the system's economic benefit indicators, thereby improving the accuracy of the evaluation results to a certain extent.

[0005] However, existing technologies still have significant shortcomings in practical applications: First, computational efficiency is low during large-scale data processing, making real-time assessment difficult; second, existing models lack adaptability to dynamically changing power grid operating environments, and the accuracy of assessment results needs improvement. Especially when facing complex and volatile power market environments, assessment models need to be able to quickly respond to parameter changes and adjust assessment results in a timely manner to provide effective support for decision-making. Summary of the Invention

[0006] In view of this, this application provides a method and system for evaluating the economic efficiency of distributed energy storage in power distribution networks, which solves the problems of low efficiency in large-scale data processing and insufficient adaptability of evaluation models in the prior art.

[0007] This application provides a method for evaluating the economic viability of distributed energy storage in a distribution network, comprising: acquiring distribution network operation data; establishing a multi-dimensional data matrix using sliding window technology; determining feature correlation thresholds; and extracting key features for evaluating the economic viability of the energy storage system; establishing an investment cost model, an operation and maintenance cost model, and a revenue model for the energy storage system based on the key features; forming a comprehensive economic benefit evaluation model through standardization; converting the comprehensive economic benefit evaluation model into an incremental graph structure; achieving dynamic optimization through approximate skip neighborhood calculation; and using the optimized evaluation model to construct an evaluation scenario library, processing the multi-scenario evaluation results using the nearest neighbor scenario aggregation analysis method, and generating an economic evaluation report for the energy storage system.

[0008] A multi-dimensional data matrix is ​​established using sliding window technology to extract key features for the economic evaluation of energy storage systems. This includes: determining the sliding window size using the 24-hour charge-discharge cycle of the energy storage system; processing the distribution network operation data using the sliding window size to obtain multiple overlapping time window segments; constructing a feature matrix based on the data in the multiple overlapping time window segments; performing calculations using the best approximation matrix multiplication algorithm to form an optimized feature matrix; and performing dimensionality reduction processing on the optimized feature matrix to obtain the key features for the economic evaluation of the energy storage system.

[0009] The optimized feature matrix is ​​subjected to dimensionality reduction processing to obtain the key features for the economic evaluation of the energy storage system. This includes: calculating the correlation coefficient matrix between each feature using the optimized feature matrix; identifying and merging features with a correlation coefficient greater than 0.8 based on the feature correlation threshold to obtain a preliminary feature set; inputting the preliminary feature set into the principal component analysis module; selecting principal components with a cumulative contribution rate of 85% to form a dimensionality-reduced feature matrix; and using the dimensionality-reduced feature matrix, applying the XGBoost feature importance scoring method to select the top 30% of features as the key features for the economic evaluation of the energy storage system.

[0010] Based on the aforementioned key characteristics, an investment cost model, an operation and maintenance cost model, and a revenue model for the energy storage system are established to form a comprehensive economic benefit evaluation model. This includes: importing full life-cycle data of the energy storage equipment to establish an investment cost model that includes equipment price and installation costs; processing the operating parameters and maintenance data of the energy storage equipment to construct an operation and maintenance cost model that includes regular maintenance and fault repair costs; analyzing electricity price arbitrage, demand response, and ancillary service data to generate a revenue model; and using the investment cost model, operation and maintenance cost model, and revenue model as inputs, integrating the models through a weighted adaptive algorithm to output the comprehensive economic benefit evaluation model.

[0011] Analyzing electricity price arbitrage, demand response, and ancillary service data to generate a revenue model includes: importing historical electricity price data and load data, using a time-series revenue forecasting model to generate a benchmark revenue for electricity price arbitrage; inputting the benchmark revenue for electricity price arbitrage into the demand response analysis module, and combining it with the demand response subsidy standard to calculate the combined revenue superimposed on the demand response; using the combined revenue, and based on ancillary service data, performing comprehensive calculations to output the revenue model.

[0012] The comprehensive economic benefit assessment model is converted into an incremental graph structure, and dynamic optimization is achieved through approximate jump neighborhood calculation. This includes: importing assessment indicators from the comprehensive economic benefit assessment model; modeling the assessment indicators as nodes with attribute sets; establishing edge connections based on the influence relationships between indicators to construct an incremental graph model; processing the incremental graph model; setting hop count parameters of 3 to 5 steps; using an importance probability sampling method to perform approximate jump neighborhood calculation between nodes to obtain key influencing factors; dynamically adjusting the parameters of the incremental graph model based on the key influencing factors; and outputting the optimized assessment model.

[0013] Based on the key influencing factors, the parameters of the incremental graph model are dynamically adjusted, including: measuring the parameter changes of the key influencing factors, identifying changes greater than 20% of the parameter standard deviation, and determining the node parameters that need to be updated; using the node parameters that need to be updated, sequentially executing update operations through an incremental calculation strategy and passing them to adjacent nodes to generate updated node parameter states; using the updated node parameter states, recalculating the overall state of the incremental graph model, and outputting the optimized evaluation model.

[0014] Using the optimized evaluation model, an evaluation scenario library is constructed. The nearest neighbor scenario aggregation analysis method is used to process the evaluation results of multiple scenarios, including: importing historical operating data to construct regular operating scenarios containing 24-hour load curves and extreme operating scenarios with load fluctuations exceeding 40%, forming an evaluation scenario library; processing the evaluation scenario library, performing cluster analysis using a scenario similarity calculation method based on Euclidean distance to generate typical scenario groups with different load characteristics; and using the typical scenario groups to perform economic index calculations for the energy storage system and output multi-scenario evaluation results.

[0015] Using the aforementioned typical scenario group, the economic indicators of the energy storage system are calculated, and multi-scenario evaluation results are output, including: extracting the operating data of the typical scenario group, calculating the average daily revenue and annualized investment payback period of the energy storage system, and forming benchmark economic indicators; using the Monte Carlo simulation method, generating an economic indicator distribution containing a 90% confidence interval based on the benchmark economic indicators; analyzing the economic indicator distribution, evaluating the revenue composition and risk factors under different scenarios, and outputting the economic evaluation report of the energy storage system.

[0016] This application also provides an economic evaluation device for distributed energy storage in power distribution networks, comprising: a data preprocessing and feature extraction module for acquiring power distribution network operation data, establishing a multi-dimensional data matrix using sliding window technology, determining feature correlation thresholds, and extracting key features for economic evaluation of the energy storage system; an economic benefit modeling module for establishing an investment cost model, operation and maintenance cost model, and revenue model of the energy storage system based on the key features, and forming a comprehensive economic benefit evaluation model through standardization processing; a dynamic evaluation and optimization module for converting the comprehensive economic benefit evaluation model into an incremental graph structure, achieving dynamic optimization through approximate skip neighborhood calculation, and obtaining an optimized evaluation model; and a multi-scenario evaluation module for constructing an evaluation scenario library using the optimized evaluation model, processing the multi-scenario evaluation results using the nearest neighbor scenario aggregation analysis method, and generating an economic evaluation report for the energy storage system.

[0017] This application also provides a computer device, the computer device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the above-described method for evaluating the economic efficiency of distributed energy storage in power distribution networks.

[0018] This application also provides a computer-readable storage medium storing computer instructions for causing a computer to execute the above-described method for evaluating the economic efficiency of distributed energy storage in a power distribution network.

[0019] This application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the above-described method for evaluating the economic efficiency of distributed energy storage in power distribution networks.

[0020] This application has the following technical effects:

[0021] (1) Feature extraction is performed using sliding window and best approximation matrix multiplication techniques, which improves the efficiency of large-scale data processing and reduces the computational complexity from the traditional O(n) to O(n) matrix multiplication. 3 ) Optimized to O(n^2.807);

[0022] (2) The incremental graph and approximate skip neighborhood algorithm are used to realize the dynamic optimization of the evaluation model, so that the model can respond to parameter changes in real time, which improves the accuracy and timeliness of the evaluation results;

[0023] (3) A comprehensive evaluation system with three sub-models (investment, operation and maintenance, and revenue) was established, realizing a full-dimensional economic analysis;

[0024] (4) By using the nearest neighbor scenario aggregation analysis method, the evaluation results of multiple scenarios are effectively integrated, providing more comprehensive support for decision-making. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the embodiments will be briefly described below. These drawings are incorporated in and constitute a part of this specification. They illustrate embodiments conforming to this disclosure and, together with the specification, serve to explain the technical solutions of this disclosure. It should be understood that the following drawings only show some embodiments of this disclosure and should not be considered as limiting the scope. Those skilled in the art can obtain other related drawings based on these drawings without creative effort.

[0026] Figure 1 A flowchart illustrating the economic evaluation method for distributed energy storage in power distribution networks provided in this application embodiment;

[0027] Figure 2 This is a detailed flowchart of the data preprocessing and feature extraction steps in the embodiments of this application;

[0028] Figure 3 This is a schematic diagram illustrating the process of constructing the feature matrix of the sliding window in an embodiment of this application;

[0029] Figure 4 This is a flowchart illustrating the economic benefit modeling steps in the embodiments of this application;

[0030] Figure 5 This is a flowchart of the dynamic evaluation and optimization process based on incremental graphs in the embodiments of this application. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. The components of the embodiments of this disclosure described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this disclosure provided in the accompanying drawings is not intended to limit the scope of the claimed disclosure, but merely represents selected embodiments of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without inventive effort are within the scope of protection of this disclosure.

[0032] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0033] In this document, the term "and / or" merely describes a relationship, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0034] Example 1

[0035] This application provides a method for evaluating the economic viability of distributed energy storage in power distribution networks, including:

[0036] Acquire power distribution network operation data, establish a multi-dimensional data matrix using sliding window technology, determine feature correlation thresholds, and extract key features for economic evaluation of energy storage systems.

[0037] Based on the key characteristics mentioned above, an investment cost model, an operation and maintenance cost model, and a revenue model for the energy storage system are established, and a comprehensive economic benefit evaluation model is formed after standardization.

[0038] The comprehensive economic benefit evaluation model is converted into an incremental graph structure, and dynamic optimization is achieved through approximate skip neighborhood calculation to obtain the optimized evaluation model.

[0039] Using the optimized evaluation model, an evaluation scenario library is constructed, and the evaluation results of multiple scenarios are processed using the nearest neighbor scenario aggregation analysis method to generate an economic evaluation report of the energy storage system.

[0040] A multi-dimensional data matrix is ​​established using sliding window technology to extract key features for the economic evaluation of energy storage systems. This includes: determining the sliding window size using the 24-hour charge-discharge cycle of the energy storage system; processing the distribution network operation data using the sliding window size to obtain multiple overlapping time window segments; constructing a feature matrix based on the data in the multiple overlapping time window segments; performing calculations using the best approximation matrix multiplication algorithm to form an optimized feature matrix; and performing dimensionality reduction processing on the optimized feature matrix to obtain the key features for the economic evaluation of the energy storage system.

[0041] The optimized feature matrix is ​​subjected to dimensionality reduction processing to obtain the key features for the economic evaluation of the energy storage system. This includes: calculating the correlation coefficient matrix between features using the optimized feature matrix; identifying and merging features with correlation coefficients greater than 0.8 based on the feature correlation threshold to obtain a preliminary feature set; inputting the preliminary feature set into the principal component analysis module; selecting principal components with a cumulative contribution rate of 85% to form the dimensionality-reduced feature matrix; and using the dimensionality-reduced feature matrix...

[0042] XGBoost's feature importance scoring method selects the top 30% of features as key features for the economic evaluation of the energy storage system.

[0043] Based on the aforementioned key characteristics, an investment cost model, an operation and maintenance cost model, and a revenue model for the energy storage system are established to form a comprehensive economic benefit evaluation model. This includes: importing full life-cycle data of the energy storage equipment to establish an investment cost model that includes equipment price and installation costs; processing the operating parameters and maintenance data of the energy storage equipment to construct an operation and maintenance cost model that includes regular maintenance and fault repair costs; analyzing electricity price arbitrage, demand response, and ancillary service data to generate a revenue model; and using the investment cost model, operation and maintenance cost model, and revenue model as inputs, integrating the models through a weighted adaptive algorithm to output the comprehensive economic benefit evaluation model.

[0044] Analyzing electricity price arbitrage, demand response, and ancillary service data to generate a revenue model includes: importing historical electricity price data and load data, using a time-series revenue forecasting model to generate a benchmark revenue for electricity price arbitrage; inputting the benchmark revenue for electricity price arbitrage into the demand response analysis module, and combining it with the demand response subsidy standard to calculate the combined revenue superimposed on the demand response; using the combined revenue, and based on ancillary service data, performing comprehensive calculations to output the revenue model.

[0045] The comprehensive economic benefit assessment model is converted into an incremental graph structure, and dynamic optimization is achieved through approximate jump neighborhood calculation. This includes: importing assessment indicators from the comprehensive economic benefit assessment model; modeling the assessment indicators as nodes with attribute sets; establishing edge connections based on the influence relationships between indicators to construct an incremental graph model; processing the incremental graph model; setting hop count parameters of 3 to 5 steps; using an importance probability sampling method to perform approximate jump neighborhood calculation between nodes to obtain key influencing factors; dynamically adjusting the parameters of the incremental graph model based on the key influencing factors; and outputting the optimized assessment model.

[0046] Based on the key influencing factors, the parameters of the incremental graph model are dynamically adjusted, including: measuring the parameter changes of the key influencing factors, identifying changes greater than 20% of the parameter standard deviation, and determining the node parameters that need to be updated; using the node parameters that need to be updated, sequentially executing update operations through an incremental calculation strategy and passing them to adjacent nodes to generate updated node parameter states; using the updated node parameter states, recalculating the overall state of the incremental graph model, and outputting the optimized evaluation model.

[0047] Using the optimized evaluation model, an evaluation scenario library is constructed. The nearest neighbor scenario aggregation analysis method is used to process the evaluation results of multiple scenarios, including: importing historical operating data to construct regular operating scenarios containing 24-hour load curves and extreme operating scenarios with load fluctuations exceeding 40%, forming an evaluation scenario library; processing the evaluation scenario library, performing cluster analysis using a scenario similarity calculation method based on Euclidean distance to generate typical scenario groups with different load characteristics; and using the typical scenario groups to perform economic index calculations for the energy storage system and output multi-scenario evaluation results.

[0048] Using the aforementioned typical scenario group, the economic indicators of the energy storage system are calculated, and multi-scenario evaluation results are output, including: extracting the operating data of the typical scenario group, calculating the average daily revenue and annualized investment payback period of the energy storage system, and forming benchmark economic indicators; using the Monte Carlo simulation method, generating an economic indicator distribution containing a 90% confidence interval based on the benchmark economic indicators; analyzing the economic indicator distribution, evaluating the revenue composition and risk factors under different scenarios, and outputting the economic evaluation report of the energy storage system.

[0049] Example 2

[0050] like Figure 1 As shown in the figure, this application provides a method for evaluating the economic viability of distributed energy storage in a distribution network, including:

[0051] S1: Data preprocessing and feature extraction;

[0052] Acquire power distribution network operation data, establish a multi-dimensional data matrix using sliding window technology, determine feature correlation thresholds, and extract key features for economic evaluation of energy storage systems.

[0053] Data preprocessing and feature extraction are fundamental steps in evaluating the economic viability of distributed energy storage in distribution networks. This step begins by collecting distribution network operation data, including load curves, electricity price data, and energy storage device parameters. Data cleaning and standardization are then performed to ensure data quality. Specifically, the system identifies and corrects outliers and missing values, using statistical methods to remove data points that significantly deviate from the normal range. Missing data is filled using time-series interpolation to ensure data continuity and reliability.

[0054] In the sliding window feature matrix construction stage, the system sets a 24-hour base window size based on the charge-discharge cycle characteristics of energy storage devices, and uses 1 / 4 of the window size (usually 6 hours) as the sliding step. This setting fully considers the daily charge-discharge characteristics of the energy storage system, ensuring sufficient data overlap. For each time window, the system constructs a multi-dimensional feature matrix, including dimensions such as load data, electricity price data, energy storage parameters, and environmental factors. The construction of the feature matrix uses an improved Strassen algorithm for computational optimization, reducing the computational complexity from the traditional O(n^2) to O(n^2).3 The process efficiency has been optimized to O(n^2.807), significantly improving the efficiency of large-scale data processing.

[0055] The key feature identification and extraction process employs a two-stage strategy of "coarse screening + fine screening." First, the system calculates the Pearson correlation coefficient matrix between features, sets a correlation threshold |θ|>0.8, identifies and merges highly correlated features, and removes redundant information. Second, the system applies Principal Component Analysis (PCA) to select principal components with a cumulative contribution rate of 85% through eigenvalue decomposition, effectively reducing dimensionality. Finally, the system uses the XGBoost algorithm to evaluate the importance of each feature to the target variable, selecting the top 30% of features as the final key feature set. This multi-stage screening method not only effectively reduces data dimensionality but also retains the key features most influential on economic evaluation, laying the foundation for subsequent modeling.

[0056] Practice has proven that this feature extraction method can significantly improve computational efficiency while maintaining model accuracy. Taking a power distribution network energy storage system as an example, the original feature dimension was 50. After processing, 15 key features were retained, improving computational efficiency by approximately 65%, while maintaining model accuracy above 98%, fully validating the effectiveness of the method.

[0057] S2: Economic benefit modeling of energy storage systems;

[0058] Based on the key characteristics mentioned above, an investment cost model, an operation and maintenance cost model, and a revenue model for the energy storage system are established, and a comprehensive economic benefit evaluation model is formed after standardization.

[0059] The economic benefit modeling of energy storage systems achieves comprehensive economic assessment through the construction of a three-layer sub-model. In the investment cost model construction phase, the system, based on the concept of the entire equipment lifecycle, divides costs into direct costs and indirect costs. Direct costs include the cost of energy storage units, power conversion systems, and energy management systems; indirect costs include engineering construction costs, design costs, and commissioning costs. The model uses the net present value (NPV) method to discount future costs and considers the residual value of the equipment. By introducing a time-varying discount rate r(t), the system can more accurately reflect the time value of money at different periods, improving the accuracy of investment cost assessment.

[0060] The operation and maintenance cost model construction process fully considers the unique characteristics of energy storage system operation and maintenance. The model categorizes operation and maintenance costs into two types: fixed and variable. Fixed costs mainly include expenses unrelated to the operating status, such as periodic maintenance fees and labor costs; variable costs are related to the actual operating status of the system and include charging and discharging losses and equipment repair costs. The model innovatively introduces a maintenance cost prediction method based on equipment health status. By analyzing historical operation and maintenance data, it establishes a correlation model between equipment failure probability and maintenance costs, significantly improving the accuracy and foresight of cost prediction.

[0061] The system revenue model employs a multi-faceted revenue aggregation approach, comprehensively considering all revenue sources for the energy storage system. First, the system imports historical electricity price and load data, using time series analysis to calculate the benchmark revenue from electricity price arbitrage—the most fundamental revenue source for the energy storage system. Then, this benchmark revenue is input into the demand response analysis module, and combined with demand response subsidy standards, the combined revenue from demand response is calculated. Finally, the system performs a comprehensive calculation based on ancillary service data, incorporating ancillary service revenue into the overall revenue model. This progressive revenue calculation method ensures the comprehensiveness and accuracy of the revenue assessment.

[0062] In the integration process of the comprehensive benefit evaluation model, the system adopts a layered integration architecture design. First, the three sub-models are standardized to unify the evaluation period and calculation benchmark, ensuring the comparability of indicators across different dimensions. Then, an adaptive weighting algorithm is designed to dynamically adjust the weight coefficients of the sub-models based on the importance of each dimension in different scenarios. This adaptive weighting mechanism allows the model to flexibly respond to changing needs in different application scenarios. Finally, the system comprehensively calculates key economic indicators such as investment payback period, net present value, and internal rate of return, providing intuitive reference for decision-making.

[0063] To enhance the reliability of the model, a sensitivity analysis module has also been introduced. By performing perturbation analysis on key parameters, the system can identify the factors that have the most significant impact on economic benefits, providing decision-makers with a risk assessment reference. For example, in a real-world project, the analysis found that changes in electricity pricing policy had the most significant impact on system revenue, with fluctuations reaching ±15%, while the impact of equipment price fluctuations was relatively small, typically within ±5%. These sensitivity analysis results are of significant guiding importance for developing risk response strategies.

[0064] S3: Dynamic evaluation and optimization based on incremental graphs;

[0065] The comprehensive economic benefit evaluation model is converted into an incremental graph structure, and dynamic optimization is achieved through approximate skip neighborhood calculation to obtain the optimized evaluation model.

[0066] The core innovation of this method is dynamic evaluation and optimization based on incremental graphs. By converting the static evaluation model into a dynamic incremental graph structure, the real-time optimization capability of the evaluation model is achieved. During the incremental graph model construction phase, the system converts each component of the evaluation model into a graph structure, where nodes represent evaluation indicators (such as investment cost, operation and maintenance cost, and revenue), and edges represent the influence relationships between indicators. Each node contains a specific set of attributes, such as numerical range, update frequency, and reliability. This graph structure design enables the model to accurately track the impact path of parameter changes on the overall evaluation results.

[0067] In its implementation, the system employs a storage structure combining adjacency lists and attribute matrices. The adjacency list stores the topological relationships between nodes, while the attribute matrix stores the dynamic characteristics of the nodes. To improve data processing efficiency, the system uses sparse matrix compression storage technology, significantly reducing storage space requirements. Practice shows that this storage structure can save approximately 60% of storage space compared to traditional methods when processing large-scale evaluation data, greatly improving the system's operating efficiency.

[0068] Approximate k-hop neighborhood calculation is a crucial step in model optimization. This algorithm identifies key influencing factors by analyzing the propagation paths of influence between nodes. The system defines the concept of a k-hop neighborhood, which is a subgraph considering all nodes reachable within k steps. Typically, k is set to 3-5 to ensure comprehensive analysis while controlling computational complexity. When the graph is large, the system employs an importance-based sampling method, prioritizing the calculation of paths with higher influence weights to avoid resource waste. Practice shows that this method can reduce computation time by approximately 70% while maintaining over 90% accuracy, significantly improving optimization efficiency.

[0069] During the dynamic optimization of the evaluation model, the system adjusts the model parameters in real time based on the calculation results of approximate skip neighborhoods. This process includes three main steps: First, the changes in parameters of key influencing factors are measured, and changes greater than 20% of the parameter standard deviation are identified to determine the node parameters that need to be updated. Second, using the node parameters that need to be updated, the update operation is sequentially executed through an incremental calculation strategy and passed to neighboring nodes to generate the updated node parameter state. Finally, using the updated node parameter state, the overall state of the incremental graph model is recalculated, and the optimized evaluation model is output.

[0070] To ensure the stability of the optimization process, a dynamic threshold mechanism is introduced. When the parameter change is less than a preset threshold, the system delays the update operation to avoid wasting computational resources due to frequent minor adjustments. The threshold is set based on historical data analysis, typically choosing 20% ​​of the parameter's standard deviation as the baseline value. This mechanism effectively balances the real-time nature of optimization with system resource consumption.

[0071] In practical applications, this optimization method demonstrates significant advantages. Taking a regional distribution network energy storage project as an example, in the event of a sudden change in electricity prices, the traditional model requires recalculating the entire evaluation result, taking approximately 30 minutes. However, using this method, the model adjustment can be completed in only 2-3 minutes, while maintaining the accuracy of the evaluation results. This efficient dynamic optimization capability provides strong support for real-time decision-making in energy storage systems.

[0072] S4: Multi-scenario economic evaluation and result output;

[0073] Using the optimized evaluation model, an evaluation scenario library is constructed, and the evaluation results of multiple scenarios are processed using the nearest neighbor scenario aggregation analysis method to generate an economic evaluation report of the energy storage system.

[0074] Multi-scenario economic evaluation and result output are crucial steps in applying the optimized evaluation model to practical decision support. During the scenario library construction phase, the system built an evaluation scenario library based on historical operational data, including typical and extreme scenarios. Typical scenarios typically include regular weekday and weekend operating conditions across four seasons, totaling eight basic scenarios; extreme scenarios include special situations such as sudden load changes, drastic electricity price fluctuations, and equipment failures, totaling twelve extreme scenarios. In particular, the system defines load fluctuations exceeding 40% as extreme operating scenarios, which pose a significant challenge to the robustness of the evaluation model.

[0075] To enhance the usability of the scenario library, the system employs a scenario generation method based on probability distribution. By analyzing the statistical characteristics of historical data, probability distribution models for various parameters are established. For example, the intraday fluctuations in daily electricity prices may follow a log-normal distribution, with a mean peak-to-valley price difference of 0.35 yuan / kWh and a standard deviation of 0.08 yuan / kWh. Based on these statistical characteristics, the system can automatically generate new scenarios that match actual conditions, thereby enriching the scenario library. Practice shows that the scenarios generated by this method achieve a consistency of over 85% with actual operating conditions, effectively enhancing the reliability of the assessment.

[0076] In the nearest neighbor scenario aggregation analysis, the system defines a scenario similarity index based on Euclidean distance, clustering basic scenarios into typical scenario groups. For example, high-temperature, high-load scenarios on summer weekdays and high-load scenarios during the winter heating season are grouped into the same scenario group due to their similar load characteristics. For each scenario group, the system calculates key economic indicators for the energy storage system, including average daily revenue, peak-valley price difference return, and ancillary service revenue. This clustering analysis method not only reduces computational load but also reveals the inherent connections between different scenarios, providing a new perspective for comprehensive evaluation.

[0077] Through aggregate analysis, the system can identify differences in the economic performance of energy storage systems under different scenarios. In a typical summer peak season, the average daily revenue can reach 15,000 yuan, while during the winter off-season, the average daily revenue may drop to around 5,000 yuan. By using nearest-neighbor aggregate queries, the system can quickly assess the economic benefits of any given scenario, greatly improving the flexibility and real-time nature of the assessment. In a practical application, when the load on the next working day is predicted to reach 90% of the historical peak, the system can provide an economic assessment prediction within 3 minutes based on the analysis results of similar scenarios, providing timely support for operational decisions.

[0078] In the assessment report generation stage, the system first extracts operational data from typical scenario groups, calculates the average daily revenue and annualized investment payback period of the energy storage system, and forms benchmark economic indicators. Then, the system uses Monte Carlo simulation to generate an economic indicator distribution containing 90% confidence intervals. This method can more accurately reflect the uncertainty of economic indicators and provide strong support for risk assessment. Finally, the system analyzes the economic indicator distribution, assesses the revenue composition and risk factors under different scenarios, and outputs a complete assessment report.

[0079] The assessment report typically includes several key sections: an investment payback period analysis to demonstrate the project's economic feasibility, such as an estimated payback period of 4.8 years under a baseline scenario, with a confidence interval of 4.2-5.5 years after considering the impact of extreme scenarios; a revenue composition analysis to reveal the proportion of different revenue sources, such as peak-valley price difference revenue accounting for 62%, demand response revenue accounting for 25%, and ancillary service revenue accounting for 13%; a sensitivity analysis to identify key influencing factors, such as how a change in peak-valley price difference of 0.1 yuan / kWh will lead to an annual revenue change of approximately 15%; and a risk warning to emphasize potential risk factors, such as the reduction in charging and discharging efficiency due to equipment aging being a major risk factor affecting later-stage revenue.

[0080] To enhance report readability, the system automatically generates a wealth of visual charts. For example, heatmaps display revenue distribution across different time periods, radar charts show the proportion of various revenue streams, and waterfall charts illustrate cost-benefit analysis. These intuitive charts help decision-makers quickly grasp the economic characteristics of a project, improving decision-making efficiency. The system also provides scenario simulation functionality, allowing users to adjust key parameters and view their impact on the evaluation results in real time, further enhancing the interactivity and practicality of the assessment.

[0081] Practice has proven that this multi-scenario evaluation method can provide comprehensive and accurate economic analysis results. In multiple real-world projects, the deviation between the evaluation results of this method and the actual operating data is controlled within ±8%, which is significantly higher than the accuracy of traditional evaluation methods, providing reliable guidance for investment decisions and operational optimization of energy storage systems.

[0082] Among them, such as Figure 2 As shown, the S1 data preprocessing and feature extraction specifically include:

[0083] S1.1: Data collection and cleaning of power distribution network operation;

[0084] Data collection and cleaning of distribution network operations is a fundamental step in the economic evaluation of energy storage systems. During the data collection phase, the system primarily gathers three types of key data: first, distribution network load data, including historical load curves with a 15-minute time resolution, reflecting the temporal distribution characteristics of electricity demand; second, electricity price data, including market information such as time-of-use pricing, peak-valley price differences, and ancillary service compensation prices; and third, energy storage equipment parameters, including technical indicators such as charge / discharge efficiency, capacity decay rate, and equipment lifespan. This raw data typically originates from power monitoring systems, power trading platforms, and technical documents provided by equipment manufacturers.

[0085] The data cleaning stage primarily addresses missing values, outliers, and noise in the raw data. For missing values, the system employs different imputation strategies based on data type: load data uses similar-date load curve substitution, electricity price data uses the average of preceding and following time periods, and equipment parameters use standard parameter substitution. Outlier identification utilizes a statistical distribution-based detection method, marking data points deviating more than three standard deviations from the mean as outliers, and correcting them using a windowed moving average. Noise removal employs wavelet transform filtering to preserve the main trend characteristics of the data while removing high-frequency disturbances. The processed data significantly improves quality while maintaining its original time-series characteristics, laying a reliable foundation for subsequent analysis.

[0086] Data standardization is the final step in the data cleaning process, aiming to eliminate differences between data of different dimensions and improve the model's generalization ability. The system uses a maximum-minimum standardization method to map various data types to the [0,1] interval, as shown in the formula.

[0087] X' = ​​(X - Xmin) / (Xmax - Xmin)

[0088] For long-tailed distributed data, a logarithmic transformation is applied before standardization to effectively compress numerical differences. Standardization not only improves the accuracy of subsequent feature extraction but also accelerates matrix operations, with measured computational efficiency improvements of approximately 30%. Through this series of processes, the system ultimately generates a standardized multidimensional dataset, providing a high-quality data source for the feature extraction stage.

[0089] S1.2: Construction of the sliding window feature matrix;

[0090] like Figure 3 As shown, the construction of the sliding window feature matrix in S1.2 specifically includes:

[0091] S1.2.1: The sliding window size is determined by the 24-hour charge and discharge cycle of the energy storage system, and the power distribution network operation data is processed using the sliding window size to obtain multiple overlapping time window segments;

[0092] S1.2.2: Construct a feature matrix based on the data in the multiple overlapping time windows, and perform calculations using the best approximation matrix multiplication algorithm to form an optimized feature matrix.

[0093] The construction of the sliding window feature matrix is ​​one of the core technologies of this method, achieving efficient feature representation through structured processing of time-series data. First, the system determines the sliding window parameters, a crucial step. Based on the physical characteristics of the energy storage system's charge-discharge cycle, the window size (w) is set to 24 hours, a setting that closely matches the daily cycle characteristics of grid load and electricity price. The sliding step size (s) is set to 1 / 4 of the window size, i.e., 6 hours. This overlapping setting ensures data continuity while avoiding redundancy caused by excessive repetition. Practice has shown that this parameter configuration achieves a good balance between computational efficiency and feature representation capability.

[0094] After the window parameters are determined, the system segments the time series. For a time series of total length T, a sliding window technique is used to decompose it into multiple overlapping time window segments, resulting in a total of [(Tw) / s+1] subsamples. Each subsample contains continuous observations within 24 hours, fully preserving the temporal characteristics and periodic variations of the data. For each time window segment, the system constructs a feature matrix M[w×d], where w is the window size and d is the feature dimension, including load data P(t), electricity price data E(t), energy storage parameters such as SOC(t) and efficiency η(t), and environmental factors such as temperature T(t), among other multidimensional information. This matrix representation transforms the time series data into a structured feature space, facilitating subsequent feature extraction and pattern recognition.

[0095] In the process of constructing the feature matrix, the system innovatively applies the best approximation matrix multiplication algorithm for computational optimization. The computational complexity of the traditional matrix multiplication algorithm is O(n^2). 3 However, this method is inefficient when dealing with large-scale data processing. This method employs an improved Strassen algorithm, reducing the computational complexity to O(n^2.807) through matrix block recursive computation. In the specific implementation, the system optimizes the block parameters based on the CPU cache size of the hardware platform, typically choosing powers of 2, such as 64 or 128. Simultaneously, an approximation calculation mechanism is introduced, further improving computation speed by discarding small elements while allowing a maximum error ε = 0.1%. Experiments show that this optimized algorithm reduces computation time by approximately 60% when processing 100,000 rows of data, while keeping the accuracy loss within an acceptable range.

[0096] By constructing a sliding window feature matrix, the system ultimately forms a structured time-series feature representation. This representation not only preserves the temporal correlation of the original data but also integrates information from different dimensions into a unified mathematical framework through the form of a feature matrix, providing a solid foundation for subsequent feature dimensionality reduction and selection. In a practical application of a 10MW energy storage project, this method processes two years of historical data in approximately 20 minutes, improving efficiency by three times compared to traditional methods. Simultaneously, the completeness of the feature representation improves the accuracy of subsequent economic assessments by approximately 15%.

[0097] S1.3: Key Feature Identification and Extraction.

[0098] Specifically, S1.3 key feature identification and extraction includes:

[0099] S1.3.1: Calculate the correlation coefficient matrix between each feature using the optimized feature matrix, identify and merge features with a correlation coefficient greater than 0.8 based on the feature correlation threshold, and obtain a preliminary selected feature set;

[0100] S1.3.2: Input the pre-screened feature set into the principal component analysis module, select the principal components with a cumulative contribution rate of 85%, and form the dimensionality-reduced feature matrix;

[0101] S1.3.3: Using the dimensionality-reduced feature matrix, the XGBoost feature importance scoring method is applied to select the top 30% of features by importance as the key features for the economic evaluation of the energy storage system.

[0102] The key feature identification and extraction stage employs a multi-level screening strategy to extract the most influential subset of key features for economic evaluation from the high-dimensional feature space. First, the system performs correlation analysis on the feature matrix, calculating the Pearson correlation coefficient matrix between features. For each element r(i,j) in the matrix, if the absolute value of the correlation coefficient is greater than a preset threshold θ (usually set to 0.8), features i and j are considered to be highly correlated. For highly correlated feature pairs, the system retains the feature with greater information content through information gain comparison, merging or deleting redundant features, thus forming a preliminary feature set. This step effectively reduces the redundancy of the feature space and decreases the dimensionality of subsequent calculations; in practice, it can remove 30%-40% of redundant features.

[0103] The second stage employs Principal Component Analysis (PCA) for feature dimensionality reduction. The system first standardizes the initially selected feature set to ensure consistency in the dimensions of each feature. Then, it calculates the covariance matrix and performs eigenvalue decomposition to obtain eigenvalues ​​and corresponding eigenvectors. The eigenvectors are sorted according to their eigenvalues, and the top k principal components with a cumulative contribution rate of 85% are selected to form the dimensionality-reduced feature matrix D'. PCA dimensionality reduction not only effectively reduces the number of features but also retains the main information of the original data, while eliminating linear correlations between features and improving the generalization ability of subsequent models. In practical applications, this step typically compresses the feature dimension to about 50% of the original dimension, significantly reducing computational complexity.

[0104] The final stage employs an XGBoost-based feature importance scoring method to further select the most influential feature subset. The system uses economic benefit indicators (such as payback period and net present value) as the target variable and the PCA-reduced features as the input variable to train the XGBoost model. By analyzing the contribution of features to the model's predictions, an importance score is calculated for each feature. The system selects the top 30% of features by importance as the final key feature set. This machine learning-based feature selection method can capture the non-linear relationship between features and the target variable, discovering potentially important features that are difficult to identify using traditional methods.

[0105] The entire key feature identification and extraction process employs a two-stage strategy of "coarse screening + fine screening": correlation analysis and PCA dimensionality reduction serve as the coarse screening stage, removing redundant and low-information features; XGBoost feature importance scoring serves as the fine screening stage, retaining the features with the most significant impact on the target. This combined strategy significantly reduces data dimensionality while maintaining feature expressive power, improving the efficiency and accuracy of subsequent modeling. In a real-world project, this method ultimately reduced the original 50-dimensional features to 15 key features, improving computational efficiency by approximately 65%, while maintaining model prediction accuracy above 98%, fully validating the effectiveness of the method. In-depth analysis revealed that these key features are mainly concentrated in several dimensions, including peak-valley electricity price differences, load fluctuation amplitude, and energy storage efficiency degradation rate, providing clear feature directions for subsequent modeling.

[0106] In the economic evaluation system for distributed energy storage in power distribution networks, XGBoost is mainly used for key feature identification and extraction. This is the foundational work for the economic evaluation of energy storage systems and is crucial for improving evaluation accuracy and computational efficiency.

[0107] The construction and training process of the XGBoost model employs a systematic approach. First, the system prepares appropriate training data. The input features are feature matrices processed by PCA dimensionality reduction, while the target variables are economic indicators of the energy storage system, such as payback period, net present value (NPV), or internal rate of return (IRR). These economic indicators directly reflect the economic performance of the energy storage system and are important targets for model learning.

[0108] The parameter configuration of the XGBoost model is crucial. In this system, an appropriate learning rate (typically 0.01-0.1) is usually set to balance learning speed and accuracy; the maximum tree depth is controlled (typically 3-7) to avoid overfitting; minimum child node weights are set to ensure model stability; and a random sampling strategy for samples and features is employed to enhance the model's generalization ability. These parameter configurations enable…

[0109] XGBoost can handle common nonlinear relationships and complex interactions in distribution network data.

[0110] The model training process uses a validation set to monitor performance and employs an early stopping strategy to avoid overfitting. After training, the system calculates feature importance scores. Typical importance evaluation metrics include "gain" (the contribution of a feature to the model's predictive ability), "cover" (the coverage of a feature in the tree), and "frequency" (the frequency of a feature appearing in the tree). The method mentioned in the patent primarily uses "gain" as the evaluation metric, which more accurately reflects the actual impact of features on economic assessment.

[0111] Based on feature importance scores, the system sorts features by importance and then selects the top 30% of features as the final key feature set. This selection ratio is explicitly stated in the patent and is determined based on practical experience, which can minimize feature dimensionality and improve subsequent computational efficiency while maintaining model performance.

[0112] In practical applications, XGBoost model analysis may reveal that peak-valley electricity price differences, energy storage charging and discharging efficiency, daily maximum load fluctuations, and equipment capacity degradation rates are the most influential characteristics. These characteristics are usually directly related to the revenue sources and cost structure of energy storage systems, thus significantly impacting their economic viability. For example, peak-valley electricity price differences directly determine the profit margin for electricity price arbitrage, while charging and discharging efficiency affects the loss costs during energy conversion.

[0113] To further improve the performance of the XGBoost model, the system also employs cross-validation to ensure model stability and may incorporate domain-specific knowledge for feature preprocessing and engineering. These measures collectively guarantee the reliability and effectiveness of the feature extraction results.

[0114] By applying XGBoost, the system significantly reduces the original feature dimensions (typically 50 or more) to approximately 15 key features while maintaining high accuracy of the evaluation model (accuracy remains above 98%). This efficient feature extraction method lays a solid foundation for subsequent economic benefit modeling and dynamic evaluation optimization, and is one of the key factors contributing to the superior performance of the entire evaluation system.

[0115] like Figure 4 As shown,

[0116] The economic benefit modeling of the S2 energy storage system specifically includes:

[0117] S2.1: Construction of Investment Cost Model;

[0118] The construction of the S2.1 investment cost model specifically includes: importing the full life cycle data of energy storage equipment and establishing an investment cost model that includes equipment price and installation cost.

[0119] The construction of an investment cost model is a fundamental step in the economic evaluation of energy storage systems. This model comprehensively considers the entire lifecycle cost of the energy storage system, from planning and design to decommissioning. First, the system imports full lifecycle data of the energy storage equipment, including basic data such as equipment specifications, pricing information, and installation requirements. This data typically comes from quotations, technical manuals, and industry standard documents provided by equipment manufacturers. Considering the specific characteristics of energy storage equipment, the system divides investment costs into two main categories: direct costs and indirect costs. Direct costs mainly include expenditures directly related to energy storage capacity and power, such as energy storage unit costs, power conversion system costs, energy management system costs, and auxiliary equipment costs. Indirect costs include non-equipment expenditures such as engineering construction costs, design fees, commissioning fees, and approval fees.

[0120] In terms of cost calculation, the system uses the Net Present Value (NPV) method to discount future cost flows. Unlike traditional valuation methods that use a fixed discount rate, this model innovatively introduces a time-varying discount rate r(t) to more accurately reflect the time value of money at different times. The time-varying discount rate is constructed based on the government bond yield curve, taking into account both time premium and risk premium. Specifically, the system uses the following formula: NPV = Σ[Ct / (1+r(t))] t ], where Ct represents the cost flow in period t, and r(t) is the discount rate in period t. This method is particularly suitable for evaluating energy storage projects with longer evaluation periods, and can more accurately measure the long-term investment value.

[0121] The model also specifically considers equipment residual value, which is crucial for accurately assessing the economic viability of energy storage projects. Based on equipment lifespan curves and secondary market value, the system establishes a residual value prediction model for energy storage equipment. This model considers both technological and market depreciation, determining the depreciation coefficient through historical data regression analysis. Practice shows that correctly assessing equipment residual value can increase the project's net present value by 5%-10%, significantly impacting the economic evaluation results.

[0122] To enhance the model's applicability, the system also features a parameterized configuration interface, allowing users to adjust key parameters such as equipment unit price, engineering cost rate, and discount rate curve according to specific projects. Through sensitivity analysis, users can quickly understand the impact of parameter changes on investment costs, enabling them to make more informed investment decisions. In a 10MW / 20MWh lithium-ion battery energy storage project, the deviation between the investment cost model's assessment results and the actual investment amount was controlled within ±5%, far exceeding the ±15% deviation level of traditional assessment methods, fully demonstrating the model's accuracy and practicality.

[0123] S2.2: Construction of Operation and Maintenance Cost Model;

[0124] The S2.2 Operation and Maintenance Cost Model Construction specifically includes: processing the operating parameters and maintenance data of energy storage devices, and constructing an operation and maintenance cost model that includes the costs of regular maintenance and fault repair.

[0125] The operation and maintenance cost model construction process fully considers the unique characteristics of energy storage system operation and maintenance. Through a systematic cost classification and prediction method, it achieves accurate assessment of operation and maintenance costs. First, the system processes the operating parameters and maintenance data of the energy storage devices, sourced from equipment monitoring systems, maintenance record systems, and industry experience databases. Unlike traditional assessment methods, this model subdivides operation and maintenance costs into two main categories: fixed operation and maintenance costs and variable operation and maintenance costs, thus more accurately reflecting the compositional characteristics of operation and maintenance costs.

[0126] Fixed operation and maintenance costs refer to recurring expenditures unrelated to the system's operational status, primarily including periodic maintenance fees, labor costs, site rental fees, and insurance premiums. These costs are typically calculated annually and are relatively stable and predictable. The system uses a benchmark costing method to determine the basic operation and maintenance rate based on the equipment's scale and complexity, and adjusts it according to regional economic levels and industry standards. For example, for a 10MW lithium-ion battery energy storage system, the annual fixed operation and maintenance cost rate is typically 2%-3% of the initial equipment investment, while for a flow battery system of the same scale, this rate may increase to 3%-4%, reflecting the difference in maintenance complexity between different technological approaches.

[0127] Variable operation and maintenance costs refer to expenditures directly related to the actual operating state of the system, mainly including charging and discharging loss costs, equipment maintenance costs, and performance degradation costs. These costs are highly dependent on the system's operating strategy and the external environment. The model innovatively introduces a maintenance cost prediction method based on the State of Health (SOH). By establishing a mapping relationship between equipment SOH and failure rate, combined with the Weibull distribution model, the system can predict the probability of equipment failure at different times, thereby assessing maintenance costs. Simultaneously, considering the performance degradation characteristics of energy storage equipment, the model also incorporates capacity degradation compensation costs, calculated using the following formula: Cdeg=C0×(1-η^N), where C0 is the initial capacity value, η is the cycle life degradation coefficient, and N is the number of cycles.

[0128] To improve prediction accuracy, the system employs a data-driven approach to establish an operation and maintenance cost database. By collecting historical operation and maintenance data from different types and scales of energy storage projects, the system uses machine learning algorithms to identify key factors affecting operation and maintenance costs and their weights. The study found that depth of charge and discharge (DOD), ambient temperature, and charge / discharge rate are the three key factors influencing the operation and maintenance costs of lithium battery systems, while electrolyte purity and flow control are key factors for flow battery systems. Based on these findings, the system constructs parameterized operation and maintenance cost models for different technical routes, significantly improving the accuracy and relevance of predictions.

[0129] Another innovation of the model is the introduction of an event-driven predictive maintenance module. By monitoring system operating parameters in real time and identifying potential fault symptoms, the system can schedule maintenance work in advance, avoiding high repair costs and downtime losses caused by serious failures. Practice has shown that this predictive maintenance strategy can reduce operation and maintenance costs by 15%-20% while improving system availability by 3%-5%. In a practical application of a 5MW energy storage system in an industrial park, the deviation between the annual predicted value and actual expenditure of this operation and maintenance cost model was controlled within ±8%, while the deviation of traditional methods is usually around ±20%, fully validating the model's advanced nature and practical value.

[0130] S2.3: Construction of the system revenue model;

[0131] Analyze electricity price arbitrage, demand response, and ancillary services data to generate a revenue model, specifically including:

[0132] S2.3.1: Import historical electricity price data and load data, use the time series revenue forecasting model to calculate and generate the benchmark revenue for electricity price arbitrage;

[0133] S2.3.2: Input the benchmark return of the electricity price arbitrage into the demand response analysis module, and calculate the combined return of superimposed demand response by combining it with the demand response subsidy standard;

[0134] S2.3.3: Using the combined revenue, perform comprehensive calculations based on the ancillary service data to output the revenue model.

[0135] The system's revenue model employs a multi-source revenue aggregation approach to comprehensively evaluate various revenue sources for the energy storage system. First, the system imports historical electricity price and load data, typically sourced from power trading systems, load monitoring systems, and publicly available market information platforms. Electricity price data includes time-of-use pricing, peak-hour pricing, off-peak pricing, and real-time price fluctuation information; load data includes typical daily load curves, seasonal load characteristics, and records of extreme load events. Based on this data, the system uses a time-series revenue forecasting model to calculate and generate a benchmark revenue for electricity price arbitrage.

[0136] Electricity price arbitrage is the most fundamental source of revenue for energy storage systems. Its core principle is to capitalize on the peak-valley electricity price difference, charging during off-peak hours and discharging during peak hours to profit from the price difference. Unlike traditional methods, this model employs a dynamically optimized arbitrage strategy. The system first uses time-series analysis methods, such as ARIMA and Prophet, to predict electricity price trends for the next 24-48 hours. Then, based on the predicted electricity prices and considering the technical constraints of the energy storage system (such as charging / discharging efficiency, power limitations, and SOC range), it optimizes charging / discharging time and power using a mixed-integer linear programming (MILP) algorithm to maximize arbitrage profits. The model also considers the efficiency loss and degradation characteristics of the energy storage system, introducing time-varying efficiency parameters η(t) and capacity parameters C(t) to more accurately calculate the actual arbitrage profits.

[0137] After the benchmark revenue calculation is completed, the system inputs the benchmark revenue from electricity price arbitrage into the demand response analysis module. Combined with the demand response subsidy standard, it calculates the combined revenue from demand response. Demand response refers to the participation of energy storage systems in grid peak shaving and load management, typically yielding revenue from both fixed capacity subsidies and actual response subsidies. The model employs a parameterized calculation method based on the demand response policies and subsidy standards of different regions. The system first assesses the technical feasibility of energy storage systems participating in demand response, including response speed, duration, and dispatchable capacity; then, based on the frequency and depth of historical demand response events, it predicts future response opportunities; finally, it calculates the expected demand response revenue by combining the subsidy standard. Research shows that in regions with well-developed demand response mechanisms, this portion of revenue can account for 20%-30% of the total revenue, representing a significant source of added value.

[0138] The final step in the revenue model is the calculation of ancillary service revenue. The system uses the combined revenue model and performs comprehensive calculations based on ancillary service data to output a complete revenue model. Ancillary services mainly include frequency regulation services, standby services, capacity services, and voltage support. For different types of ancillary services, the model employs differentiated evaluation methods: for frequency regulation services, the available frequency regulation capacity and expected revenue are evaluated based on the power response characteristics of the energy storage system and market demand; for standby services, the system's continuous discharge capacity and standby call probability are considered; for capacity services and voltage support, the evaluation is mainly based on capacity value and system availability.

[0139] A key innovation of the revenue model is its consideration of the mutual exclusion and synergistic relationships among various revenue streams. For example, providing frequency regulation services may consume some energy storage capacity, potentially impacting arbitrage opportunities; while participating in demand response may create synergies with certain types of backup services. The model constructs a constraint matrix to explicitly define the time and resource conflicts among different revenue streams and employs a multi-objective optimization algorithm to maximize total revenue while satisfying technical constraints. In a real-world project, this synergistic optimization method improved total revenue by approximately 12% compared to simple revenue aggregation, fully demonstrating the practical value of this approach.

[0140] S2.4: Integration of comprehensive benefit evaluation models.

[0141] The S2.4 comprehensive benefit evaluation model integration specifically includes: taking the investment cost model, operation and maintenance cost model and revenue model as inputs, integrating the models through a weight adaptive algorithm, and outputting the comprehensive economic benefit evaluation model.

[0142] The integrated benefit assessment model organically combines the investment cost model, operation and maintenance cost model, and revenue model to form a complete economic evaluation system. First, the system uses the three sub-models as input, standardizing the evaluation period and calculation benchmark. Considering the differences in time scales between the different models (investment costs are typically one-time expenditures, while operation and maintenance costs and revenues are periodic flows), the system uses the equivalent annuity method to convert one-time expenditures into equivalent annual costs, ensuring the comparability of the indicators across the models. Simultaneously, to address the differences in the lifespan of energy storage systems using different technologies, the model sets an adjustable evaluation period parameter, typically taking the design life of the energy storage system or the financial evaluation period (e.g., 10 years, 15 years), ensuring the reasonableness of the evaluation results.

[0143] In the model integration process, the system uses an adaptive weight algorithm for model fusion, which is a major innovation of this method. Traditional evaluation methods typically use fixed weights, making it difficult to adapt to the changing needs of different scenarios. This model adopts an adaptive weight mechanism based on a combination of the Analytic Hierarchy Process (AHP) and fuzzy evaluation. The system first defines the hierarchical structure of evaluation objectives, decomposing economic benefit evaluation into three dimensions: cost control, profit maximization, and risk management. Then, based on expert knowledge and historical data, a judgment matrix is ​​established, and the initial weights of each dimension are calculated. Finally, the weight allocation is adjusted through fuzzy logic reasoning, taking into account user preferences and specific project characteristics. This adaptive weight mechanism enables the model to flexibly respond to the evaluation needs of different types of projects. For example, commercial projects may focus more on short-term benefits, while power grid projects may focus more on system stability and long-term benefits.

[0144] After model integration, the system outputs a series of comprehensive economic indicators, including payback period, net present value (NPV), internal rate of return (IRR), and benefit-cost ratio (BCR). Unlike traditional methods, this model not only provides point estimates but also confidence intervals, reflecting the uncertainty of the assessment results more comprehensively. Specifically, the system uses Monte Carlo simulation to randomly perturb key parameters, generating a large number of simulated samples and calculating the probability distribution of economic indicators. For example, for a 10MW energy storage project, the system gives a payback period of 4.8 years and a 90% confidence interval of [4.2 years, 5.5 years]. This interval representation provides richer decision-making information than a single point estimate.

[0145] To enhance the model's practicality, the system also includes a sensitivity analysis module. By performing perturbation analysis on key parameters, it identifies the factors most significantly impacting economic benefits, providing decision-makers with risk assessment references. Typical sensitivity analyses include the impact of electricity price fluctuations, equipment cost changes, and policy subsidy changes. The study found that in most energy storage projects, changes in electricity price policies have the most significant impact on system revenue, with fluctuations reaching ±15%, while the impact of equipment price fluctuations is relatively small, typically within ±5%. These analytical results provide important guidance for developing risk response strategies and optimizing operational plans.

[0146] Another innovation of the comprehensive benefit assessment model is the introduction of scenario simulation functionality. The system pre-sets various typical application scenarios, such as grid-side energy storage, user-side energy storage, and renewable energy-supporting energy storage, each with corresponding parameter configurations and assessment priorities. Users can select appropriate scenario templates based on project characteristics or create new scenarios by customizing parameter combinations. In this way, the model can quickly adapt to the assessment needs of different types of energy storage projects, greatly improving its versatility and practical value. In multiple real-world projects, this comprehensive assessment model has demonstrated excellent performance, with the deviation between the assessment results and actual operating data controlled within ±7%, providing reliable guidance for investment decisions in energy storage projects.

[0147] like Figure 5 As shown,

[0148] S3's dynamic evaluation optimization based on incremental graphs specifically includes:

[0149] S3.1: Incremental graph model construction;

[0150] The construction of the incremental graph model in S3.1 specifically includes: importing the evaluation indicators from the comprehensive economic benefit evaluation model, modeling the evaluation indicators as nodes with attribute sets, establishing edge connections based on the influence relationship between the indicators, and constructing the incremental graph model.

[0151] Incremental graph model construction is a crucial step in transforming a static evaluation model into a dynamic graph structure. The system first imports various evaluation indicators from the comprehensive economic benefit evaluation model, including investment indicators, cost indicators, revenue indicators, and comprehensive economic indicators. These indicators are modeled as nodes with rich attribute sets, each containing at least five types of attributes: numerical attributes (e.g., current value, historical value, predicted value), time attributes (e.g., update frequency, timeliness), reliability attributes (e.g., data quality, prediction accuracy), sensitivity attributes (e.g., change threshold, influence radius), and correlation attributes (e.g., upstream and downstream relationships, influence weight). This multi-dimensional attribute design enables nodes to comprehensively describe the dynamic characteristics of the evaluation indicators.

[0152] After the nodes are defined, the system establishes edge connections based on the influence relationships between indicators, constructing a complete incremental graph model. Edge connections represent the dependencies and influence relationships between nodes, and each edge contains at least three key attributes: influence direction (unidirectional or bidirectional), influence weight (quantitatively describing the intensity of the influence), and time lag characteristics (reflecting the delayed effect of the influence). The system uses statistical methods such as multiple regression analysis and Granger causality tests to mine potential correlations between indicators from historical data and determine the attribute values ​​of the edges. For example, the analysis might find that the influence weight of electricity price changes on the revenue model is 0.85, with a time lag of 1 day; while the influence weight of equipment efficiency changes on operation and maintenance costs is 0.63, with a time lag of 7 days. This data-driven correlation analysis ensures the objectivity and accuracy of the incremental graph structure.

[0153] In terms of storage structure, the system adopts a hybrid approach combining adjacency lists and attribute matrices. Adjacency lists efficiently store the topological relationships between nodes, suitable for handling sparsely connected graph structures; the attribute matrix specifically stores the dynamic characteristics of nodes and edges, facilitating fast access and updates. To further improve data processing efficiency, the system introduces sparse matrix compression storage technology, saving only non-zero elements and their position information, significantly reducing storage space requirements. Experimental results show that for a medium-sized evaluation model containing 100 nodes and approximately 500 edges, this hybrid storage structure saves about 60% of storage space compared to traditional methods, while improving the speed of query and update operations by 2-3 times.

[0154] A key feature of incremental graph models is their support for dynamic topology. As the evaluation process progresses, the system can dynamically adjust the attributes of nodes and edges, and even add new nodes and connections, based on new data discoveries and correlation analysis results. This adaptive structure allows the model to continuously evolve, gradually improving its ability to represent real-world systems. In a practical application of a 10MW / 20MWh energy storage project, the initial incremental graph constructed by the system contained 78 nodes and 342 edges. After three months of operation, it expanded to 93 nodes and 417 edges through an automatic learning mechanism, improving the model accuracy by approximately 12%, fully demonstrating the value of dynamic topology.

[0155] S3.2: Approximate skip neighborhood calculation;

[0156] S3.2 Approximate hop neighborhood calculation specifically includes: processing the incremental graph model, setting hop count parameters of 3 to 5 steps, using the importance probability sampling method to perform approximate hop neighborhood calculation between nodes, and obtaining key influencing factors.

[0157] Approximate hop neighborhood computation is the core algorithm for dynamic optimization of incremental graphs. It identifies key influencing factors in the evaluation model by analyzing the propagation paths of influence between nodes. The system first processes the constructed incremental graph model and sets the hop count parameter k, typically ranging from 3 to 5. The hop count parameter determines the depth and breadth of the influence analysis: smaller k values ​​are computationally efficient but may miss long-range influences; larger k values ​​provide a more comprehensive analysis but significantly increase computational overhead. Practice shows that k=4 usually achieves a good balance between efficiency and comprehensiveness, capturing over 95% of significant influence paths while maintaining acceptable computational complexity.

[0158] For large-scale incremental graphs, calculating the complete k-hop neighborhood can lead to combinatorial explosion, consuming excessive computational resources. To address this issue, the system innovatively introduces an approximate calculation method based on importance probability sampling. This method first defines the propagation probability p(e) of an edge, which is positively correlated with the edge's influence weight w(e): p(e) = w(e)^α, where α is an adjustment parameter (typically ranging from 1.5 to 2.0) used to control the sampling bias. Then, the system employs a random walk strategy, starting from the source node and selecting the next node to visit according to the propagation probability, until k steps are reached or the walk terminates. By repeatedly performing the random walk process, the system generates an approximate k-hop neighborhood of the source node.

[0159] To improve the representativeness of the sampling, the system employs a hierarchical sampling strategy. First, all edges are divided into three levels—high, medium, and low—based on their weights. Then, different sampling frequencies are used for each level to ensure that high-weight edges are adequately sampled while not completely ignoring low-weight edges. Furthermore, the system introduces an early stopping mechanism: if no new significantly influential paths are found after N consecutive sampling iterations (typically set to 100-200), the sampling process is terminated early. This optimization allows the system to reduce computation time by approximately 70% while maintaining over 90% accuracy, significantly improving the algorithm's practicality.

[0160] Based on the calculation results of the approximate jump neighborhood, the system can identify the key factors that have the most significant impact on the evaluation model. The system first calculates the influence centrality of each node, i.e., the weighted frequency of that node in its approximate jump neighborhood. Then, based on the ranking of influence centrality, it determines the set of key influencing nodes. Practice has shown that in most energy storage system evaluation models, the top 10% of nodes typically contribute more than 80% of the total influence. This "Pareto principle" characteristic allows the system to focus optimization efforts on a few key nodes, significantly improving computational efficiency.

[0161] To verify the accuracy of the approximate skip neighborhood calculation, a comparative experiment was conducted. On small- to medium-scale models, both the exact and approximate algorithms were used to calculate skip neighborhoods, and the sets of key influencing factors identified by the two methods were compared. Experimental results show that, with k=4 and 1000 samples, the approximate algorithm achieves a 93.5% consistency with the exact algorithm, but its computation time is only 18% of the latter. This high performance enables the system to quickly respond to parameter changes in real-time application scenarios, providing reliable support for dynamic optimization.

[0162] S3.3: Evaluate the dynamic optimization of the evaluation model.

[0163] The dynamic optimization of the evaluation model specifically includes:

[0164] S3.3.1: Measure the parameter changes of the key influencing factors, identify changes greater than 20% of the parameter standard deviation, and determine the node parameters that need to be updated;

[0165] S3.3.2: Using the node parameters that need to be updated, the update operation is executed sequentially through an incremental calculation strategy and passed to adjacent nodes to generate the updated node parameter state;

[0166] S3.3.3: Using the updated node parameter states, recalculate the overall state of the incremental graph model and output the optimized evaluation model.

[0167] The evaluation model's dynamic optimization is a process of real-time adjustment and updating of the incremental graph model based on approximate skip neighborhood calculations. First, the system measures the parameter changes of key influencing factors and compares them with preset thresholds to determine whether an update operation needs to be triggered. The system uses statistical methods to define dynamic thresholds: the standard deviation σ of historical data is calculated for each parameter, and the change threshold is set to 0.2σ. This means that only when a parameter's change exceeds 20% of its standard deviation is it considered a significant change, triggering the update process. This dynamic threshold mechanism effectively balances sensitivity and stability, avoiding frequent calculations caused by small parameter fluctuations.

[0168] Once the system identifies the node parameters that need updating, it begins incremental update operations. Unlike traditional methods, this system does not recalculate the entire evaluation model; instead, it employs an incremental calculation strategy, updating only the affected nodes and their downstream nodes. Specifically, the system first updates the parameters of the changed source nodes, and then, based on the topology of the incremental graph, propagates the impact layer by layer downstream nodes in a breadth-first order. For each affected node, the system calculates its new state based on the new state of its upstream nodes and the influence relationships between edges. This ordered update propagation ensures computational consistency and efficiency.

[0169] During the update propagation process, the system employs several key optimization techniques. First, a local caching mechanism caches intermediate computation results from each node to avoid redundant calculations. Second, a lazy update strategy delays update operations for low-priority nodes, allowing multiple changes to accumulate before processing them all at once, reducing computational overhead. Furthermore, an impact decay mechanism is introduced: as the propagation distance increases, the impact intensity decays exponentially; propagation terminates when the impact intensity falls below a preset threshold, further improving computational efficiency.

[0170] After completing the incremental update, the system uses the updated node parameter states to recalculate the overall state of the incremental graph model and outputs the optimized evaluation model. This process includes updating all key economic indicators, such as payback period, net present value, and internal rate of return, and generating a new evaluation report. Unlike the previous static evaluation, the dynamically optimized evaluation report not only displays the latest evaluation results but also includes parameter change trend analysis and key influencing factor analysis, providing decision-makers with more comprehensive information support.

[0171] This dynamic optimization method demonstrates significant advantages in practical applications. Taking a regional distribution network energy storage project as an example, in the event of a sudden change in electricity pricing policy, traditional assessment methods require recalculating the entire model, taking approximately 30 minutes; while this method only requires 2-3 minutes to complete the model adjustment, improving computational efficiency by about 90%. Furthermore, through multiple iterative optimizations, the model's predictive accuracy continuously improves; after three months, the prediction deviation for the investment payback period decreased from the initial ±12% to ±5%, fully validating the practical value of the dynamic optimization method.

[0172] The S4 multi-scenario economic assessment and results output specifically include:

[0173] S4.1: Scene library construction;

[0174] The scenario library construction specifically includes: importing historical operating data, constructing regular operating scenarios containing 24-hour load curves and extreme operating scenarios with load fluctuations exceeding 40%, forming an evaluation scenario library.

[0175] Scenario library construction is a fundamental step in multi-scenario economic assessment, aiming to create a set of scenarios covering various operating conditions to provide data support for comprehensive evaluation. The system first imports historical operating data, including at least two years of load records, electricity price data, and time-series information such as the operating status of the energy storage system. Based on this historical data, the system constructs two basic scenarios: normal operating scenarios and extreme operating scenarios. Normal operating scenarios reflect the system's daily operating status, typically including weekday and weekend conditions for the four seasons (spring, summer, autumn, and winter), totaling eight basic scenarios. Each scenario contains complete 24-hour load curves, electricity price data, and environmental parameters, representing the system's typical operating mode under normal conditions.

[0176] Extreme operating scenarios simulate the system's operation under unconventional conditions, including sudden load changes, drastic electricity price fluctuations, and equipment failures. The system defines extreme load scenarios as load fluctuations exceeding 40%, extreme electricity price scenarios as electricity price fluctuations exceeding twice the historical average standard deviation, and extreme failure scenarios as energy storage equipment efficiency degradation exceeding 20%. By analyzing extreme events in historical data, the system can typically identify and construct 10-15 extreme scenarios. These extreme scenarios pose significant challenges to the robustness and risk adaptability of the assessment model and are an indispensable component of the comprehensive evaluation.

[0177] To enhance the representativeness and comprehensiveness of the scenario library, the system employs a scenario generation method based on probability distribution. This method first establishes probability distribution models for various parameters through statistical analysis. For example, by fitting historical data, the system may discover that intraday fluctuations in daily electricity prices follow a log-normal distribution, with a mean peak-to-valley price difference of 0.35 yuan / kWh and a standard deviation of 0.08 yuan / kWh; load fluctuations may conform to a normal or bimodal distribution, with characteristic parameters varying seasonally. Based on these probability distribution models, the system uses Monte Carlo simulation to generate a large number of synthetic scenarios that conform to statistical characteristics, further enriching the scenario library.

[0178] The scene synthesis process considers not only the distribution characteristics of individual parameters but also the correlations between them. The system utilizes multivariate joint distribution models or copula functions to capture the dependencies between parameters, ensuring the rationality of the generated scenes in multidimensional space. For example, high temperatures in summer are often accompanied by high loads and high electricity prices; this correlation is preserved in the scene generation. In this way, the system can generate synthetic scenes that statistically closely match actual operating conditions, with a measured consistency exceeding 85%.

[0179] Maintaining and updating the scenario library is also a crucial function of the system. As new data accumulates, the system regularly (usually monthly or quarterly) updates the scenario library, adjusts probability distribution parameters, and adds newly discovered extreme scenarios to ensure its timeliness and representativeness. In a practical application of a 10MW energy storage project in an industrial park, the system initially built 25 basic scenarios, expanding to 38 scenarios three months later. The coverage rate of the scenario library (the proportion of historical situations it can represent) increased from the initial 87% to 95%, providing a solid foundation for the reliability of the evaluation results.

[0180] S4.2: Nearest Neighbor Scene Aggregation Analysis;

[0181] S4.2 Nearest Neighbor Scene Aggregation Analysis specifically includes: processing the evaluation scene library, performing cluster analysis using a scene similarity calculation method based on Euclidean distance, and generating typical scene groups with different load characteristics.

[0182] Nearest neighbor scene aggregation analysis aims to extract key information from a large number of scenes to form effective evaluation results. The system first processes the evaluation scene library, employing a scene similarity calculation method based on Euclidean distance. For any two scenes si and sj, the similarity calculation formula is: d(si,sj)=Σ(wk×(fik-fjk)) 2 ), where fik and fjk represent the values ​​of the k-th feature in the two scenarios, and wk is the weight coefficient of that feature. Feature weights are determined through sensitivity analysis, with features that have a greater impact on economic evaluation receiving higher weights. Typical high-weight features include peak-valley electricity price differences, peak load levels, and the charging and discharging efficiency of energy storage systems.

[0183] Based on scene similarity, the system performs cluster analysis, grouping similar scenes together to form typical scene groups. The system employs an improved K-means++ algorithm for clustering, which significantly improves the stability and efficiency of clustering by optimizing the selection of initial cluster centers. The number of clusters K is automatically determined using indicators such as the silhouette coefficient and gap statistic, typically between 4 and 7. This data-driven clustering method avoids human intervention, ensuring the objectivity and scientific rigor of the grouping results.

[0184] After typical scenario groups are formed, the system analyzes the characteristic distribution of each group to determine its representative characteristics and economic impact. For example, one scenario group might exhibit the characteristics of "high temperature and high load on summer workdays," while another scenario group might exhibit the characteristics of "high load and low electricity price during the winter heating season." The system calculates the central value and distribution range of the characteristics for each scenario group, forming a statistical profile of that group. These statistical profiles not only help decision-makers understand the economic impact of different operating conditions but also provide a classification basis for subsequent aggregate analysis.

[0185] Based on cluster analysis, the system implements a predictive nearest neighbor aggregation query function. Given a set of future prediction parameters, the system can quickly determine the scenario group to which the parameters belong and predict the corresponding economic indicators based on the historical performance of that group. Specifically, the system uses the KNN (K-Nearest Neighbors) algorithm to calculate the distance between the prediction parameters and the center of each scenario group, selects the N nearest scenarios (usually N=3 or N=5), and generates the prediction result based on a weighted average principle. This method uses similar patterns in historical data to predict future performance, offering advantages such as strong intuitiveness and computational simplicity.

[0186] The system also features a scenario importance assessment function, using sensitivity analysis to identify key scenarios with the greatest impact on overall economics. Specifically, the system quantifies the impact of each scenario by adjusting its weight in the overall assessment and observing the magnitude of changes in economic indicators. Research has found that in typical energy storage projects, approximately 20% of scenarios contribute 80% of economic fluctuations, confirming the "Pareto effect" of scenario impact. Identifying these key scenarios helps decision-makers focus on optimizing the most influential operating conditions and improving resource utilization efficiency.

[0187] Nearest neighbor scenario aggregation analysis has demonstrated significant practical application results. In a regional power grid energy storage project, this method can complete the economic prediction of a new scenario within 3 minutes, with the deviation between the prediction results and actual operating data controlled within ±10%. Compared with the traditional full-scenario recalculation method, the time efficiency is improved by approximately 95%, while the prediction accuracy remains comparable, fully validating the practical value of this method.

[0188] S4.3: Evaluation report generation.

[0189] The evaluation report generation specifically includes:

[0190] S4.3.1: Extract the operating data of the typical scenario group, calculate the average daily revenue and annualized investment payback period of the energy storage system, and form benchmark economic indicators;

[0191] S4.3.2: Using the Monte Carlo simulation method, based on the benchmark economic indicators, generate an economic indicator distribution containing 90% confidence intervals;

[0192] S4.3.3: Analyze the distribution of the economic indicators, assess the revenue composition and risk factors under different scenarios, and output the economic assessment report of the energy storage system.

[0193] The assessment report generation is the final step in the multi-scenario economic evaluation, transforming the analysis results into decision support information. The system first extracts operational data from typical scenario groups and calculates key economic indicators for the energy storage system. For each scenario group, the system calculates benchmark economic indicators such as average daily revenue, annualized payback period, net present value (NPV), and internal rate of return (IRR). The calculation process fully considers the impact of scenario characteristics on various costs and revenues, such as reduced revenue due to decreased equipment efficiency in high-temperature scenarios and increased maintenance costs due to accelerated equipment aging in extreme load scenarios. Through this detailed analysis, the system creates economic profiles under different scenario conditions.

[0194] To reasonably reflect the uncertainty of the assessment results, the system uses the Monte Carlo simulation method to generate a distribution of economic indicators that includes confidence intervals. Specifically, based on the statistical characteristics of historical data, the system randomly perturbs key parameters to generate a large number (typically 10,000) of simulation samples and calculates the corresponding economic indicator values. Then, the distribution characteristics of these indicator values ​​are statistically analyzed to determine the 90% confidence interval. For example, the payback period for an energy storage project under the baseline scenario is 4.8 years, but considering parameter uncertainty, the 90% confidence interval is [4.2 years, 5.5 years]. This interval representation provides more comprehensive risk information than a single-point estimate, helping decision-makers make more robust decisions.

[0195] During the simulation, the system pays particular attention to the impact of extreme scenarios. By increasing the weight of extreme scenarios in the simulation, the system can assess the economic performance under "stress testing." For example, if a major change in electricity pricing policy leads to a 50% reduction in peak-valley price differences, the payback period for the energy storage system may be extended to 7.5 years; if the frequency of extreme weather events increases, leading to a 30% increase in the number of days operating at high temperatures, equipment lifespan may be shortened by 15%, thus affecting long-term economic benefits. This extreme scenario analysis helps decision-makers understand the project's risk tolerance and critical conditions.

[0196] Based on the comprehensive analysis results, the system generates a structured assessment report, which typically includes four core parts: First, an investment payback period analysis, showing the distribution of investment payback periods under different scenarios and assessing the financial feasibility of the project; second, a revenue composition analysis, detailing the proportion and stability of various revenue sources, such as electricity price arbitrage revenue accounting for 62%, demand response revenue accounting for 25%, and ancillary service revenue accounting for 13%; third, a sensitivity analysis, quantifying the impact of changes in key parameters on economics, such as a change of 0.1 yuan / kWh in peak-valley electricity prices leading to an annual revenue change of approximately 15%; and finally, risk warnings, emphasizing potential risk factors and their impact pathways, such as risks associated with changes in electricity price policies, equipment aging, and market competition.

[0197] To enhance report readability and intuitiveness, the system automatically generates various visualization charts. Commonly used chart types include: heatmaps, used to display the distribution of benefits over different time periods; radar charts, used to compare the economic performance of different scenarios from multiple dimensions; waterfall charts, visually displaying the cumulative effect of various costs and benefits; butterfly charts, showing the impact of positive and negative factors on economics; and probability density curves of Monte Carlo simulation results, reflecting the uncertainty distribution of economic indicators. These intuitive charts greatly improve information communication efficiency, helping decision-makers quickly grasp the economic characteristics of projects.

[0198] The system also offers scenario simulation capabilities, allowing users to adjust key parameters online and view their impact on evaluation results in real time. For example, users can simulate changes in electricity pricing policies (such as changes in peak-valley price differences), load characteristics (such as extended peak periods), or equipment parameters (such as reduced charging and discharging efficiency). The system will immediately recalculate and display the adjusted economic indicators. This interactive evaluation method significantly improves the efficiency of decision-making, enabling users to quickly identify key economic influencing factors and optimization directions.

[0199] Practice has proven that this assessment report generation method can provide comprehensive and accurate economic analysis results. In multiple real-world projects, the deviation between the assessment results obtained by this method and subsequent actual operating data is controlled within ±8%, which is significantly higher than the accuracy of traditional assessment methods, providing reliable guidance for investment decisions and operational optimization of energy storage systems.

[0200] Example 3

[0201] In another embodiment, the data preprocessing and feature extraction steps of the present invention mainly use a sliding window technique combined with the best approximation matrix multiplication algorithm to process time-series data. In the distribution network operation data acquisition and cleaning stage, the system collects raw data including load curves, electricity price data, and energy storage device parameters, and performs standardization processing to ensure data quality.

[0202] The construction of the sliding window feature matrix is ​​the core step in this process. In practice, based on the charge-discharge cycle characteristics of the energy storage device, the system typically sets a base window size of 24 hours and uses 1 / 4 of the window size as the sliding step (typically 6 hours) to ensure sufficient data overlap. In this way, the time series T = {t1, t2, ..., tn} is divided into multiple overlapping time window segments. For each time window, the system constructs a matrix M[w×d], where w is the window size and d is the feature dimension, including load data P(t), electricity price data E(t), energy storage parameters SOC(t) and efficiency η(t), as well as environmental factors such as temperature T(t).

[0203] During the construction of the feature matrix, the system employs an improved Strassen algorithm for matrix computation optimization. This algorithm reduces the computational complexity from the traditional O(n log n) by using matrix block processing. 3 The computational efficiency is optimized to O(n^2.807). The choice of block size is optimized based on the CPU cache size, typically taking a power of 2, and fast matrix operations are performed while allowing a maximum error ε = 0.1%. This optimization method is particularly suitable for processing large-scale time-series data in distribution networks, significantly improving computational efficiency.

[0204] The key feature identification and extraction process employs a two-stage strategy of "coarse screening + fine screening." First, the system calculates the Pearson correlation coefficient matrix between features, sets a correlation threshold (typically |θ|>0.8), and identifies and merges highly correlated features. This step, known as "coarse screening," primarily aims to remove redundant features. Next, the system inputs the preliminarily selected feature set into the Principal Component Analysis (PCA) module. Through eigenvalue decomposition, it selects principal components with a cumulative contribution rate of 85%, forming a dimensionality-reduced feature matrix. Finally, the system uses a method based on...

[0205] XGBoost's feature importance scoring method selects the top 30% of features by importance as the final key features. This step is called "fine screening" to ensure that the retained features have a significant impact on the economic evaluation.

[0206] It should be noted that this feature extraction method effectively reduces the feature dimensionality and improves the computational efficiency of the subsequent evaluation model. Taking a distribution network energy storage system as an example, the original feature dimensionality was 50 (including load, electricity price, energy storage parameters, etc.). After feature extraction, 15 key features were retained, improving computational efficiency by about 65%, while maintaining model accuracy above 98%.

[0207] Secondly, the economic benefit modeling step of energy storage systems is a key step in constructing a comprehensive evaluation system. This invention adopts a three-layer sub-model structure (investment cost model, operation and maintenance cost model, and revenue model), and forms a comprehensive evaluation system through model integration.

[0208] In the investment cost model construction phase, the system is modeled based on the concept of the entire life cycle of equipment. Investment costs are divided into two main categories: direct costs and indirect costs. Direct costs include the cost of energy storage units, power conversion systems, and energy management systems; indirect costs include engineering construction costs, design costs, and commissioning costs. The model uses the net present value (NPV) method to discount future costs and considers the residual value of the equipment. By introducing a time-varying discount rate r(t), the model can more accurately reflect the time value of money at different periods.

[0209] The construction of the operation and maintenance cost model fully considers the unique characteristics of energy storage system operation and maintenance. The model divides operation and maintenance costs into fixed and variable costs. Fixed costs include periodic maintenance expenses and labor costs; variable costs are related to the actual operating status of the system and include charging and discharging losses and equipment repair costs. The model innovatively introduces a maintenance cost prediction method based on equipment health status. By analyzing historical operation and maintenance data, it establishes a correlation model between equipment failure probability and maintenance costs, improving the accuracy of cost prediction.

[0210] The system's revenue model employs a multi-source revenue aggregation approach. Revenue sources primarily include electricity price arbitrage revenue, demand response subsidies, and ancillary service revenue. Specifically, the system first imports historical electricity price and load data, then uses a time-series revenue forecasting model to calculate the benchmark revenue from electricity price arbitrage. This benchmark revenue is then input into the demand response analysis module, and combined with subsidy standards to calculate the combined revenue. Finally, ancillary service data is used for comprehensive calculation to form a complete revenue model. This multi-source aggregation approach fully considers all revenue sources of the energy storage system, improving the accuracy of the assessment.

[0211] In the integration process of the comprehensive benefit evaluation model, the system adopted a hierarchical integration architecture design. First, the three sub-models were standardized to unify the evaluation period and calculation benchmark. Then, by designing a weight adaptive algorithm, the weight coefficients of the sub-models were dynamically adjusted according to the importance of each dimension in different scenarios. Finally, the overall economic evaluation indicators of the system were obtained through comprehensive calculation, including key indicators such as investment payback period, net present value, and internal rate of return.

[0212] Furthermore, to improve the reliability of the model, the system also incorporates a sensitivity analysis module. By performing disturbance analysis on key parameters, it identifies the factors that have the most significant impact on economic benefits, providing decision-makers with a risk assessment reference. For example, the analysis revealed that in a certain distribution network energy storage project, changes in electricity price policies have the most significant impact on system revenue, with fluctuations reaching ±15%, while the impact of equipment price fluctuations is relatively small, typically within ±5%.

[0213] Furthermore, the core innovation of this invention is the dynamic evaluation and optimization based on incremental graphs. By converting the static evaluation model into a dynamic incremental graph structure, the real-time optimization capability of the evaluation model is realized.

[0214] In the incremental graph model construction phase, the system first converts the various components of the evaluation model into a graph structure. Nodes represent evaluation metrics (such as investment costs, operating costs, and revenues), and edges represent the relationships between metrics. Each node contains a specific set of attributes, such as numerical range, update frequency, and reliability. This graph structure design allows the model to accurately track the impact path of parameter changes on the overall evaluation results.

[0215] For the specific implementation of the incremental graph, the system adopts a storage structure combining adjacency lists and attribute matrices. The adjacency list stores the topological relationships between nodes, while the attribute matrix stores the dynamic characteristics of the nodes. To improve data processing efficiency, the system employs sparse matrix compression storage technology, significantly reducing storage space requirements. Practice shows that this storage structure can save approximately 60% of storage space compared to traditional methods when processing large-scale evaluation data.

[0216] The computation of approximate k-hop neighborhoods is a crucial step in model optimization. This algorithm identifies key influencing factors by analyzing the propagation paths of influence between nodes. Specifically, the system defines the concept of a k-hop neighborhood, which is a subgraph considering all nodes reachable within k steps. By setting different values ​​for k (typically 3-5), the scope and depth of influence analysis can be controlled. When the graph is large, calculating the entire neighborhood can consume excessive resources. Therefore, the algorithm employs an importance-based sampling method, prioritizing the computation of paths with higher influence weights. Practice shows that this method can reduce computation time by approximately 70% while maintaining over 90% accuracy.

[0217] During the dynamic optimization of the evaluation model, the system adjusts the model parameters in real time based on the calculation results of approximate skip neighborhoods. This process includes three main steps: First, the changes in parameters of key influencing factors are measured, and changes greater than 20% of the parameter standard deviation are identified to determine the node parameters that need to be updated. Second, using the node parameters that need to be updated, the update operation is sequentially executed through an incremental calculation strategy and passed to neighboring nodes to generate the updated node parameter state. Finally, using the updated node parameter state, the overall state of the incremental graph model is recalculated, and the optimized evaluation model is output.

[0218] To ensure the stability of the optimization process, a dynamic threshold mechanism is introduced. When the parameter change is less than a preset threshold, the system delays the update operation to avoid wasting computational resources caused by frequent minor adjustments. The threshold is set based on historical data analysis, typically choosing 20% ​​of the parameter's standard deviation as the baseline value.

[0219] In practical applications, this optimization method has demonstrated significant advantages. Taking a regional distribution network energy storage project as an example, in the event of a sudden change in electricity prices, the traditional model requires recalculating the entire assessment result, which takes about 30 minutes. However, using this method, the model adjustment can be completed in only 2-3 minutes while maintaining the accuracy of the assessment result.

[0220] Finally, the multi-scenario economic evaluation and results output step applies the optimized evaluation model to practical decision support. This step achieves a comprehensive economic evaluation through three stages: scenario library construction, nearest neighbor scenario aggregation analysis, and evaluation report generation.

[0221] In the scenario library construction phase, the system built an evaluation scenario library based on historical operational data, including typical and extreme scenarios. Typical scenarios typically include regular weekday and weekend operating conditions in each season; extreme scenarios include special situations such as sudden load changes, drastic fluctuations in electricity prices, and equipment failures. In particular, the system defines load fluctuations exceeding 40% as extreme operating scenarios, which pose challenges to the robustness of the evaluation model.

[0222] To enhance the usability of the scenario library, the system employs a scenario generation method based on probability distribution. By analyzing the statistical characteristics of historical data, probability distribution models for various parameters are established. For example, the intraday fluctuations in weekday electricity prices may follow a log-normal distribution, with a mean peak-to-valley price difference of 0.35 yuan / kWh and a standard deviation of 0.08 yuan / kWh. Based on these statistical characteristics, the system can automatically generate new scenarios that match actual conditions, thereby enriching the content of the scenario library.

[0223] In the nearest neighbor scenario aggregation analysis, the system defines a scenario similarity index based on Euclidean distance, clustering basic scenarios into typical scenario groups. For example, high-temperature, high-load scenarios on summer weekdays and high-load scenarios during the winter heating season are grouped into the same scenario group due to their similar load characteristics. For each scenario group, the system calculates key economic indicators for the energy storage system, including average daily revenue, peak-valley price difference return, and ancillary service revenue.

[0224] Through aggregation analysis, the system can identify differences in economic performance across various scenarios. In typical peak scenarios, average daily revenue may be significantly higher than during off-peak periods. By using nearest-neighbor aggregation queries, the system can quickly assess the economic benefits of any given scenario, greatly improving the flexibility and real-time nature of the evaluation.

[0225] In the assessment report generation stage, the system first extracts operational data from typical scenario groups, calculates the average daily revenue and annualized investment payback period of the energy storage system, and forms benchmark economic indicators. Then, the system uses Monte Carlo simulation to generate an economic indicator distribution containing 90% confidence intervals; this method can more accurately reflect the uncertainty of economic indicators. Finally, the system analyzes the economic indicator distribution, assesses the revenue composition and risk factors under different scenarios, and outputs a complete assessment report.

[0226] The evaluation report typically includes investment payback period analysis, revenue composition analysis, sensitivity analysis, and risk warnings. The system also automatically generates visual charts, such as heatmaps to show the distribution of revenue over different periods, radar charts to show the proportion of various revenue streams, and waterfall charts to show cost-benefit analysis. These intuitive charts help decision-makers quickly grasp the economic characteristics of the project.

[0227] This application also provides an economic evaluation device for distributed energy storage in power distribution networks, comprising: a data preprocessing and feature extraction module for acquiring power distribution network operation data, establishing a multi-dimensional data matrix using sliding window technology, determining feature correlation thresholds, and extracting key features for economic evaluation of the energy storage system; an economic benefit modeling module for establishing an investment cost model, operation and maintenance cost model, and revenue model of the energy storage system based on the key features, and forming a comprehensive economic benefit evaluation model through standardization processing; a dynamic evaluation and optimization module for converting the comprehensive economic benefit evaluation model into an incremental graph structure, achieving dynamic optimization through approximate skip neighborhood calculation, and obtaining an optimized evaluation model; and a multi-scenario evaluation module for constructing an evaluation scenario library using the optimized evaluation model, processing the multi-scenario evaluation results using the nearest neighbor scenario aggregation analysis method, and generating an economic evaluation report for the energy storage system.

[0228] This application also provides a computer device, the computer device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the above-described method for evaluating the economic efficiency of distributed energy storage in power distribution networks.

[0229] In summary, the economic evaluation method and system for distributed energy storage in distribution networks provided by this invention achieves a comprehensive, accurate, and real-time evaluation of the economic performance of energy storage systems through four steps: data preprocessing and feature extraction, economic benefit modeling of energy storage systems, dynamic evaluation and optimization based on incremental graphs, and multi-scenario economic evaluation. Specifically, the sliding window technique combined with the best approximation matrix multiplication improves data processing efficiency; the three-layer sub-model construction ensures the comprehensiveness of the evaluation; the incremental graph and approximate skip neighborhood algorithm enhance the model's dynamic adaptability; and the nearest neighbor scenario aggregation analysis method improves the effectiveness of multi-scenario evaluation. These innovations collectively constitute an efficient, accurate, and practical evaluation system, providing strong support for the planning and decision-making of energy storage systems in distribution networks.

[0230] This disclosure also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program performs the steps of the economic evaluation method and system method for distributed energy storage in power distribution networks described in the above-described method embodiments. The storage medium can be a volatile or non-volatile computer-readable storage medium.

[0231] In addition, this disclosure also provides a computer program product, which stores a computer program. When the computer program is run by a processor, it executes the steps of the economic evaluation method and system method for distributed energy storage in distribution networks provided in any of the above embodiments of this disclosure. For details, please refer to the above method embodiments, which will not be repeated here.

[0232] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium, which can be a volatile or non-volatile computer-readable storage medium. In another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0233] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices and apparatuses described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this disclosure, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0234] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0235] In addition, the functional units in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0236] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0237] Finally, it should be noted that the above-described embodiments are merely specific implementations of this disclosure, used to illustrate the technical solutions of this disclosure, and not to limit it. The protection scope of this disclosure is not limited thereto. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this disclosure; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this disclosure, and should all be covered within the protection scope of this disclosure. Therefore, the protection scope of this disclosure should be determined by the protection scope of the claims.

Claims

1. A method for evaluating the economic viability of distributed energy storage in power distribution networks, characterized in that, include: Acquire power distribution network operation data, establish a multi-dimensional data matrix using sliding window technology, determine feature correlation thresholds, and extract key features for economic evaluation of energy storage systems. Based on the key characteristics, an investment cost model, an operation and maintenance cost model, and a revenue model for the energy storage system are established, and a comprehensive economic benefit evaluation model is formed after standardization. The comprehensive economic benefit evaluation model is converted into an incremental graph structure, and dynamic optimization is achieved through approximate skip neighborhood calculation to obtain the optimized evaluation model. Using the optimized evaluation model, an evaluation scenario library is constructed, and the evaluation results of multiple scenarios are processed using the nearest neighbor scenario aggregation analysis method to generate an economic evaluation report of the energy storage system.

2. The method according to claim 1, characterized in that, A multi-dimensional data matrix is ​​established using the sliding window technique to extract key features for the economic evaluation of energy storage systems, including: The size of the sliding window is determined by the 24-hour charge and discharge cycle of the energy storage system. The sliding window size is then used to process the power distribution network operation data to obtain multiple overlapping time window segments. A feature matrix is ​​constructed based on the data in the multiple overlapping time windows, and the best approximation matrix multiplication algorithm is used to perform the operation to form an optimized feature matrix. The optimized feature matrix is ​​subjected to dimensionality reduction processing to obtain the key features for the economic evaluation of the energy storage system.

3. The method according to claim 2, characterized in that, The optimized feature matrix is ​​subjected to dimensionality reduction processing to obtain the key features for the economic evaluation of the energy storage system, including: The correlation coefficient matrix between each feature is calculated using the optimized feature matrix. Features with a correlation coefficient greater than 0.8 are identified and merged according to the feature correlation threshold to obtain a preliminary feature set. The preliminary feature set is input into the principal component analysis module, and the principal components with a cumulative contribution rate of 85% are selected to form the dimensionality-reduced feature matrix. Using the reduced feature matrix, the XGBoost feature importance scoring method is applied to select the top 30% of features as the key features for the economic evaluation of the energy storage system.

4. The method according to claim 1, characterized in that, Based on the aforementioned key characteristics, an investment cost model, an operation and maintenance cost model, and a revenue model for the energy storage system are established, forming a comprehensive economic benefit evaluation model, including: Import the full lifecycle data of energy storage equipment to establish an investment cost model that includes equipment price and installation cost; Process the operating parameters and maintenance data of energy storage devices to build an operation and maintenance cost model that includes the costs of regular maintenance and fault repair; Analyze data on electricity price arbitrage, demand response, and ancillary services to generate a revenue model; The investment cost model, operation and maintenance cost model, and revenue model are used as inputs, and the models are integrated through a weighted adaptive algorithm to output the comprehensive economic benefit evaluation model.

5. The method according to claim 4, characterized in that, Analyze electricity price arbitrage, demand response, and ancillary services data to generate revenue models, including: Import historical electricity price data and load data, and use a time series revenue forecasting model to calculate and generate benchmark revenue for electricity price arbitrage; The benchmark return from the electricity price arbitrage is input into the demand response analysis module, and combined with the demand response subsidy standard, the combined return from superimposed demand response is calculated. The combined revenue is used to perform comprehensive calculations based on ancillary service data, and the revenue model is output.

6. The method according to claim 1, characterized in that, The comprehensive economic benefit evaluation model is converted into an incremental graph structure, and dynamic optimization is achieved through approximate skip neighborhood calculation, including: Import the evaluation indicators from the comprehensive economic benefit evaluation model, model the evaluation indicators as nodes with attribute sets, establish edge connections based on the influence relationship between indicators, and construct an incremental graph model. The incremental graph model is processed by setting a hop count parameter of 3 to 5 steps and using an importance probability sampling method to perform approximate hop neighborhood calculation between nodes to obtain key influencing factors. Based on the key influencing factors, the parameters of the incremental graph model are dynamically adjusted to output the optimized evaluation model.

7. The method according to claim 6, characterized in that, Based on the aforementioned key influencing factors, the parameters of the incremental graph model are dynamically adjusted, including: Measure the parameter changes of the key influencing factors, identify changes greater than 20% of the parameter standard deviation, and determine the node parameters that need to be updated; Using the node parameters that need to be updated, the update operation is executed sequentially through an incremental calculation strategy and passed to adjacent nodes to generate the updated node parameter state; Using the updated node parameter states, the overall state of the incremental graph model is recalculated, and the optimized evaluation model is output.

8. The method according to claim 1, characterized in that, Using the optimized evaluation model, an evaluation scenario library is constructed, and the evaluation results of multiple scenarios are processed using the nearest neighbor scenario aggregation analysis method, including: Import historical operating data to construct a library of evaluation scenarios, including regular operating scenarios with 24-hour load curves and extreme operating scenarios with load fluctuations exceeding 40%. The evaluation scenario library is processed, and cluster analysis is performed using a scenario similarity calculation method based on Euclidean distance to generate typical scenario groups with different load characteristics; Using the aforementioned typical scenario group, the economic indicators of the energy storage system are calculated, and multi-scenario evaluation results are output.

9. The method according to claim 8, characterized in that, Using the aforementioned typical scenario group, economic indicators for the energy storage system are calculated, and multi-scenario evaluation results are output, including: Extract the operational data of the typical scenario group, calculate the average daily revenue and annualized investment payback period of the energy storage system, and form benchmark economic indicators; Using the Monte Carlo simulation method, an economic indicator distribution containing a 90% confidence interval is generated based on the benchmark economic indicators. Analyze the distribution of the economic indicators, assess the revenue composition and risk factors under different scenarios, and output the economic evaluation report of the energy storage system.

10. An economic evaluation device for distributed energy storage in power distribution networks, characterized in that, include: The data preprocessing and feature extraction module is used to acquire power distribution network operation data, establish a multi-dimensional data matrix using sliding window technology, determine feature correlation thresholds, and extract key features for the economic evaluation of energy storage systems. The economic benefit modeling module is used to establish an investment cost model, operation and maintenance cost model and revenue model of the energy storage system based on the key characteristics, and form a comprehensive economic benefit evaluation model after standardization. The dynamic evaluation and optimization module is used to convert the comprehensive economic benefit evaluation model into an incremental graph structure and achieve dynamic optimization through approximate skip neighborhood calculation to obtain the optimized evaluation model. The multi-scenario evaluation module is used to construct an evaluation scenario library using the optimized evaluation model, process the multi-scenario evaluation results using the nearest neighbor scenario aggregation analysis method, and generate an economic evaluation report for the energy storage system.

Images