Pumped storage AGC control strategy evaluation standard method and system

Abstract

The embodiment of the invention discloses a pumped storage AGC control strategy evaluation standard method and system. The method comprises the steps of obtaining a comprehensive index set through a multi-dimensional index extraction algorithm based on operation data of a pumped storage unit and an AGC scheduling instruction; based on the comprehensive index set, constructing a hierarchical relationship among indexes through a phylogenetic tree topological distribution algorithm to obtain an evaluation index tree with a credible topological structure; on the basis of the evaluation index tree, index weight optimization calculation is carried out through a single-index time FPT algorithm, and an objectively quantified weight distribution matrix is obtained; based on the comprehensive index set and the weight distribution matrix, obtaining a comprehensive evaluation score of the AGC control strategy through a weighted comprehensive evaluation algorithm; and based on the comprehensive evaluation score, obtaining an optimal AGC control strategy recommendation result through a strategy sorting and optimization algorithm. According to the method, objective and comprehensive evaluation of the pumped storage AGC control strategy can be realized.

Description

Technical Field

[0001] This disclosure relates to the field of power system regulation and control technology, and in particular to a method and system for evaluating the AGC control strategy of pumped storage power plants, applicable to the comprehensive performance evaluation and optimization of automatic generation control strategies for pumped storage power plants. Background Technology

[0002] Pumped storage AGC (Automatic Generation Control) is an important technical means for power system frequency regulation and load tracking control. As an important frequency regulation resource for the power grid, pumped storage power stations need to respond quickly to grid dispatch commands and flexibly switch between generation and pumping modes to provide bidirectional regulation capabilities for the power grid.

[0003] Traditional AGC control strategy evaluation methods primarily employ single-indicator assessments, such as focusing solely on response time or adjustment deviation. Typical evaluation approaches include frequency deviation assessment based on CPS (Control Performance Standard) and tracking performance evaluation based on ACE (Control Area Deviation). These methods typically utilize statistical analysis to calculate fundamental indicators such as average response time and maximum deviation.

[0004] Currently, the most relevant technology is the comprehensive performance evaluation method for AGC (Automatic Guided Collection) based on multi-objective optimization theory. This technology establishes evaluation indicators encompassing multiple dimensions such as response speed, adjustment accuracy, and stability, and uses the analytic hierarchy process (AHP) or fuzzy comprehensive evaluation method to comprehensively score AGC strategies. The technical principle involves normalizing each performance indicator, determining weight coefficients through expert experience, and finally calculating the comprehensive evaluation score.

[0005] However, existing technologies have significant shortcomings: First, the evaluation index system is not comprehensive enough and lacks consideration for the special operating characteristics of pumped storage units, such as the impact of frequent operating condition switching on equipment lifespan; second, the weight determination method is highly subjective, relying on expert experience and lacking objective quantitative basis, making it difficult to guarantee the consistency and credibility of the evaluation results. Therefore, a more comprehensive and objective evaluation standard method for pumped storage AGC control strategies is needed. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for evaluating pumped storage AGC control strategies, aiming to solve the problems of insufficient comprehensiveness of evaluation index systems and strong subjectivity in weight determination in existing technologies, and to achieve an objective and comprehensive evaluation of pumped storage AGC control strategies.

[0007] To achieve the above objectives, this invention provides a method for evaluating the AGC control strategy of pumped storage hydroelectric power generation, comprising the following steps: Based on the operating data of pumped storage units and AGC scheduling instructions, a comprehensive set of indicators including response speed, regulation accuracy, regulation mileage, energy conversion efficiency, and equipment life impact indicators is obtained through a multi-dimensional indicator extraction algorithm. Based on the comprehensive index set, a hierarchical relationship between the indicators is constructed using the phylogenetic tree topology distribution algorithm to obtain an evaluation index tree with a reliable topological structure; Based on the evaluation index tree, the index weights are optimized and calculated using the single exponential time FPT algorithm to obtain an objectively quantified weight allocation matrix. Based on the comprehensive index set and the weight allocation matrix, the comprehensive evaluation score of the AGC control strategy is obtained through a weighted comprehensive evaluation algorithm. Based on the comprehensive evaluation score, the optimal AGC control strategy recommendation result is obtained through strategy ranking and optimization algorithms.

[0008] In the above method, based on the operating data of pumped storage units and AGC scheduling instructions, a comprehensive set of indicators is obtained through a multi-dimensional indicator extraction algorithm. This set includes indicators such as response speed, regulation accuracy, regulation mileage, energy conversion efficiency, and equipment lifespan impact. Based on the pumped storage unit's operating data, the time difference between the command and the unit's response is analyzed using a response time calculation algorithm to obtain the response speed index. Based on the response speed index and power tracking error data, the deviation statistics between the actual output and the target value are calculated using the adjustment accuracy algorithm to obtain the adjustment accuracy index. Based on the aforementioned adjustment accuracy index and the unit operating condition switching record, the total amount of adjustment actions per unit time is calculated using the mileage accumulation algorithm to obtain the adjustment mileage index. Based on the aforementioned mileage adjustment index and energy consumption monitoring data, the energy conversion ratio under power generation and pumping modes is analyzed through an efficiency calculation algorithm to obtain the energy conversion efficiency index. Based on the energy conversion efficiency index and equipment operating parameters, the wear and tear on key components caused by frequent adjustments is analyzed using a lifespan impact assessment algorithm to obtain equipment lifespan impact indexes, forming the comprehensive index set.

[0009] In the above method, based on the comprehensive index set, a hierarchical relationship between the indicators is constructed using a phylogenetic tree topology distribution algorithm to obtain an evaluation index tree with a reliable topological structure, including: Based on the comprehensive set of indicators, the mutual influence relationship between the indicators is calculated by the indicator correlation analysis algorithm to obtain the correlation matrix; Based on the correlation matrix, a hierarchical classification structure among the indicators is established using a phylogenetic tree construction algorithm to obtain an initial topology tree structure; Based on the initial topology tree structure, the rationality and stability of the tree structure are verified by the topology distribution verification algorithm to obtain the credibility evaluation result; Based on the credibility assessment results, the branching relationship of the tree structure is adjusted by a topology optimization algorithm to obtain the evaluation index tree with a credibility topology structure.

[0010] In the above method, based on the evaluation index tree, the index weights are optimized and calculated using the single exponential time (FPT) algorithm to obtain an objectively quantified weight allocation matrix, including: Based on the evaluation index tree, the complex weight allocation problem is transformed into multiple sub-problems through a tree structure decomposition algorithm, resulting in a set of sub-problems. Based on the set of subproblems, each subproblem is parameterized and solved using the single exponential time FPT algorithm to obtain a set of locally optimal weight vectors; Based on the aforementioned set of locally optimal weight vectors, the solutions to each subproblem are integrated using a weight merging algorithm to obtain an initial weight allocation scheme. Based on the initial weight allocation scheme, the objectively quantified weight allocation matrix is ​​obtained by ensuring that the sum of the weights is 1 and the actual constraint conditions are met through weight normalization and constraint optimization algorithms.

[0011] In the above method, based on the comprehensive index set and the weight allocation matrix, a weighted comprehensive evaluation algorithm is used to obtain the comprehensive evaluation score of the AGC control strategy, including: Based on the comprehensive index set and the weight allocation matrix, the index values ​​of different dimensions are converted into a unified evaluation scale through the index standardization algorithm to obtain a standardized index matrix. Based on the standardized index matrix and the weight allocation matrix, the weighted values ​​of each dimension index are calculated using a weighted summation algorithm to obtain a weighted index vector. Based on the weighted index vector, the multi-dimensional evaluation results are synthesized into a single evaluation index through a comprehensive scoring algorithm to obtain the comprehensive evaluation score of the AGC control strategy.

[0012] In the above method, based on the operating data of the pumped storage unit, the response speed index, regulation accuracy index, and regulation mileage index are obtained through response time calculation algorithm, regulation accuracy algorithm, and mileage accumulation algorithm, including: Time-series data extraction is performed on the pumped storage unit's operating data to obtain time-series data containing unit power, frequency deviation, and dispatch instructions; Based on the aforementioned time series data, the time difference between the command being issued and the unit's response is analyzed using a response time calculation algorithm to obtain the response speed index. Based on the response speed index and power tracking error data, the deviation statistics between the actual output and the target value are calculated using the adjustment accuracy algorithm to obtain the adjustment accuracy index. Based on the aforementioned adjustment accuracy index and the unit operating condition switching record, the total amount of adjustment actions per unit time is calculated using the mileage accumulation algorithm to obtain the adjustment mileage index. The response speed index, adjustment accuracy index, and adjustment mileage index are subjected to data verification and outlier processing to obtain the verified index data.

[0013] In the above method, based on the adjusted mileage index and energy consumption monitoring data, energy conversion efficiency index and equipment lifespan impact index are obtained through efficiency calculation algorithm and lifespan impact assessment algorithm, including: Based on the aforementioned mileage adjustment index and energy consumption monitoring data, the energy conversion ratio under the power generation mode is analyzed through an efficiency calculation algorithm to obtain the power generation efficiency index. Based on the power generation efficiency index and pumping mode energy consumption data, the energy conversion ratio under the pumping mode is analyzed by the efficiency calculation algorithm to obtain the pumping efficiency index. Based on the power generation efficiency index and the pumping efficiency index, the energy conversion efficiency index is obtained through a comprehensive efficiency calculation algorithm. Based on the energy conversion efficiency index and equipment operating parameters, the wear degree of key components caused by frequent adjustments is analyzed through a lifespan impact assessment algorithm to obtain the equipment lifespan impact index. The energy conversion efficiency index and the equipment lifespan impact index are normalized to obtain standardized efficiency and lifespan indices.

[0014] In the above method, based on the comprehensive index set, a correlation matrix and an initial topological tree structure are obtained through an index correlation analysis algorithm and a phylogenetic tree construction algorithm, including: Based on the aforementioned comprehensive index set, the linear correlation between each index is calculated using the Pearson correlation coefficient algorithm to obtain the linear correlation coefficient matrix; Based on the linear correlation coefficient matrix, the partial correlation relationship between each index is calculated using a partial correlation analysis algorithm to obtain the partial correlation coefficient matrix; Based on the linear correlation coefficient matrix and the partial correlation coefficient matrix, a correlation matrix is ​​obtained through a correlation synthesis algorithm; Based on the correlation matrix, an unrooted tree structure is constructed using the adjacency algorithm to obtain an initial unrooted topological tree; Based on the initial unrooted topology tree, the root node position is determined by the maximum likelihood estimation algorithm to obtain the initial topology tree structure.

[0015] In the above method, based on the evaluation index tree, a set of sub-problems and a set of locally optimal weight vectors are obtained through a tree structure decomposition algorithm and a single exponential time FPT algorithm, including: Based on the evaluation index tree, the key nodes and branch structures in the tree are identified by the depth-first traversal algorithm to obtain the tree structure analysis results; Based on the tree structure analysis results, the complex weight allocation problem is transformed into multiple low-dimensional sub-problems through a problem decomposition algorithm, resulting in a set of sub-problems. Based on the set of subproblems, a state transition equation is established using a dynamic programming algorithm, and weights are enumerated to obtain the state transition matrix. Based on the state transition matrix, the optimal weight combination is found by using the branch and bound algorithm under the condition of satisfying the normalization constraint, and the optimal solution of each subproblem is obtained. The optimal solutions to each subproblem are processed in parallel and the results are integrated to obtain the set of locally optimal weight vectors.

[0016] Furthermore, this invention also provides a system for evaluating pumped storage AGC control strategies, comprising: The multi-dimensional indicator extraction module is used to obtain a comprehensive set of indicators, including response speed indicators, regulation accuracy indicators, regulation mileage indicators, energy conversion efficiency indicators, and equipment life impact indicators, based on the pumped storage unit operation data and AGC scheduling instructions, through a multi-dimensional indicator extraction algorithm. The phylogenetic tree topology construction module is used to construct hierarchical relationships between indicators based on the comprehensive indicator set and to obtain an evaluation indicator tree with a reliable topological structure through the phylogenetic tree topology distribution algorithm. The weight optimization calculation module is used to perform indicator weight optimization calculation based on the evaluation indicator tree using the single exponential time FPT algorithm to obtain an objectively quantified weight allocation matrix. The weighted comprehensive evaluation module is used to obtain the comprehensive evaluation score of the AGC control strategy based on the comprehensive index set and the weight allocation matrix through a weighted comprehensive evaluation algorithm. The strategy optimization module is used to obtain the optimal AGC control strategy recommendation result based on the comprehensive evaluation score through strategy ranking and optimization algorithms.

[0017] This invention constructs a multi-dimensional quantitative evaluation index system for pumped storage AGC control, comprehensively considering key factors such as response speed, regulation accuracy, regulation mileage, energy conversion efficiency, and equipment lifespan. This effectively solves the problems of insufficient comprehensiveness and lack of consideration for the special operating characteristics of pumped storage units in existing technologies. A phylogenetic tree topology distribution algorithm is introduced to construct hierarchical relationships between evaluation indicators. A reliable index system structure is established through objective data analysis, avoiding the subjectivity of manually setting index relationships. A single exponential time FPT algorithm is used for weight optimization calculation. Objective quantitative weight allocation is achieved through parameterized solution and constraint optimization, overcoming the shortcomings of traditional methods that rely on expert experience and are highly subjective. A standardized comprehensive evaluation method for AGC control strategies is established. Through index standardization and weighted comprehensive evaluation algorithms, objective and comprehensive performance evaluation of different strategies is achieved. A scientific strategy optimization mechanism based on comprehensive evaluation scores is provided. Through strategy ranking and optimization algorithms, a quantitative basis is provided for the optimization selection and parameter tuning of AGC strategies, improving the consistency and reliability of evaluation results. Attached Figure Description

[0018] Figure 1 A flowchart illustrating the method for evaluating the AGC control strategy of pumped storage provided in this embodiment of the invention; Figure 2 This is a flowchart of the multi-dimensional indicator extraction process in an embodiment of the present invention; Figure 3 This is a flowchart of the phylogenetic tree topology construction process in an embodiment of the present invention; Figure 4 This is a flowchart of the weight optimization calculation process in an embodiment of the present invention; Figure 5 This is a flowchart of the weighted comprehensive evaluation process in an embodiment of the present invention; Figure 6 This is a flowchart of the strategy optimization process in an embodiment of the present invention; Figure 7 A schematic diagram of the system structure for the evaluation standard of pumped storage AGC control strategy provided in an embodiment of the present invention. Detailed Implementation

[0019] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0020] like Figure 1 As shown, the present invention provides a method for evaluating the AGC control strategy of pumped storage, comprising the following steps: Step S1: Based on the pumped storage unit operation data and AGC scheduling instructions, a comprehensive set of indicators is obtained through a multi-dimensional indicator extraction algorithm, including response speed indicators, regulation accuracy indicators, regulation mileage indicators, energy conversion efficiency indicators, and equipment life impact indicators.

[0021] like Figure 2 As shown, in step S1, based on the pumped storage unit's operating data and AGC scheduling instructions, a comprehensive set of indicators is obtained through a multi-dimensional indicator extraction algorithm. This set includes indicators such as response speed, regulation accuracy, regulation mileage, energy conversion efficiency, and equipment lifespan impact. Specifically, it includes: Step S1.1: Based on the operating data of the pumped storage unit, the time difference between the command and the unit's response is analyzed by the response time calculation algorithm to obtain the response speed index; For example, time-series data is extracted from the pumped storage unit's operating data to obtain time-series data including unit power, frequency deviation, and dispatch instructions; based on the time-series data, the time difference between the instruction and the unit's response is analyzed using a response time calculation algorithm to obtain a response speed index.

[0022] Step S1.2: Based on the response speed index and power tracking error data, calculate the deviation statistics between the actual output and the target value using the adjustment accuracy algorithm to obtain the adjustment accuracy index; Step S1.3: Based on the adjustment accuracy index and the unit operating condition switching record, the total amount of adjustment actions per unit time is calculated using the mileage accumulation algorithm to obtain the adjustment mileage index; After steps S1.1, S1.2 and S1.3, the response speed index, adjustment accuracy index and adjustment mileage index can be verified and outlier processed to obtain the verified index data.

[0023] Step S1.4: Based on the aforementioned mileage adjustment index and energy consumption monitoring data, analyze the energy conversion ratio under power generation and pumping modes using an efficiency calculation algorithm to obtain the energy conversion efficiency index; For example, based on the adjustment mileage index and energy consumption monitoring data, the energy conversion ratio under the power generation mode is analyzed by an efficiency calculation algorithm to obtain the power generation efficiency index; based on the power generation efficiency index and the energy consumption data under the pumping mode, the energy conversion ratio under the pumping mode is analyzed by an efficiency calculation algorithm to obtain the pumping efficiency index; based on the power generation efficiency index and the pumping efficiency index, the energy conversion efficiency index is obtained by a comprehensive efficiency calculation algorithm.

[0024] Step S1.5: Based on the energy conversion efficiency index and equipment operating parameters, the wear degree of key components caused by frequent adjustments is analyzed through the life impact assessment algorithm to obtain the equipment life impact index and form the comprehensive index set.

[0025] Optionally, the energy conversion efficiency index and the equipment lifespan impact index can be normalized to obtain standardized efficiency and lifespan indices.

[0026] Step S2: Based on the comprehensive index set, construct the hierarchical relationship between the indicators using the phylogenetic tree topology distribution algorithm to obtain an evaluation index tree with a reliable topology structure.

[0027] like Figure 3 As shown, in step S2, based on the comprehensive index set, a hierarchical relationship between the indicators is constructed using a phylogenetic tree topology distribution algorithm to obtain an evaluation index tree with a reliable topological structure, specifically including: Step S2.1: Based on the comprehensive index set, calculate the mutual influence relationship between each index using an index correlation analysis algorithm to obtain the correlation matrix; Optionally, based on the comprehensive index set, the linear correlation between each index is calculated using the Pearson correlation coefficient algorithm to obtain a linear correlation coefficient matrix; based on the linear correlation coefficient matrix, the partial correlation between each index is calculated using a partial correlation analysis algorithm to obtain a partial correlation coefficient matrix; based on the linear correlation coefficient matrix and the partial correlation coefficient matrix, a correlation matrix is ​​obtained using a correlation comprehensive algorithm.

[0028] Step S2.2: Based on the correlation matrix, a hierarchical classification structure among the indicators is established using a phylogenetic tree construction algorithm to obtain an initial topology tree structure; Optionally, based on the correlation matrix, an unrooted tree structure is constructed using the adjacency algorithm to obtain an initial unrooted topological tree; based on the initial unrooted topological tree, the root node position is determined using the maximum likelihood estimation algorithm to obtain the initial topological tree structure.

[0029] Step S2.3: Based on the initial topology tree structure, the rationality and stability of the tree structure are verified by the topology distribution verification algorithm to obtain the credibility assessment result; Step S2.4: Based on the credibility assessment results, adjust the branching relationship of the tree structure using a topology optimization algorithm to obtain the evaluation index tree with a credibility topology structure.

[0030] Step S3: Based on the evaluation index tree, the index weights are optimized and calculated using the single exponential time FPT algorithm to obtain an objectively quantified weight allocation matrix.

[0031] like Figure 4 As shown, in step S3, based on the evaluation index tree, the index weights are optimized and calculated using the single exponential time FPT algorithm to obtain an objectively quantified weight allocation matrix, specifically including: Step S3.1: Based on the evaluation index tree, the complex weight allocation problem is transformed into multiple sub-problems through a tree structure decomposition algorithm to obtain a set of sub-problems; Optionally, based on the evaluation index tree, the key nodes and branch structures in the tree are identified by a depth-first traversal algorithm to obtain the tree structure analysis results; based on the tree structure analysis results, the complex weight allocation problem is transformed into multiple low-dimensional sub-problems by a problem decomposition algorithm to obtain a set of sub-problems.

[0032] Step S3.2: Based on the set of subproblems, each subproblem is solved parametrically using the single exponential time FPT algorithm to obtain a set of locally optimal weight vectors; Optionally, based on the set of subproblems, a state transition equation is established using a dynamic programming algorithm to enumerate weights and obtain a state transition matrix; based on the state transition matrix, an optimal weight combination is found using a branch and bound algorithm under the condition of satisfying normalization constraints to obtain the optimal solution for each subproblem; the optimal solutions for each subproblem are processed in parallel and the results are integrated to obtain the set of locally optimal weight vectors.

[0033] Step S3.3: Based on the set of locally optimal weight vectors, the solutions to each subproblem are integrated using a weight merging algorithm to obtain an initial weight allocation scheme; Step S3.4: Based on the initial weight allocation scheme, the weights are normalized and constrained by an optimization algorithm to ensure that the sum of the weights is 1 and the actual constraints are met, thereby obtaining the objectively quantified weight allocation matrix.

[0034] Step S4: Based on the comprehensive index set and the weight allocation matrix, obtain the comprehensive evaluation score of the AGC control strategy through a weighted comprehensive evaluation algorithm.

[0035] like Figure 5 As shown, in step S4, based on the comprehensive index set and the weight allocation matrix, a weighted comprehensive evaluation algorithm is used to obtain the comprehensive evaluation score of the AGC control strategy, specifically including: Step S4.1: Based on the comprehensive index set and the weight allocation matrix, the index values ​​of different dimensions are converted into a unified evaluation scale through the index standardization algorithm to obtain the standardized index matrix; Step S4.2: Based on the standardized index matrix and the weight allocation matrix, calculate the weighted values ​​of each dimension index using a weighted summation algorithm to obtain a weighted index vector; Step S4.3: Based on the weighted index vector, the multi-dimensional evaluation results are synthesized into a single evaluation index through a comprehensive scoring algorithm to obtain the comprehensive evaluation score of the AGC control strategy.

[0036] Step S5: Based on the comprehensive evaluation score, the optimal AGC control strategy recommendation result is obtained through strategy ranking and optimization algorithms.

[0037] like Figure 6As shown, in step S5, based on the comprehensive evaluation score, the optimal AGC control strategy recommendation result is obtained through strategy ranking and optimization algorithms, specifically including: Step S5.1: Based on the comprehensive evaluation score of the AGC control strategy, the performance of multiple candidate strategies is ranked using a score ranking algorithm to obtain a strategy ranking list; Step S5.2: Based on the strategy ranking list, determine the optimal strategy selection criteria by combining the optimization criterion algorithm with the actual application requirements, and obtain the optimal AGC control strategy recommendation result.

[0038] like Figure 7 As shown, the present invention also provides a system for evaluating the AGC control strategy of pumped storage hydroelectric power generation, comprising: The multi-dimensional indicator extraction module is used to obtain a comprehensive set of indicators, including response speed indicators, regulation accuracy indicators, regulation mileage indicators, energy conversion efficiency indicators, and equipment life impact indicators, based on the pumped storage unit operation data and AGC scheduling instructions, through a multi-dimensional indicator extraction algorithm. The phylogenetic tree topology construction module is used to construct hierarchical relationships between indicators based on the comprehensive indicator set and to obtain an evaluation indicator tree with a reliable topological structure through the phylogenetic tree topology distribution algorithm. The weight optimization calculation module is used to perform indicator weight optimization calculation based on the evaluation indicator tree using the single exponential time FPT algorithm to obtain an objectively quantified weight allocation matrix. The weighted comprehensive evaluation module is used to obtain the comprehensive evaluation score of the AGC control strategy based on the comprehensive index set and the weight allocation matrix through a weighted comprehensive evaluation algorithm. The strategy optimization module is used to obtain the optimal AGC control strategy recommendation result based on the comprehensive evaluation score through strategy ranking and optimization algorithms.

[0039] The following details the specific implementation process of this invention: Example 1: In this embodiment, the process of obtaining a comprehensive set of indicators, including response speed, regulation accuracy, regulation mileage, energy conversion efficiency, and equipment lifespan impact indicators, based on pumped storage unit operation data and AGC scheduling instructions, through a multi-dimensional indicator extraction algorithm, includes: The first step involves extracting time-series data from the pumped-storage unit's operating data to obtain time-series data including unit power, frequency deviation, and dispatch instructions. The unit's operating data for the most recent 30 days is extracted from the SCADA system or historical database, including the actual output power value, AGC target value, frequency deviation value, unit operating status flags, and dispatch instruction reception timestamps for each second. Data preprocessing algorithms are used to remove erroneous data points caused by communication failures or sensor anomalies, and the data is then reformatted according to the time series.

[0040] The second step, based on time-series data, analyzes the time difference between the command and the unit's response using a response time calculation algorithm to obtain the response speed index R. The response time calculation employs an edge detection algorithm, first identifying the step change point of the AGC command, then measuring the time point at which the power output begins to change, and calculating the difference between the two time points. The calculation formula is: R = T response - T command T response T represents the actual response time of the unit. command This refers to the time point when the instruction was issued. For multiple instruction changes, the statistical average of all effective response times over 30 days is calculated as the final response speed indicator, R value.

[0041] The third step involves calculating the deviation statistic between the actual output and the target value based on the response speed index R and the power tracking error data, using a regulation accuracy algorithm to obtain the regulation accuracy index A. The regulation accuracy algorithm employs the root mean square error (RMSE) method, and the calculation formula is as follows: Where P actual P represents the actual output power. target Let n be the target power value and n be the number of sampling points. Simultaneously, considering the influence of the hysteresis component, a time-weighted deviation index is introduced, assigning a smaller weight to the deviation in the initial stage of the response and a larger weight to the deviation during the steady-state period. The overall calculation yields the regulation accuracy index A.

[0042] The fourth step involves using the regulation accuracy index A and the unit's operating condition switching records to calculate the total amount of regulation actions per unit time through a mileage accumulation algorithm, thus obtaining the regulation mileage index M. The mileage accumulation algorithm employs a power change accumulation method, calculated as: M = ∑|P(t+Δt) - P(t)|, where P(t) represents the actual output power at time t, and Δt is the sampling interval. The algorithm pays particular attention to the number of switches between power generation and pumping modes, adding an extra mileage accumulation value for each operating condition transition. Finally, the cumulative regulation amount over 30 days is converted into a daily average regulation mileage value, which serves as the regulation mileage index M.

[0043] The fifth step involves analyzing the energy conversion ratio under the power generation mode using an efficiency calculation algorithm, based on the mileage adjustment index M and energy consumption monitoring data, to obtain the power generation efficiency index. The efficiency calculation formula is: Eg = (P g ·T g ) / (V·H·ρ·g), where P g T represents the average power generated. g Let V be the power generation time, V be the water consumption, H be the effective head, ρ be the water density, and g be the acceleration due to gravity. The relative efficiency index is calculated by comparing the energy conversion efficiency under different AGC control strategies.

[0044] Step 6: Based on the power generation efficiency index and pumping mode energy consumption data, analyze the energy conversion ratio under pumping mode using an efficiency calculation algorithm to obtain the pumping efficiency index. The calculation formula is: Ep = (V·H·ρ·g) / (P p ·T p ), where P p T represents the average pumping power. p The pumping time is used as an example. Similarly, by comparing the pumping efficiency under different control strategies, the relative efficiency index is calculated.

[0045] Step 7: Based on the power generation efficiency index and the pumping efficiency index, the energy conversion efficiency index E is obtained through a comprehensive efficiency calculation algorithm. The comprehensive efficiency calculation considers the operating time ratio of the two modes, and the calculation formula is: E = (E g ·T g +E p ·T p ) / (T g + T p The comprehensive energy conversion efficiency index E value is obtained.

[0046] Step 8: Based on the energy conversion efficiency index E and equipment operating parameters, the wear and tear on key components caused by frequent adjustments is analyzed using a lifespan impact assessment algorithm to obtain the equipment lifespan impact index L. The lifespan impact assessment algorithm is based on an equipment wear model and considers three key factors: the number of start-ups and shutdowns, the load change rate, and the operating time. The calculation formula is: L = w1·N ss +w2·∑|dP / dt| + w3·T, where N ss Let dP / dt be the number of start-stop cycles, dP / dt be the power change rate, T be the operating time, and w1, w2, and w3 be weighting coefficients. The lifespan impact index L is calculated by comparing it with the expected lifespan under rated operating conditions.

[0047] The ninth step is to normalize the response speed index R, adjustment accuracy index A, adjustment mileage index M, energy conversion efficiency index E, and equipment lifespan impact index L to obtain a standardized index set I = {R', A', M', E', L'}, where R', A', M', E', and L' are the normalized index values, all mapped to the interval [0,1].

[0048] Example 2: In this embodiment, the process of constructing a hierarchical relationship between the indicators based on the comprehensive indicator set and obtaining an evaluation indicator tree with a reliable topological structure through a phylogenetic tree topology distribution algorithm includes: The first step is to calculate the linear correlation between the indicators based on the comprehensive indicator set I using the Pearson correlation coefficient algorithm, thus obtaining the linear correlation coefficient matrix CL. For any two indicators Xi and Xj, calculate their Pearson correlation coefficient r(Xj). i ,X j ) = Cov(X i ,X j ) / (σX i ·σX j ), where Cov represents the covariance and σ represents the standard deviation. A complete 5×5 linear correlation coefficient matrix CL is calculated by analyzing daily indicator data over 30 days.

[0049] The second step involves calculating the partial correlation between indicators based on the linear correlation coefficient matrix CL, using a partial correlation analysis algorithm to obtain the partial correlation coefficient matrix CP. Partial correlation analysis is used to eliminate the influence of other indicators and measure the direct correlation between two indicators. For any two indicators X... i and X j Calculate its partial correlation coefficient rp(X) i ,X j |X^k), where X^k represents the remainder of X. i and X j All other indicators besides these are used. The complete 5×5 partial correlation coefficient matrix CP is calculated using the matrix inversion method.

[0050] The third step involves obtaining the correlation matrix C based on the linear correlation coefficient matrix CL and the partial correlation coefficient matrix CP using a correlation synthesis algorithm. This algorithm calculates a weighted average of the linear and partial correlation results using the formula: C = α·CL + (1-α)·CP, where α is the weighting coefficient with a value of 0.6. By combining the results of these two correlation analysis methods, a more accurate correlation description matrix C between the indicators is obtained.

[0051] The fourth step involves constructing an unrooted tree structure based on the correlation matrix C using the adjacency join algorithm, resulting in an initial unrooted topological tree T0'. The adjacency join algorithm first transforms the correlation matrix C into a distance matrix D = 1 - |C|, then iteratively selects the closest pair of indicators to connect, while avoiding loops. In this example, response speed R and regulation accuracy A have the strongest correlation and are connected first; regulation mileage M and the impact of equipment lifespan L have the next strongest correlation and are connected second; energy conversion efficiency E has a relatively strong correlation with regulation accuracy A and is connected third; finally, all connections are integrated to form a complete unrooted tree structure T0'.

[0052] The fifth step involves determining the root node position based on the initial unrooted topology tree T0' using the maximum likelihood estimation algorithm, thus obtaining the initial topology tree structure T0. The algorithm calculates the overall likelihood value of each node as the root node and selects the node with the highest likelihood value as the root node. In this example, the adjustment precision A is determined as the root node, forming a hierarchical tree structure T0 with A as the root.

[0053] Step 6: Based on the initial topological tree structure T0, the rationality and stability of the tree structure are verified using a topological distribution verification algorithm to obtain the credibility assessment result V. The verification algorithm employs the bootstrap resampling method, randomly sampling from the original data to reconstruct the tree structure, repeating this process 1000 times, and calculating the support (bootstrap value) of each branch node. Simultaneously, the Robinson-Foulds distance between tree structures constructed using different sampling methods is calculated to assess the consistency of the tree structure. Through comprehensive stability testing and logical rationality analysis, a credibility assessment result V is generated, containing the credibility values ​​of each node and the overall structural rationality score.

[0054] Step 7: Based on the credibility assessment result V, the branching relationships of the tree structure are adjusted using a topology optimization algorithm to obtain an evaluation index tree T with a credible topology. The optimization algorithm rearranges the branches of nodes with a credibility below 70%, trying different connection methods and selecting the structure with the highest likelihood value and logical rationality. For example, if the credibility of the connection between energy conversion efficiency E and regulation accuracy A in the original tree structure is low, the algorithm might try to change E to a connection with regulation mileage M. Through iterative optimization, until the credibility value of all nodes exceeds 80%, the final evaluation index tree T with high credibility and logical rationality is obtained.

[0055] Example 3: In this embodiment, the process of optimizing the index weights based on the evaluation index tree using the single exponential time FPT algorithm to obtain an objectively quantified weight allocation matrix includes: The first step, based on the evaluation index tree T, is to identify key nodes and branch structures in the tree using a depth-first traversal algorithm, obtaining the tree structure analysis result SA. The traversal algorithm starts from the root node and recursively visits all child nodes, recording the depth, number of child nodes, and connection relationships of each node. For example, if the regulation accuracy A is the root node, its direct child nodes may include response speed R and a subtree containing regulation mileage M, energy conversion efficiency E, and equipment lifespan impact L. Through structural analysis, two main subtrees are identified: the performance index tree (containing R and A) and the impact index tree (containing M, E, and L).

[0056] The second step, based on the tree structure analysis result SA, transforms the complex weight allocation problem into multiple low-dimensional subproblems using a problem decomposition algorithm, resulting in a subproblem set SP. The decomposition algorithm, based on the structural characteristics of trees, breaks down the 5-dimensional weight optimization problem into multiple 2-3 dimensional subproblems. In this example, two subproblems are formed: Subproblem 1 focuses on the weight allocation of performance-related indicators {R, A}, and Subproblem 2 focuses on the weight allocation of influence-related indicators {M, E, L}. Simultaneously, the weight ratios of the two subtrees themselves also need to be determined. Through this decomposition, the original high-dimensional combinatorial optimization problem is transformed into multiple solvable low-dimensional problems.

[0057] The third step involves establishing a state transition equation based on the subproblem set SP, and then enumerating the weights using a dynamic programming algorithm to obtain the state transition matrix STM. For each subproblem, the weight values ​​are discretized into 100 equally spaced points between 0 and 1, and a state transition equation is established. Taking subproblem 1 as an example, the state transition equation is: W(i,s) = max{W(i-1,sw} i ) + v i} where i represents the index number, s represents the total weight constraint, and wi and vi represent the weight value and utility value, respectively. By filling the dynamic programming table, the utility value matrix under different weight combinations is obtained, forming the state transition matrix STM.

[0058] The fourth step involves using the State Transition Matrix (STM) to find the optimal weight combination under normalization constraints through the branch and bound algorithm, thereby obtaining the optimal solution (SPS) for each subproblem. The branch and bound algorithm prunes the search space by setting upper and lower bounds, finding the optimal solution for all subproblems that satisfy ∑w. i =1 and w i The algorithm seeks the optimal solution from the weight combinations with ≥0 constraints. For subproblem 1, the algorithm may obtain the optimal weight allocation as {w R =0.4, w A =0.6}; For subproblem 2, the optimal weight allocation may be {w M =0.3, w E =0.4, w L =0.3}; the weight ratio of the two subtrees may be {w性能 =0.6, w 影响 =0.4}. By solving all subproblems, we obtain the set of optimal solutions to the subproblems, SPS.

[0059] The fifth step involves parallel processing and result integration of the optimal solutions to each subproblem, resulting in a set of locally optimal weight vectors, LW. The parallel processing algorithm handles multiple subproblems simultaneously and then organizes the results according to a tree structure. In this example, the internal weights {w} of the performance index tree are... R =0.4, w A =0.6} and the overall weight w of the subtree 性能 Multiplying by 0.6 yields the global weight {w}. R =0.24, w A =0.36}; Similarly, the internal weights {w} of the influence class indicator tree M =0.3, w E =0.4, w L =0.3} and the overall weight w of the subtree 影响 Multiplying by 0.4 yields the global weight {w}. M =0.12, w E =0.16, w L =0.12}. In this way, a complete set of locally optimal weight vectors LW is formed.

[0060] Step 6: Based on the locally optimal weight vector set LW, the solutions to each subproblem are integrated using a weight merging algorithm to obtain the initial weight allocation scheme W0. The merging algorithm combines local weight values ​​according to the hierarchical relationship of the index tree T. In this example, the weight results of two subtrees are merged to obtain the initial weight vector W0 = [w R =0.24, w A =0.36, w M =0.12, w E =0.16, w L =0.12].

[0061] Step 7: Based on the initial weight allocation scheme W0, the weights are normalized and a constraint optimization algorithm is used to ensure that the sum of the weights is 1 and satisfies the actual constraints, thus obtaining an objectively quantified weight allocation matrix W. The normalization algorithm first checks if the sum of the weights equals 1; if not, a proportional adjustment is made. The constraint optimization part handles various practical constraints, including the range of values ​​for individual weights (0.05 ≤ w). i ≤0.5), minimum weight constraints for key indicators (e.g., response speed weight should not be lower than 0.15), etc. The constrained optimization problem is solved using the Lagrange multiplier method to obtain the final weight allocation matrix W = [w R=0.25, w A =0.35, w M =0.15, w E =0.15, w L =0.10], this weight allocation satisfies all constraints and is calculated entirely based on objective data, avoiding the bias of subjective judgment.

[0062] Example 4: In this embodiment, the process of obtaining the comprehensive evaluation score of the AGC control strategy based on the comprehensive index set and the weight allocation matrix through a weighted comprehensive evaluation algorithm includes: The first step, based on the comprehensive indicator set I and the weight allocation matrix W, is to convert the indicator values ​​of different dimensions into a unified evaluation scale using an indicator standardization algorithm, resulting in the standardized indicator matrix NI. For benefit-type indicators (such as regulation accuracy A and energy conversion efficiency E, where higher values ​​are better), a forward standardization formula is used: NI i = (X i - X min ) / (X max - X min For cost-related indicators (such as response speed R and equipment lifespan impact L, the smaller the value, the better), the inverse standardization formula is used: NI i =(X max - X i ) / (X max - X min After standardization, we obtain the matrix NI = [NR, NA, NM, NE, NL], where all elements are dimensionless values ​​in the interval [0,1].

[0063] The second step involves calculating the weighted values ​​of each dimension's indicators based on the standardized indicator matrix NI and the weight allocation matrix W using a weighted summation algorithm, resulting in the weighted indicator vector WV. Each standardized indicator value is then multiplied by its corresponding weight coefficient: WVR = NR × 0.25, WVA = NA × 0.35, WVM = NM × 0.15, WVE = NE × 0.15, WVL = NL × 0.10, yielding the weighted indicator vector WV = [WVR, WVA, WVM, WVE, WVL].

[0064] The third step involves synthesizing the multi-dimensional evaluation results into a single evaluation index based on the weighted index vector WV using a comprehensive scoring algorithm, thus obtaining the comprehensive evaluation score S of the AGC control strategy. The comprehensive score is calculated using a weighted linear combination model: S = WVR + WVA + WVM + WVE + WVL, while introducing a nonlinear correction factor to handle the interaction effects between indicators. The correction formula is: Sfinal = S × (1 + α × synergistic effect coefficient - β × penalty coefficient), where α = 0.1 and β = 0.15. For example, when both response speed and adjustment accuracy are excellent (both greater than 0.8), an additional 5% score is given; when a key indicator performs extremely poorly (such as response speed less than 0.3), 10% of the total score is deducted. Through this comprehensive scoring method, a comprehensive evaluation score S reflecting the overall performance of the AGC control strategy is obtained, with a value ranging from [0,1].

[0065] Example 5: In this embodiment, the process of obtaining the optimal AGC control strategy recommendation result based on the comprehensive evaluation score through strategy ranking and optimization algorithms includes: The first step involves ranking and classifying multiple candidate strategies based on their comprehensive evaluation score S using a score ranking algorithm, resulting in a strategy ranking list SL. Assuming five different AGC control strategies are to be evaluated, their comprehensive evaluation scores are calculated using the above steps: Strategy A scores 0.85, Strategy B scores 0.72, Strategy C scores 0.78, Strategy D scores 0.65, and Strategy E scores 0.70. First, the strategies are ranked from highest to lowest score: Strategy A > Strategy C > Strategy B > Strategy E > Strategy D. Then, a statistical significance test is performed, calculating the p-value of the score difference between any two strategies. The difference between Strategy A and Strategy C is found to be significant (p = 0.02 < 0.05), while the difference between Strategy B and Strategy E is not significant (p = 0.09 > 0.05). Using the K-means clustering algorithm, the five strategies are divided into three levels: Excellent (Strategy A), Good (Strategy C, Strategy B), and Average (Strategy E, Strategy D). Finally, a strategy ranking list SL containing score, ranking, level classification, and statistical significance information is generated.

[0066] The second step involves determining the optimal strategy selection criteria based on the strategy ranking list SL, using an optimization criterion algorithm combined with practical application requirements, to obtain the recommended optimal AGC control strategy. The optimization algorithm first performs hard constraint screening, excluding strategies that do not meet basic technical requirements. For example, strategy D's response time exceeds the specified limit and is excluded. Then, soft constraint evaluation is performed, assessing adaptability based on the specific conditions of the power plant. For example, for older equipment, greater emphasis is placed on equipment lifespan protection, potentially prioritizing strategy B; for newly built power plants, frequency regulation performance is prioritized, potentially prioritizing strategy A. Finally, a comprehensive trade-off is made, using the TOPSIS method to calculate the relative closeness of each strategy to the ideal solution. The calculated closeness values ​​are 0.92 for strategy A, 0.78 for strategy C, and 0.75 for strategy B. Considering all factors, strategy A is ultimately recommended as the optimal AGC control strategy, and a detailed report including the reasons for the recommendation, applicable conditions, implementation suggestions, and potential risks is provided.

[0067] This invention constructs a multi-dimensional quantitative evaluation index system for pumped storage AGC control, comprehensively considering key factors such as response speed, regulation accuracy, regulation mileage, energy conversion efficiency, and equipment lifespan. This effectively solves the problems of insufficient comprehensiveness and lack of consideration for the special operating characteristics of pumped storage units in existing technologies. A phylogenetic tree topology distribution algorithm is introduced to construct hierarchical relationships between evaluation indicators. A reliable index system structure is established through objective data analysis, avoiding the subjectivity of manually setting index relationships. A single exponential time FPT algorithm is used for weight optimization calculation. Objective quantitative weight allocation is achieved through parameterized solution and constraint optimization, overcoming the shortcomings of traditional methods that rely on expert experience and are highly subjective. A standardized comprehensive evaluation method for AGC control strategies is established. Through index standardization and weighted comprehensive evaluation algorithms, objective and comprehensive performance evaluation of different strategies is achieved. A scientific strategy optimization mechanism based on comprehensive evaluation scores is provided. Through strategy ranking and optimization algorithms, a quantitative basis is provided for the optimization selection and parameter tuning of AGC strategies, improving the consistency and reliability of evaluation results.

[0068] The embodiments described above are merely illustrative of the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that various changes or substitutions can be made to the technical solutions of the present invention without departing from the scope of protection of the present invention.

Claims

1. A method for evaluating the AGC control strategy of pumped storage hydroelectric power generation, characterized in that, include: Based on the operating data of pumped storage units and AGC scheduling instructions, a comprehensive set of indicators including response speed, regulation accuracy, regulation mileage, energy conversion efficiency, and equipment life impact indicators is obtained through a multi-dimensional indicator extraction algorithm. Based on the comprehensive index set, a hierarchical relationship between the indicators is constructed using the phylogenetic tree topology distribution algorithm to obtain an evaluation index tree with a reliable topological structure; Based on the evaluation index tree, the index weights are optimized and calculated using the single exponential time FPT algorithm to obtain an objectively quantified weight allocation matrix. Based on the comprehensive index set and the weight allocation matrix, the comprehensive evaluation score of the AGC control strategy is obtained through a weighted comprehensive evaluation algorithm. Based on the comprehensive evaluation score, the optimal AGC control strategy recommendation result is obtained through strategy ranking and optimization algorithms.

2. The method according to claim 1, characterized in that, Based on pumped storage unit operating data and AGC scheduling instructions, a comprehensive set of indicators is obtained through a multi-dimensional indicator extraction algorithm. This set includes response speed indicators, regulation accuracy indicators, regulation mileage indicators, energy conversion efficiency indicators, and equipment lifespan impact indicators. Based on the pumped storage unit's operating data, the time difference between the command and the unit's response is analyzed using a response time calculation algorithm to obtain the response speed index. Based on the response speed index and power tracking error data, the deviation statistics between the actual output and the target value are calculated using the adjustment accuracy algorithm to obtain the adjustment accuracy index. Based on the aforementioned adjustment accuracy index and the unit operating condition switching record, the total amount of adjustment actions per unit time is calculated using the mileage accumulation algorithm to obtain the adjustment mileage index. Based on the aforementioned mileage adjustment index and energy consumption monitoring data, the energy conversion ratio under power generation and pumping modes is analyzed through an efficiency calculation algorithm to obtain the energy conversion efficiency index. Based on the energy conversion efficiency index and equipment operating parameters, the wear and tear on key components caused by frequent adjustments is analyzed using a lifespan impact assessment algorithm to obtain equipment lifespan impact indexes, forming the comprehensive index set.

3. The method according to claim 2, characterized in that, Based on the aforementioned comprehensive index set, a hierarchical relationship between the indicators is constructed using a phylogenetic tree topology distribution algorithm to obtain an evaluation index tree with a reliable topological structure, including: Based on the comprehensive set of indicators, the mutual influence relationship between the indicators is calculated by the indicator correlation analysis algorithm to obtain the correlation matrix; Based on the correlation matrix, a hierarchical classification structure among the indicators is established using a phylogenetic tree construction algorithm to obtain an initial topology tree structure; Based on the initial topology tree structure, the rationality and stability of the tree structure are verified by the topology distribution verification algorithm to obtain the credibility evaluation result; Based on the credibility assessment results, the branching relationship of the tree structure is adjusted by a topology optimization algorithm to obtain the evaluation index tree with a credibility topology structure.

4. The method according to claim 3, characterized in that, Based on the evaluation index tree, the index weights are optimized and calculated using the single exponential time FPT algorithm to obtain an objectively quantified weight allocation matrix, including: Based on the evaluation index tree, the complex weight allocation problem is transformed into multiple sub-problems through a tree structure decomposition algorithm, resulting in a set of sub-problems. Based on the set of subproblems, each subproblem is parameterized and solved using the single exponential time FPT algorithm to obtain a set of locally optimal weight vectors; Based on the aforementioned set of locally optimal weight vectors, the solutions to each subproblem are integrated using a weight merging algorithm to obtain an initial weight allocation scheme. Based on the initial weight allocation scheme, the objectively quantified weight allocation matrix is ​​obtained by ensuring that the sum of the weights is 1 and the actual constraint conditions are met through weight normalization and constraint optimization algorithms.

5. The method according to claim 4, characterized in that, Based on the comprehensive index set and the weight allocation matrix, a weighted comprehensive evaluation algorithm is used to obtain the comprehensive evaluation score of the AGC control strategy, including: Based on the comprehensive index set and the weight allocation matrix, the index values ​​of different dimensions are converted into a unified evaluation scale through the index standardization algorithm to obtain a standardized index matrix. Based on the standardized index matrix and the weight allocation matrix, the weighted values ​​of each dimension index are calculated using a weighted summation algorithm to obtain a weighted index vector. Based on the weighted index vector, the multi-dimensional evaluation results are synthesized into a single evaluation index through a comprehensive scoring algorithm to obtain the comprehensive evaluation score of the AGC control strategy.

6. The method according to claim 5, characterized in that, Based on the operating data of the pumped storage unit, response speed indicators, regulation accuracy indicators, and regulation mileage indicators are obtained through response time calculation algorithms, regulation accuracy algorithms, and mileage accumulation algorithms, including: Time-series data extraction is performed on the pumped storage unit's operating data to obtain time-series data containing unit power, frequency deviation, and dispatch instructions; Based on the aforementioned time series data, the time difference between the command being issued and the unit's response is analyzed using a response time calculation algorithm to obtain the response speed index. Based on the response speed index and power tracking error data, the deviation statistics between the actual output and the target value are calculated using the adjustment accuracy algorithm to obtain the adjustment accuracy index. Based on the aforementioned adjustment accuracy index and the unit operating condition switching record, the total amount of adjustment actions per unit time is calculated using the mileage accumulation algorithm to obtain the adjustment mileage index. The response speed index, adjustment accuracy index, and adjustment mileage index are subjected to data verification and outlier processing to obtain the verified index data.

7. The method according to claim 6, characterized in that, Based on the aforementioned mileage adjustment index and energy consumption monitoring data, energy conversion efficiency index and equipment lifespan impact index are obtained through efficiency calculation algorithm and lifespan impact assessment algorithm, including: Based on the aforementioned mileage adjustment index and energy consumption monitoring data, the energy conversion ratio under the power generation mode is analyzed through an efficiency calculation algorithm to obtain the power generation efficiency index. Based on the power generation efficiency index and pumping mode energy consumption data, the energy conversion ratio under the pumping mode is analyzed by the efficiency calculation algorithm to obtain the pumping efficiency index. Based on the power generation efficiency index and the pumping efficiency index, the energy conversion efficiency index is obtained through a comprehensive efficiency calculation algorithm. Based on the energy conversion efficiency index and equipment operating parameters, the wear degree of key components caused by frequent adjustments is analyzed through a lifespan impact assessment algorithm to obtain the equipment lifespan impact index. The energy conversion efficiency index and the equipment lifespan impact index are normalized to obtain standardized efficiency and lifespan indices.

8. The method according to claim 7, characterized in that, Based on the aforementioned comprehensive index set, a correlation matrix and an initial topological tree structure are obtained through an index correlation analysis algorithm and a phylogenetic tree construction algorithm, including: Based on the aforementioned comprehensive index set, the linear correlation between each index is calculated using the Pearson correlation coefficient algorithm to obtain the linear correlation coefficient matrix; Based on the linear correlation coefficient matrix, the partial correlation relationship between each index is calculated using a partial correlation analysis algorithm to obtain the partial correlation coefficient matrix; Based on the linear correlation coefficient matrix and the partial correlation coefficient matrix, a correlation matrix is ​​obtained through a correlation synthesis algorithm; Based on the correlation matrix, an unrooted tree structure is constructed using the adjacency algorithm to obtain an initial unrooted topological tree; Based on the initial unrooted topology tree, the root node position is determined by the maximum likelihood estimation algorithm to obtain the initial topology tree structure.

9. The method according to claim 8, characterized in that, Based on the aforementioned evaluation index tree, a set of sub-problems and a set of locally optimal weight vectors are obtained through a tree structure decomposition algorithm and a single exponential time FPT algorithm, including: Based on the evaluation index tree, the key nodes and branch structures in the tree are identified by the depth-first traversal algorithm to obtain the tree structure analysis results; Based on the tree structure analysis results, the complex weight allocation problem is transformed into multiple low-dimensional sub-problems through a problem decomposition algorithm, resulting in a set of sub-problems. Based on the set of subproblems, a state transition equation is established using a dynamic programming algorithm, and weights are enumerated to obtain the state transition matrix. Based on the state transition matrix, the optimal weight combination is found by using the branch and bound algorithm under the condition of satisfying the normalization constraint, and the optimal solution of each subproblem is obtained. The optimal solutions to each subproblem are processed in parallel and the results are integrated to obtain the set of locally optimal weight vectors.

10. A system for evaluating the AGC control strategy of pumped storage hydroelectric power generation, characterized in that, include: The multi-dimensional indicator extraction module is used to obtain a comprehensive set of indicators, including response speed indicators, regulation accuracy indicators, regulation mileage indicators, energy conversion efficiency indicators, and equipment life impact indicators, based on the pumped storage unit operation data and AGC scheduling instructions, through a multi-dimensional indicator extraction algorithm. The phylogenetic tree topology construction module is used to construct hierarchical relationships between indicators based on the comprehensive indicator set and to obtain an evaluation indicator tree with a reliable topological structure through the phylogenetic tree topology distribution algorithm. The weight optimization calculation module is used to perform indicator weight optimization calculation based on the evaluation indicator tree using the single exponential time FPT algorithm to obtain an objectively quantified weight allocation matrix. The weighted comprehensive evaluation module is used to obtain the comprehensive evaluation score of the AGC control strategy based on the comprehensive index set and the weight allocation matrix through a weighted comprehensive evaluation algorithm. The strategy optimization module is used to obtain the optimal AGC control strategy recommendation result based on the comprehensive evaluation score through strategy ranking and optimization algorithms.

Images