A rural power supply station service carrying capacity evaluation and optimization method

Abstract

The application provides a rural power supply station business carrying capacity evaluation and optimization method. It relates to the technical field of electric power. It is used to solve the problem of distorted evaluation results caused by layer-by-layer transmission of errors. It includes: constructing a hierarchical index library, with the lowest level being specific quantifiable indicators; obtaining the structural weight, subjective weight and objective weight of each quantifiable indicator in the lowest level; determining the comprehensive weight of each indicator from the three weights through game theory, weighting the data of each quantifiable indicator in the lowest level based on the comprehensive weight to obtain the evaluation value of each indicator in the intermediate level; iterating layer by layer until the evaluation value of each indicator in the highest level is obtained; obtaining the business carrying capacity evaluation value based on the evaluation value of each indicator in the highest level; and selecting a power supply optimization strategy that matches the business carrying capacity evaluation value. In each layer of transmission, the method corrects the input bias of the current layer and blocks the transmission of errors to the upper layer through game theory.

Description

Technical Field

[0001] This invention relates to the field of power technology, and in particular to a method for assessing and optimizing the business carrying capacity of rural power supply stations. Background Technology

[0002] The business carrying capacity of rural power supply stations refers to the comprehensive support capacity of power supply stations based on existing power grid facilities, personnel allocation, and service capabilities, in order to meet the regular electricity demand in the region, as well as to cope with the electricity demand of various types of electricity load growth or new business formats such as cold chain logistics and rural tourism. Assessing and optimizing this capacity can accurately identify the shortcomings of power supply services, improve the adaptability of the power grid and the quality of services, and provide stable and efficient energy security for the development of new rural business formats.

[0003] The existing evaluation model adopts a linear aggregation paradigm of indicators-weights-scores. That is, by constructing a hierarchical static indicator system, the actual operational data collected at the bottom level is standardized and then weighted and summed according to preset fixed weights, and finally reduced to a single comprehensive evaluation score, thereby realizing a hierarchical leap from micro-level business data to macro-level carrying capacity assessment.

[0004] However, this process carries the risk that small errors caused by inaccurate weighting can accumulate and amplify during hierarchical transmission. Such errors, when propagated layer by layer to the macro-assessment, may lead to systematic distortion of the assessment results, causing the macro-capacity judgment to deviate from the actual business status of the power supply station, and making the ability to characterize and interpret dynamic and nonlinear business scenarios unreliable. Summary of the Invention

[0005] Based on this, it is necessary to provide a method for assessing and optimizing the business carrying capacity of rural power supply stations to address the above-mentioned technical problems. This method aims to solve the shortcomings of the linear weighted assessment model based on a static index system in suppressing cascading errors in the existing technology. It also aims to realize an error circuit breaker mechanism based on a hierarchical index structure, thereby achieving a synergistic improvement in assessment accuracy and scenario credibility.

[0006] The following technical solution is adopted in this specification: This manual provides a method for assessing and optimizing the service capacity of rural power supply stations, including: Construct a hierarchical indicator library, including a top level, a bottom level, and at least one intermediate level between the top and bottom levels; wherein, the top level includes indicators of rural power supply support capacity, rural power consumption capacity, and rural development regulation capacity; the bottom level includes quantifiable indicators corresponding to rural power supply support capacity, rural power consumption capacity, and rural development regulation capacity, respectively. Obtain the structural weight, objective weight, and subjective weight of each quantifiable indicator in the lowest level; among them, the structural weight is used to quantify the structural importance of each indicator in the hierarchical indicator library; the objective weight is used to quantify the objective importance of each indicator based on the actual operation data of the power supply station; and the subjective weight is used to quantify the subjective importance of each indicator based on expert experience judgment. The comprehensive weight of each quantifiable indicator is determined from structural weight, objective weight and subjective weight by game theory, and the operational quantitative data of each quantifiable indicator is weighted and integrated based on the comprehensive weight to obtain the evaluation value of the intermediate level indicator to which the lowest level belongs. Based on the evaluation values ​​of intermediate-level indicators, the same game-theoretic weighting method as the lowest level is used to iterate and aggregate upwards level by level until the evaluation values ​​of each indicator in the highest level are obtained. Based on the evaluation values ​​of each indicator in the highest level, the business carrying capacity assessment value of the target rural power supply station is obtained; a power supply optimization strategy that matches the business carrying capacity assessment value is selected from the preset condition IF-rule THEN strategy knowledge base.

[0007] Furthermore, the methods for obtaining the objective weights of each quantifiable indicator in the lowest level include: The evaluation model is trained by taking the historical operational data of all quantifiable indicators of the target rural power supply station as input and the historical assessment value of the business carrying capacity of the target rural power supply station as output. Based on the completed evaluation model, the contribution of each quantifiable indicator to the historical evaluation value of business carrying capacity is determined, and key quantifiable indicators that match the target rural power supply station are selected from the quantifiable indicators according to the contribution. The entropy weight method is used to process the operational quantification data of key quantifiable indicators to obtain the objective weights of each key quantifiable indicator.

[0008] Furthermore, the process of training the evaluation model by using historical operational data of all quantifiable indicators of the target rural power supply station as input and historical assessment values ​​of the target rural power supply station's business carrying capacity as output specifically includes: Historical operational data with quantifiable indicators are collected from data sources associated with the target rural power supply stations. After data cleaning, outlier processing, and normalization, standardized time-series data is obtained. Statistical and temporal features of normalized time-series data are extracted, and multi-source data alignment and feature fusion are performed based on a unified spatiotemporal benchmark to generate multimodal fusion features. The multimodal fusion features are labeled according to the preset service carrying capacity level standard, and the training set, validation set and test set are divided. Using random forest as the base learner, the model is trained using the training set, and hyperparameters are optimized and overfitting is suppressed based on the validation set. After the performance is verified on the test set to meet the preset accuracy threshold, the evaluation model is completed.

[0009] Furthermore, the selection of key quantifiable indicators from the quantifiable indicators that match the target rural power supply station based on contribution specifically includes: The interpretability analysis of the trained evaluation model is carried out using the SHAP interpretation model based on game theory. The marginal contribution of each quantifiable indicator to the evaluation value of business carrying capacity is calculated using each quantifiable indicator as input feature. The global importance measure of each quantifiable indicator is calculated based on the marginal contribution, and the importance ranking of the indicators is generated. By combining the importance of built-in features in the evaluation model with global importance metrics, redundant indicators are eliminated, and indicators that have a significant impact on the business carrying capacity evaluation results are retained as key quantifiable indicators.

[0010] Furthermore, the subjective weights of all indicators in the hierarchical indicator library are obtained through the G1 subjective weighting method.

[0011] Furthermore, the structural weights of all indicators in the hierarchical indicator library are obtained through DEMATEL-ANP; specifically including: The influence relationships between indicators are determined using the DEMATEL method: a direct influence matrix between indicators in a hierarchical indicator library is constructed, and the direct influence matrix is ​​normalized to obtain a normalized influence matrix. A comprehensive influence matrix is ​​calculated based on the normalized influence matrix. The influence degree and the degree of being influenced by each indicator are calculated based on the comprehensive influence matrix, and then the centrality and causality of each indicator are determined. The causal relationship and importance of each indicator are analyzed based on the centrality and causality, and key driving indicators and key result indicators at each level are identified. A hierarchical network structure is constructed and weights are calculated based on the ANP method: a hierarchical network structure model of each level of indicators is constructed according to causal relationships, and the dependency and feedback relationships between each indicator are determined; an unweighted hypermatrix is ​​constructed based on the dependency and feedback relationships, and the unweighted hypermatrix is ​​weighted by combining centrality to obtain a weighted hypermatrix; a limit operation is performed on the weighted hypermatrix, and the limit hypermatrix is ​​obtained when the matrix converges; the feature vectors corresponding to each indicator in the limit hypermatrix are extracted as the structural weight values ​​of each indicator. The structural weight values ​​of each level are normalized so that the sum of the weights of all indicators at the same level is 1, thus obtaining the final structural weights of all indicators in the hierarchical indicator library.

[0012] Furthermore, the determination of the comprehensive weight of each quantifiable indicator from structural weights, objective weights, and subjective weights using game theory specifically includes: The structural weight, objective weight, and subjective weight of the same quantifiable indicator are regarded as the strategy sets of the three game participants. Each participant aims to maximize the reflection of its own weight information in the comprehensive weight, while minimizing the deviation of the weights of other participants. Establish a game optimization model: construct three original weight vectors consisting of structural weight, objective weight, and subjective weight, set a comprehensive weight vector, and construct an objective function to minimize the sum of squared deviations between the comprehensive weight and the three original weight vectors; introduce combination coefficients to perform linear weighted aggregation of the three types of original weights, and set the constraint that the sum of the combination coefficients is one and each coefficient is positive. Solve the game optimization model to obtain the optimal combination coefficients that minimize the sum of squared deviations; Based on the optimal combination coefficients, the three original weight vectors are weighted and fused to obtain the preliminary game comprehensive weights of each quantifiable indicator. The preliminary game comprehensive weights are then normalized to obtain the comprehensive weights of each quantifiable indicator.

[0013] Furthermore, the evaluation values ​​based on intermediate-level indicators are iteratively aggregated upwards level by level using the same game-theoretic weighting method as the lowest level, until the evaluation values ​​of each indicator in the highest level are obtained, specifically including: The evaluation value of the intermediate level indicator to which the lowest level belongs is obtained based on the operational quantitative data of quantifiable indicators: in, This indicates that it belongs to the rural power supply support capacity index. The middle level Evaluation values ​​of each indicator; This indicates that it belongs to the rural electricity supply pressure index. The middle level Evaluation values ​​of each indicator; This indicates that it belongs to the indicators of rural development regulation and control. The middle level Evaluation values ​​of each indicator; This indicates that it belongs to the rural power supply support capacity index. The middle level The first indicator Historical quantitative data of each quantifiable indicator; This indicates that it belongs to the rural electricity supply pressure index. The middle level The first indicator Historical quantitative data of each quantifiable indicator; This indicates that it belongs to the indicators of rural development regulation and control. The middle level The first indicator Historical quantitative data of each quantifiable indicator; , and This indicates that the quantifiable indicators belong to the rural power supply support capacity indicators. Rural electricity supply pressure indicators Rural development regulation and control indicators The number of key quantifiable indicators; , and They represent , and The overall weight; The evaluation values ​​of each indicator in the highest level are obtained based on the evaluation values ​​of the intermediate-level indicators: in, Indicators of rural power supply support capacity Evaluation value; Indicators of rural electricity supply pressure Evaluation value; Indicators representing the regulatory capacity for rural development Evaluation value; , and They represent , and The overall weight; , and This indicates that they belong to the rural power supply support capacity indicators respectively. Rural electricity supply pressure indicators Rural development regulation and control indicators The number of key quantifiable indicators.

[0014] Furthermore, the assessment value of the business carrying capacity of the target rural power supply station, obtained based on the evaluation values ​​of each indicator at the highest level, is achieved through the following formula: Where K represents the current assessment value of the business carrying capacity of the target rural power supply station.

[0015] Furthermore, the step of selecting a power supply optimization strategy that matches the service carrying capacity assessment value from a preset conditional IF-rule THEN strategy knowledge base specifically includes: A knowledge base for the IF-rule THEN strategy is constructed. The IF part of the IF-rule THEN strategy knowledge base includes the business carrying capacity assessment value level range, the evaluation value of the rural power supply support capacity index, the evaluation value of the rural electricity consumption capacity index, the evaluation value of the rural development regulation capacity index, and the carrying capacity change trend. The rule THEN part of the IF-rule THEN strategy knowledge base includes the corresponding current situation diagnosis conclusion, cause analysis, risk warning information, and actionable governance measures suggestions. Multiple quantifiable indicators of current operation data are collected from multiple data sources associated with the target rural power supply station. The multiple current operation data are then merged and input into a pre-trained long short-term memory network model to obtain the changing trend of the business carrying capacity assessment value of the target rural power supply station within a future preset assessment period. The changing trend of the business carrying capacity assessment value is combined with the evaluation values ​​of rural power supply support capacity indicators, rural electricity consumption capacity indicators, rural development regulation capacity indicators, and business carrying capacity assessment value level to form a multi-dimensional comprehensive diagnostic condition set. The multi-dimensional comprehensive diagnostic condition set is matched with the conditional IF-rule THEN strategy knowledge base, and a differentiated power supply optimization strategy is generated based on the matching results.

[0016] The above-mentioned technical solutions adopted in this specification can achieve the following beneficial effects: This invention constructs a multi-level indicator library and introduces a game-theoretic weighting mechanism. It deconstructs the macro-level business carrying capacity into a multi-dimensional hierarchical system of support, pressure, and control forces, along with their subordinate quantifiable indicators. When the weights are passed upwards, structural weights, objective weights, and subjective weights within the same level are optimized using game theory to generate adaptive comprehensive weights. This allows the weight configuration of each level's indicators to not only dynamically adjust with the power supply station but also to instantly correct input deviations at that level and prevent errors from propagating to higher levels, forming an error circuit breaker mechanism. This prevents the gradual transmission of micro-level errors to macro-level assessments. Finally, relying on nonlinear fidelity mapping, it ensures the scientific conversion of micro-level operational data to macro-level assessment values. This achieves error isolation and gradual correction within the multi-level indicator system, adaptive optimization of weight configuration across scenarios, and high-fidelity mapping between assessment results and actual business conditions, thereby improving the accuracy, robustness, and scenario adaptability of rural power supply station business carrying capacity assessments. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0018] Figure 1 This is one of the flowcharts illustrating a method for assessing and optimizing the business carrying capacity of rural power supply stations, as provided in this manual. Figure 2 This is the second flowchart illustrating a method for assessing and optimizing the business carrying capacity of rural power supply stations, as provided in this manual. Figure 3 This document provides a schematic diagram illustrating the application environment of a method for assessing and optimizing the business carrying capacity of rural power supply stations. Figure 4 This is a schematic diagram of a random forest unit structure provided in this specification; Figure 5 A schematic diagram of the unit structure of a long short-term memory network provided in this specification; Figure 6 This document presents a schematic diagram of the functional modules of a rural power supply station business capacity assessment and optimization system. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this specification clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments in this specification without creative effort are within the scope of protection of this application.

[0020] The following is combined with Figures 1-6 This invention describes a method for assessing and optimizing the service carrying capacity of rural power supply stations.

[0021] Figure 1 This is one of the flowcharts illustrating a method for assessing and optimizing the business carrying capacity of rural power supply stations, as provided in this specification. Figure 1 As shown, the method includes the following: S1. Construct a hierarchical indicator library, including a top level, a bottom level, and at least one intermediate level between the top and bottom levels; wherein, the top level includes indicators of rural power supply support capacity, rural power consumption capacity, and rural development regulation capacity; the bottom level includes quantifiable indicators corresponding to rural power supply support capacity, rural power consumption capacity, and rural development regulation capacity, respectively.

[0022] S2. Obtain the structural weight, objective weight, and subjective weight of each quantifiable indicator in the lowest level; whereby the structural weight is used to quantify the structural importance of each indicator in the hierarchical indicator library; the objective weight is used to quantify the objective importance of each indicator based on the actual operation data of the power supply station; and the subjective weight is used to quantify the subjective importance of each indicator based on expert experience judgment.

[0023] S3. Using game theory, determine the comprehensive weight of each quantifiable indicator from structural weight, objective weight, and subjective weight, and then weight and fuse the operational quantitative data of each quantifiable indicator based on the comprehensive weight to obtain the evaluation value of the intermediate level indicator to which the lowest level belongs.

[0024] S4. Based on the evaluation values ​​of intermediate-level indicators, the same game-theoretic weighting method as the lowest level is used to iteratively aggregate upwards level by level until the evaluation values ​​of each indicator in the highest level are obtained.

[0025] S5. Based on the evaluation values ​​of each indicator in the highest level, obtain the business carrying capacity assessment value of the target rural power supply station; select the power supply optimization strategy that matches the business carrying capacity assessment value from the preset condition IF-rule THEN strategy knowledge base.

[0026] As can be seen, this invention provides a method for assessing and optimizing the service carrying capacity of rural power supply stations based on dynamic perception and situational analysis. Figure 2 This is the second flowchart illustrating a method for assessing and optimizing the business carrying capacity of rural power supply stations, as provided in this manual. Figure 2 As shown, this invention collects and integrates internal operational data and external multimodal environmental data from rural power supply stations. It then dynamically filters key influencing indicators related to the target business, calculates the current support capacity (S), pressure (P), and control capacity (A) of the power supply station, and the comprehensive evaluation value (K) of its business carrying capacity. Next, it uses an LSTM prediction model to deduce the future trend of the business carrying capacity (K) value, and finally generates differentiated business carrying capacity optimization strategies. By constructing a complete technical system encompassing multimodal data perception and fusion, dynamic indicator diagnosis and management, intelligent carrying capacity assessment, future trend prediction and early warning, and differentiated strategy generation, it achieves accurate assessment and intelligent optimization of the business carrying capacity of rural power supply stations. This invention employs a dynamic key indicator screening mechanism combining random forest and SHAP models, determines indicator weights based on a subjective-objective combined weighting method, and achieves accurate quantification of carrying capacity through a nonlinear calculation model. Furthermore, it uses an LSTM prediction model to deduce future carrying capacity trends and provide risk warnings, ultimately generating differentiated optimization strategies based on a pre-set strategy rule engine. This invention addresses the discrepancies between static indicators for rural power supply and the demands of new business models. The aforementioned models, due to the potential inclusion of redundant interference and fixed weights in preset indicators, lack dynamic adaptability, making them difficult to adapt to the business scenarios of different power supply stations. This results in discrepancies between the evaluation results and the actual business carrying capacity, leading to low evaluation accuracy. The invention achieves full automation from data collection and evaluation calculation to strategy generation, significantly improving the level of precise investment and lean management of rural power grids.

[0027] Based on the above Figure 1In the illustrated embodiment, for example, in S1 above, a hierarchical indicator library is constructed, including a highest level, a lowest level, and at least one intermediate level between the highest and lowest levels. The highest level includes rural power supply support capacity indicators (S), rural electricity consumption pressure indicators (P), and rural development regulation capacity indicators (A). The lowest level includes quantifiable indicators corresponding to rural power supply support capacity, rural electricity consumption pressure, and rural development regulation capacity, respectively. In a specific embodiment, the hierarchical indicator library includes a highest level, a lowest level, and an intermediate level. The highest level includes three primary indicators (support capacity S, pressure P, and regulation capacity A), the intermediate level includes eight secondary indicators, and the lowest level includes 34 tertiary indicators. This library contains 34 detailed indicators, such as employee skill level, average monthly transformer load, 10kV line loss compliance rate, and agricultural, forestry, animal husbandry, and fishery output value, as shown in Table 1. The detailed indicator system is shown in the following table:

[0028] Table 1. Business Capacity Indicators of Rural Power Supply Stations For example, based on S1 above, after constructing the hierarchical indicator library, this embodiment also provides a method for dynamically selecting key impact indicators related to the target business scenario. The method uses the historical operational quantitative data of all quantifiable indicators of the target rural power supply station as input and the historical assessment value of the business carrying capacity of the target rural power supply station as output to train an assessment model. Based on the trained assessment model, the contribution of each quantifiable indicator to the historical assessment value of business carrying capacity is determined, and key quantifiable indicators that match the target rural power supply station are selected from the quantifiable indicators according to the contribution. Figure 4 This is a schematic diagram of a random forest unit structure provided in this specification. The classification process is as follows: Figure 4 As shown. Specifically, it includes the following steps:

[0029] Step 111: To verify the rationality of the indicator classification, this invention selects the Random Forest classification algorithm. The Random Forest algorithm is an ensemble learning algorithm based on decision trees. It constructs multiple decision trees and integrates the prediction results of these decision trees, exhibiting good generalization performance and high classification accuracy. It includes: (1) During the training process, Bootstrap sampling is performed on the original dataset to generate K independent datasets by random sampling with replacement from the original training data.

[0030] (2) For each generated dataset, randomly select different feature subsets from all features. P 1. Feature 1, 2. Feature 2... P k Based on the features, a CART decision tree is constructed for each dataset and its corresponding feature subset, resulting in K different decision trees.

[0031] (3) Form a random forest by K different CART decision trees. Each decision tree independently outputs the prediction result of the ship category. Finally, the prediction results of all decision trees are comprehensively judged by the majority voting algorithm to obtain the final classification result.

[0032] Random forest algorithms have significant advantages in handling high-dimensional data, avoiding overfitting, and being robust to missing and outlier values, thus finding widespread application in numerous fields. However, random forest algorithms are highly dependent on data; biased or incomplete training data may affect model performance. This invention introduces a random forest classification algorithm combined with Gaussian noise into the determination of the business carrying capacity index of rural power supply stations in China, and performs cross-validation on its output results.

[0033] Step 112: To further enhance the interpretability of the model and ensure the robustness of the screening results, the SHAP model is introduced for interpretability analysis. SHAP calculates the marginal contribution of each feature (i.e., indicator) to the model output based on game theory, providing a consistent and quantifiable ranking of influence. Combined with the feature importance ranking provided by random forest and the detailed contribution decomposition provided by SHAP analysis, it can dynamically and evidence-basedly select the set of key indicators most relevant to a specific business scenario from a pre-defined indicator library. This set includes:

[0034] (1) Load the trained random forest carrying capacity assessment model as the interpretation object. Determine the input feature set N={1,2,...,M}, where each feature corresponds to a third-level indicator in the indicator library set as shown in Table 1.

[0035] (2) Randomly select a representative subset of samples from the training data in the unified data resource pool to construct the background dataset. This dataset is used to estimate the baseline prediction level of the model in a "no specific information" state. The expected prediction value of the model on this background dataset is calculated. As a benchmark for SHAP interpretation :

[0036] .

[0037] (3) Call the TreeExplainer interpreter, which is optimized for tree models, and use the TreeSHAP algorithm to efficiently calculate the SHAP value, avoiding direct calculation of combinatorial complexity. For the sample set that needs to be explained (which can be all samples or specific samples), calculate the SHAP value for each sample. Each feature SHAP value This value is approximated by the following formula:

[0038] ; in, It is a feature subset. Representation Model Using only a subset of features Time for sample The predicted value is M, where M is the total number of features and N represents the total feature set.

[0039] (4) To identify key indicators from a global perspective, calculate the global importance score for each feature i. The score is obtained by averaging the absolute values ​​of the SHAP values ​​for this feature across all samples:

[0040] ; Where m is the total number of samples. (According to...) The features are sorted in descending order to generate a global feature importance ranking based on SHAP. This ranking quantifies the average influence of each indicator on the carrying capacity assessment model.

[0041] (5) For any sample Its final predicted bearing capacity value It can be decomposed into: ; In this decomposition: if , then represents the feature This sample had a positive driving effect on bearing capacity prediction, that is, it improved the predicted bearing capacity level or score. If , then represents the feature This sample had a negative inhibitory effect on bearing capacity prediction, that is, it reduced the predicted bearing capacity level or score.

[0042] Based on any of the above embodiments, for example, based on S1 above, this embodiment also provides a data collection method. This method is executed by a computer device, specifically by a computer device such as a terminal or server alone, or by both a terminal and a server. Specifically, it includes the following steps: Step 121: Establish an automated data acquisition channel and extract real-time or near-real-time data from multiple heterogeneous data sources through a preset application programming interface (API) and data bus technology.

[0043] Step 122: Obtain the internal operation data of the rural power supply station. The data source is the power supply station's internal information system. Figure 3 This document provides a schematic diagram illustrating the application environment of a method for assessing and optimizing the business carrying capacity of rural power supply stations. Figure 3As shown, the internal data includes, but is not limited to: SCADA system data, electricity consumption information collection system data, asset management system data, and customer service work order data. SCADA system data includes substation operating status, voltage, and current; electricity consumption information collection system data includes user electricity consumption, average monthly electricity sales, and distribution area line loss; asset management system data includes the configuration of information technology equipment and self-service terminals; and customer service work order data includes annual customer complaints, fault repair work orders, and business expansion installation work orders.

[0044] Step 123: Obtain external multimodal environment data. For example... Figure 3 As shown, external multimodal environmental data includes, but is not limited to: meteorological data, geographic information data, macroeconomic data, demographic data, and data from third-party commercial platforms. Meteorological data includes monthly temperature and other environmental factor data obtained from meteorological department interfaces; geographic information data includes topography, road conditions, and geographic coordinates of the power supply station's jurisdiction; macroeconomic data includes GDP and energy prices; demographic data includes the rural population size change rate and per capita disposable income of rural residents. Third-party commercial platform data includes agricultural product transaction volume on e-commerce platforms and rural logistics order data.

[0045] Step 124 involves employing multimodal data fusion technology to standardize and preprocess the aforementioned heterogeneous data from multiple internal and external sources, eliminating the influence of unit dimensions. Data is then correlated and fused based on unified timestamps and geospatial identifiers, ultimately forming a high-quality unified data resource pool to serve the carrying capacity assessment model. This includes:

[0046] (1) The internal operating data and the external multimodal environment data are cleaned, transformed and normalized in sequence to obtain preprocessed data. By cleaning, transforming and normalizing the internal operating data and the external multimodal environment data, noise, outliers and inconsistencies in the data are eliminated, which provides a guarantee for subsequent feature extraction and fusion processing.

[0047] (2) Feature extraction is performed on the preprocessed data to obtain multimodal feature data. Specifically, meaningful information and features are extracted from the preprocessed data for subsequent fusion and processing. Feature extraction needs to consider the characteristics of the data and the application scenario, and select appropriate feature extraction methods and algorithms. For example, for continuous electrical measurement data such as distribution transformer load rate and line voltage qualification rate, the sliding window method is used to extract statistical features such as mean, variance, peak value, and peak-to-valley difference, and wavelet transform is combined to capture the multi-scale time-frequency features of the load curve to identify periodic power consumption patterns and abnormal fluctuation inflection points. For discrete event data such as equipment fault alarms and work order types, TF-IDF weighting and Word processing are used. The 2Vec embedding technology transforms text descriptions into low-dimensional dense vectors and constructs event time-series graphs to extract frequent fault modes and association rule features. For spatial data such as transformer area geographical coordinates and topological connections, graph neural networks are used to encode the adjacency matrix and node attributes of the power grid physical structure to generate topological embedding features that characterize the spatial coupling of power supply units. For external correlation data such as meteorological temperature and holiday labels, one-hot encoding and periodic sine and cosine transforms are used to extract seasonal fluctuations and special date influence factors. The final output is a multimodal feature set covering electrical operation state, equipment event state, spatial structure state, and external environment state.

[0048] (3) Align the multimodal feature data in time and space to obtain aligned multimodal feature data. Specifically, align the data from different sources in time and space to ensure the consistency of the fused data. For example, in the time dimension, establish a global time axis with a uniform sampling interval of 15 minutes. Use resampling and interpolation algorithms to downsample and align high-frequency measurement data (such as minute-level load curves). Use forward filling and moving average to extend the time series for low-frequency business data (such as monthly maintenance records). In the spatial dimension, establish a unified spatial reference system based on the GIS geocoding of the power supply station's jurisdiction. Map the scattered line monitoring points, distribution transformer terminals and user meter data to standard grid cells according to the topological hierarchy. Eliminate the coordinate system differences of the multi-source positioning system through coordinate transformation and projection correction. Finally, achieve accurate registration of electrical measurement features, event semantic features and spatial topological features under a unified spatiotemporal reference.

[0049] (4) The aligned multimodal feature data are fused to obtain multimodal fused feature data. Taking into account the features of different data sources and the fusion objective, a reasonable fusion model is designed and optimized. The fusion models that can be selected include weighted average, neural network, support vector machine, etc.

[0050] (5) The server uses a method for assessing and optimizing the business carrying capacity of rural power supply stations based on dynamic perception and situational analysis to generate differentiated operation strategies for power supply stations.

[0051] Based on any of the above embodiments, for example, in S2 above, the structural weight characterizes the structural importance of each indicator in the hierarchical indicator library, that is, the theoretical weight determined after comprehensively considering the causal relationship, mutual influence strength and network dependency structure between indicators, reflecting the systematic contribution of the indicators in the rural power supply station business carrying capacity assessment system, which is achieved through DEMATEL-ANP. Specifically, it includes the following steps:

[0052] Step 211: Determine the influence relationships between indicators based on the DEMATEL method: Construct a direct influence matrix among the indicators in the hierarchical indicator library; normalize the direct influence matrix to obtain a normalized influence matrix; calculate the comprehensive influence matrix based on the normalized influence matrix; calculate the influence degree and affected degree of each indicator according to the comprehensive influence matrix, and then determine the centrality and causality of each indicator; analyze the causal relationship and importance of each indicator based on the centrality and causality, and identify the key driving indicators and key result indicators at each level. This includes:

[0053] (1) After determining the set of evaluation indicators, the DEMATEL questionnaire was designed and the influence between indicators was assessed using a 0-4 five-point scale, as shown in Table 2.

[0054] Table 2. G1 Assignment Method Assignment Standard Table (2) Based on the DEMATEL analysis results, construct the ANP network structure; design an ANP questionnaire, including three types of comparisons: relative importance comparison among indicators, mutual influence comparison among indicators, and importance comparison among clusters; collect expert judgment data and calculate ANP according to the formula.

[0055] (3) Analyze the weight results of DEMATEL-ANP: Analyze the characteristics of high weight indicators, compare the relative relationship of the weights of each indicator, and verify whether the weights conform to the domain knowledge; conduct sensitivity analysis and method comparison verification.

[0056] Step 212: Construct a network hierarchy and calculate weights based on the ANP method: Construct a network hierarchy model for each level of indicators based on the causal relationships, and determine the dependency and feedback relationships between each indicator; construct an unweighted hypermatrix based on the dependency and feedback relationships, and weight the unweighted hypermatrix by combining it with the centrality to obtain a weighted hypermatrix; perform a limit operation on the weighted hypermatrix, and obtain a limit hypermatrix when the matrix converges; extract the feature vectors corresponding to each indicator in the limit hypermatrix as the structural weight values ​​of each indicator. This includes:

[0057] (1) Clarify the influence relationships between the primary indicators. If the value of primary indicator Ui affects or is affected by the value of Uj, then there is an influence relationship between the two. This paper sets the degree of influence of primary indicator Ui on Uj into five levels: no influence, weak influence, moderate influence, significant influence, and strong influence, and corresponds to the numbers 0-4 respectively. The direct influence matrix A can be obtained by analyzing the mutual influence relationships between all primary indicators, and its calculation formula is as follows:

[0058] ; in, This indicates the degree of influence of the primary indicator Ui on Uj, where 1≤i and j≤n.

[0059] (2) Obtain the standard influence matrix. Based on the direct influence matrix A, the standard influence matrix B can be further obtained, and its calculation formula is shown below:

[0060] .

[0061] (3) Calculate the comprehensive impact matrix. The comprehensive impact matrix reflects the causal relationships between the primary indicators in the entire indicator system. Its calculation method is as follows:

[0062] ; In the above formula, I is the identity matrix.

[0063] (4) Calculate the centrality and causality of the primary indicators. Specifically, the influence degree D represents the overall influence of the i-th primary indicator on the other primary indicators, which is the sum of the elements in the i-th row of the overall influence matrix T; the degree of influence R represents the overall influence of the j-th primary indicator on the other primary indicators, which is the sum of the elements in the j-th column. Then we have:

[0064] ; .

[0065] Based on this, we further define D+R as the centrality of the indicator and DR as the causality of the indicator, representing the importance and value of the indicator in the entire indicator system, respectively.

[0066] (5) Construct the network hierarchy of the evaluation index system. Based on the centrality and causality of the first-level indicators, clarify the mutual influence relationship between the control layer indicators and the network layer indicators, and construct the network hierarchy of the evaluation index system. The control layer corresponds to the highest level, and the network layer corresponds to the middle and lowest levels.

[0067] (6) Construct the network hierarchy analysis hypermatrix. Based on the control layer element S as the criterion, further construct the hypermatrix based on the elements in Ui. For this criterion, compare the pairs of elements in the element group Ui. The influence degree is used to obtain the judgment matrix, which is then obtained through the eigenvalue method. , , ..., Equal-weighted vectors. The final result is a matrix. and the weight matrix W of the control layer (S) :

[0068] ; .

[0069] (7) Construct a weighted hypermatrix using ANP. For W (S) After column normalization, the weighted hypermatrix B is obtained. (S) :

[0070] ; And for the supermatrix W (S) We perform a weighted summation operation on each submatrix to obtain the weighted supermatrix W. (BS) : ; ; (8) Calculate the limiting hypermatrix. Stabilize the hypermatrix to obtain the limiting hypermatrix W. (S)* The calculation method is as follows:

[0071] ; If the limit is convergent and unique, then the value of the corresponding row in the original matrix is ​​the stable weight of each evaluation index.

[0072] Step 213: Finally, normalize the structural weight values ​​of each level so that the sum of the weights of all indicators at the same level is 1, and obtain the final structural weights of all indicators in the level indicator library.

[0073] Based on any of the above embodiments, for example, in S2 above, the operational quantification data of key quantifiable indicators are processed by the entropy weight method to obtain the objective weights of each key quantifiable indicator. The entropy weight method specifically includes the following steps: Step 221: Collect and clean the data, process missing and outlier values, and construct the original data matrix.

[0074] Step 222: Standardize the data for positive and negative indicators according to the formula; calculate the indicator weight, entropy value of each indicator, and information entropy redundancy; determine the objective weight of the indicators. The calculation formula is as follows:

[0075] First, we perform dimensionless transformation on each indicator. Assume we have m indicators: X 1 , X 2 , ..., X m Each indicator includes n data points. X i ={ X i1 , X i2 , ..., X in Assume the standardized values ​​for each indicator are: Y 1 , Y 2 , ..., Y m ,in, Y i ={ Y i1 , Y i2 , ..., Y in}; Positive indicator normalization Y ij Obtained using the following formula:

[0076] ; Here, min() represents the minimum function and max() represents the maximum function.

[0077] Negative indicator normalization: ; Then, calculate the ratio of each indicator under each plan: ; Next, calculate the information entropy of each indicator. : ; in ≥0, if =0, then define =0.

[0078] Finally, determine the weight of each indicator: .

[0079] Step 223: Interpret the weighting results, compare the relative magnitudes of the indicator weights, examine the impact of minor data changes on the weighting results, and evaluate the stability of the weighting results.

[0080] Based on any of the above embodiments, for example, in S2 above, the step of obtaining the subjective weights of each key quantifiable indicator through the G1 subjective weighting method includes: Step 231: Establish an expert group consisting of 8-15 experts in the field, ensuring a diverse composition of experts; develop an indicator explanation manual and design a questionnaire specifically for the G1 method.

[0081] Step 232: Each expert independently ranks the indicators from highest to lowest importance to determine the order of the indicators; for the ranked indicators, the experts evaluate the importance ratio between adjacent indicators in turn; and the collected data undergoes quality control.

[0082] Step 233: Each expert independently calculates their individual weight; the group weights are synthesized using a weighted average method; the weights are then normalized. This includes the following steps:

[0083] (1) For the evaluation index set X={ X 1 , X 2 , ..., X m}, sorted in descending order of importance, denoted as L1≥L2≥…≥Lm.

[0084] (2) Invite experts to score the relative importance of adjacent indicators after the indicators are sorted. The scoring expression for each indicator and its weight coefficient is as follows: ; In the formula, k = {m, m-1, ..., 2, 1}. , These represent the experts' evaluation indicators. and Weighting coefficients for scoring; for experts to evaluate indicators and Score the relative importance of the two.

[0085] (3) Based on relative importance The weighting coefficients for each indicator by the experts were calculated separately: ; .

[0086] (4) Perform sensitivity analysis on the weighting results: adjust the influence of changing the order relation. The degree of influence of the value; conduct practical application tests and evaluate the rationality of the weighting results.

[0087] Based on any of the above embodiments, for example, in S3 above, the comprehensive weight of each quantifiable indicator is determined from structural weight, objective weight, and subjective weight using a game theory approach, including: Step 310: Construct a game theory model, treating the entropy weight method, G1 method, and Dematel-ANP method as three game participants. Player 1 is an objective weight based on data dispersion, Player 2 is a subjective weight based on expert ranking, and Player 3 is a modified weight based on the network relationship between indicators. Define the strategy space as the adjustment of each participant's weight allocation to maximize the overall consistency of the final combined weight. Design a payoff function based on KL divergence to measure the difference between the weights of each method and the combined weight. Use Nash equilibrium as the optimization objective and solve the model through a system of linear equations.

[0088] Step 320: Combine the entropy weight method, G1 method, and DEMATEL-ANP method to construct the covariance matrix and solve the system of equations. Normalize the obtained weights. This includes:

[0089] (1) Calculation and Correlation coefficient: ; In the formula: for and The correlation coefficient; The weights obtained using the i-th method; This represents the average weight of the n indicators obtained using the i-th method. The combined weights of (m-1) methods excluding the i-th method; for The average of the weights of the n indicators.

[0090] (2) Normalize the correlation coefficient. After normalization, we get : .

[0091] (3) Calculate the combined weights : .

[0092] (4) Conduct a consistency test on the combined weights to verify the consistency between the weights of each method and analyze the correlation between the combined weights and the weights of each method; conduct a sensitivity analysis to observe the stability of the combined weights by perturbing the initial weights and eliminate indicators one by one to identify key indicators; conduct practical application verification to evaluate the applicability of the weights in the actual scenario.

[0093] Based on any of the above embodiments, for example, in S3 above, the operational quantitative data of each quantifiable indicator are weighted and fused based on comprehensive weights to obtain the evaluation values ​​of each indicator in the intermediate level immediately adjacent to the lowest level, including: ; in, This indicates that it belongs to the rural power supply support capacity index. The middle level Evaluation values ​​of each indicator; This indicates that it belongs to the rural electricity supply pressure index. The middle level Evaluation values ​​of each indicator; This indicates that it belongs to the indicators of rural development regulation and control. The middle level Evaluation values ​​of each indicator; This indicates that it belongs to the rural power supply support capacity index. The middle level The first indicator Historical quantitative data of each quantifiable indicator; This indicates that it belongs to the rural electricity supply pressure index. The middle level The first indicator Historical quantitative data of each quantifiable indicator; This indicates that it belongs to the indicators of rural development regulation and control. The middle level The first indicator Historical quantitative data of each quantifiable indicator; , and This indicates that the quantifiable indicators belong to the rural power supply support capacity indicators. Rural electricity supply pressure indicators Rural development regulation and control indicators The number of key quantifiable indicators; , and They represent , and The overall weight.

[0094] Based on any of the above embodiments, for example, in S4 above, the evaluation values ​​of each indicator in the highest level are obtained by iterating layer by layer according to the same comprehensive weight acquisition method as the lowest level, including: The evaluation values ​​of each indicator in the intermediate level are obtained from the evaluation values ​​of each indicator in the highest level: ; in, Indicators of rural power supply support capacity Evaluation value; Indicators of rural electricity supply pressure Evaluation value; Indicators representing the regulatory capacity for rural development Evaluation value; , and They represent , and The overall weight; , and This indicates that they belong to the rural power supply support capacity indicators respectively. Rural electricity supply pressure indicators Rural development regulation and control indicators The number of key quantifiable indicators.

[0095] When there are two or more intermediate levels, the evaluation values ​​of the lower-level indicators are used as the quantitative data for the operation of the upper-level indicators, following the direction from the lowest level to the highest level. The same game-theoretic weighting method as the lowest level is used to determine the comprehensive weight of each level's indicators, and the comprehensive weight is used for weighted fusion to obtain the evaluation values ​​of each indicator in the next higher level. This process is iterated upwards until the evaluation values ​​of each indicator in the highest level are obtained.

[0096] Based on any of the above embodiments, for example, in S5 above, the business carrying capacity assessment value of the target rural power supply station is obtained based on the evaluation values ​​of each indicator in the highest level, including: ; Where K represents the current assessment value of the business carrying capacity of the target rural power supply station.

[0097] Based on any of the above embodiments, for example, in S5 above, selecting a power supply optimization strategy that matches the service carrying capacity assessment value from a preset conditional IF-rule THEN strategy knowledge base includes: Step 510: Construct a Condition IF-Rule THEN strategy knowledge base. The Condition IF part of the Condition IF-Rule THEN strategy knowledge base includes the business carrying capacity assessment value level range, the evaluation value of rural power supply support capacity index, the evaluation value of rural electricity consumption pressure index, the evaluation value of rural development regulation capacity index, and the carrying capacity change trend. The Rule THEN part of the Condition IF-Rule THEN strategy knowledge base includes the corresponding current situation diagnosis conclusion, cause analysis, risk warning information, and actionable governance measures suggestions.

[0098] Step 520: Collect multiple current operational data points with quantifiable indicators from multiple data sources associated with the target rural power supply station. Merge these multiple current operational data points and input them into a pre-trained Long Short-Term Memory (LSTM) network model to obtain the changing trend of the target rural power supply station's business carrying capacity assessment value within a future preset assessment period. The LSTM network is trained using a training sample set. The training sample set includes historical multimodal fusion feature data of the target area and the corresponding rural power supply station business carrying capacity evaluation levels. Specifically, this includes the following steps:

[0099] Step 521: Based on relevant literature and reports, we have compiled existing research on the evaluation standards for the business carrying capacity of power supply stations. After consulting with relevant experts, we divided the business carrying capacity of rural power supply stations into five intervals: [0, 0.1), [0.1, 0.3), [0.3, 0.4], [0.4, 0.7], and [0.7, 1], corresponding to five levels. These intervals are then set as the final evaluation standards in this paper, as shown in Table 3.

[0100] Table 3 Evaluation Level of Rural Power Supply Station Bearing Capacity Step 522: Construct a bearing capacity prediction model, including the following steps: (1) Clean, normalize and standardize the historical multimodal fusion feature data, handle missing values ​​and outliers to ensure data quality, extract time series features, spatial features and business features, and construct time series sequence samples suitable for LSTM input.

[0101] (2) Construct a multi-layer LSTM network structure, including an input layer, a hidden layer and an output layer. Set the time step, number of hidden units, activation function (such as tanh, ReLU) and other hyperparameters, and initialize the model weights and bias terms.

[0102] (3) Use the training sample set to train the LSTM model, use the backpropagation algorithm (BPTT) to optimize the model parameters, set the loss function (such as mean square error MSE) and the optimizer (such as Adam), perform multiple rounds of iterative training, and use early stopping to prevent overfitting.

[0103] (4) Use the validation set to evaluate the performance of the trained model, adjust hyperparameters (such as learning rate, batch size, number of network layers) to improve prediction accuracy, and finally verify the generalization ability of the model through the test set to ensure its stable performance on unknown data.

[0104] Specifically, historical multimodal fusion feature data is divided into training and testing sets to train and test the Long Short-Term Memory (LSTM) network, obtain the optimal hyperparameters of the LSM network, and obtain the load-bearing capacity prediction model.

[0105] LSTM consists of an input layer, an LSTM layer, a fully connected layer, and an output layer. The input layer receives standardized time-series data, typically represented as a three-dimensional tensor (number of samples, number of time steps, number of features). Multiple LSTM units in the LSTM layer can be stacked to form a deep LSTM network to capture more complex time-series patterns. Each LSTM unit includes an input gate, a forget gate, an output gate, and a unit state. The output of the LSTM layer is passed to the fully connected layer for further feature extraction and prediction. The output layer generates the final prediction result.

[0106] LSTM introduces a logic unit, the cell state, to determine the validity of input data. Three gate structures control the updating of the cell state, enabling long-term data flow across the network. The input gate controls the data input into the memory cell, the forget gate controls the deletion of data and resetting of the memory cell, and the output gate controls the data output.

[0107] Input gate: Based on the input data, the input gate conditionally determines which data can be used to update the memory state. The specific formula is as follows: .

[0108] Forget gate: Based on the input data, the forget gate conditionally determines the information needed for the cell state according to the activation function. The specific formula is as follows: .

[0109] Output gate: Based on the input data and the information currently stored in the memory unit, it conditionally determines the output information. The specific formula is as follows: .

[0110] in, , , These are the output data for the input gate, forget gate, and output gate, respectively. The proportion of forgetting determines the cellular state. Determine the proportion of current data input into the cell state. Determine the cell state output to the current state. proportion, The input data is at time t. The state at time t-1 For activation function, , , These are the weight vectors for the input gate, forget gate, and output gate, respectively. , , These are the phase vectors of the input gate, forget gate, and output gate, respectively, and are all parameters to be trained.

[0111] Figure 5 This specification provides a schematic diagram of the unit structure of a Long Short-Term Memory (LSTM) network, such as... Figure 5 As shown, the input data consists of the input data at time t. and state of time The calculation results show that, The function restricts the function value to To create a new candidate state vector, the memory information at time t. Memory information at time t-1 after passing through the forgetting gate And the input information passing through the input gate determines: ; in, For memory cells, This refers to the phase of memory cells.

[0112] State at time t From the output gate Decide how to update: ; The LSTM is trained using the training set, its performance is evaluated using the test set, and its hyperparameters are tuned to optimize it. If the LSTM reaches the required accuracy, it is saved and used; otherwise, it is trained repeatedly until the required accuracy is achieved.

[0113] Step 530: Combine the changing trend of the business carrying capacity assessment value with the evaluation values ​​of the rural power supply support capacity index, the rural electricity consumption pressure index, the rural development regulation capacity index, and the business carrying capacity assessment value level to form a multi-dimensional comprehensive diagnostic condition set.

[0114] Step 540: Match the multi-dimensional comprehensive diagnostic condition set with the conditional IF-rule THEN strategy knowledge base, traverse all matching rules that meet the conditions, eliminate logically conflicting items, and determine rule priority based on the risk level and the degree of influence of key indicators. Use the current K, S, P, and A values, the future K value trend, and the obtained key quantifiable indicator information as comprehensive diagnostic conditions. Match these comprehensive diagnostic conditions with the rules in the strategy knowledge base, and generate a differentiated strategy through logical reasoning that includes current status diagnosis, weakness analysis, risk warning, and specific optimization suggestions. This includes the following steps:

[0115] Step 541: Construct a strategy knowledge base. Integrate historical operation and maintenance data, industry guidelines, expert experience, and management procedures to form an initial rule set. Adopt a standardized structure of "IF<condition>THEN<action>": the condition part is a multi-dimensional logical combination, covering carrying capacity status (K-value level), key weakness indicators, future risk level (S / P / A trend), etc.; the action part corresponds to a specific optimization suggestion project library.

[0116] Step 542, input comprehensive diagnostic information. The engine receives the processing results from the previous steps in real time. The core inputs include: the current quantitative scores and levels of K, S, P, and A, the future trend of K value changes based on situational projections, and a list of key influencing indicators derived from the source analysis in step S2.

[0117] Step 543: Perform logical matching and reasoning. The comprehensive diagnostic conditions described above are automatically matched with rules in the strategy knowledge base. Using efficient algorithms such as Rete, rapid logical reasoning is performed on combinations of multiple conditions to identify all applicable rules.

[0118] Step 544: Generate and output differentiated strategies. Based on the matched rules, the engine automatically synthesizes a structured report, which includes: Current Status Diagnosis: clarifying the current carrying capacity level and stage; Shortcoming Analysis: accurately identifying the key indicators that lead to the current state; Risk Warning: indicating potential future business risks if no intervention is taken; Optimization Suggestions: outputting specific and actionable measures, such as "initiating specialized skills training" or "supplementing the procurement of specific emergency supplies," forming a differentiated optimization plan tailored to each institution.

[0119] Step 550: Generate differentiated power supply optimization strategies based on the matching results. The strategies include current status diagnosis conclusions, cause analysis, risk warning information, and actionable governance measures recommendations.

[0120] In one exemplary embodiment, Figure 6 This manual provides a functional module diagram of a rural power supply station business capacity assessment and optimization system, as shown below. Figure 6 As shown in the figure, this application provides a rural power supply station business carrying capacity assessment and optimization system, which specifically includes: a multimodal data perception and fusion module, a dynamic indicator diagnosis and management module, a carrying capacity intelligent assessment module, a future situation prediction and early warning module, and a differentiated strategy generation module.

[0121] The multimodal data perception and fusion module is used to collect and fuse internal operational data and external multimodal environmental data from rural power supply stations.

[0122] The dynamic indicator diagnosis and management module is connected to the multimodal data perception and fusion module. The dynamic indicator diagnosis and management module is used to dynamically filter out key impact indicators related to the target business scenario.

[0123] The intelligent load-bearing capacity assessment module is connected to the dynamic index diagnosis and management module. The intelligent load-bearing capacity assessment module is used to calculate the comprehensive evaluation value K of the current business load-bearing capacity of the power supply station and the sub-evaluation values ​​of the three dimensions of support capacity S, pressure P, and regulation capacity A.

[0124] The future situation prediction and early warning module is connected to the carrying capacity intelligent assessment module. The future situation prediction and early warning module is used to calculate the trend of change of the business carrying capacity K value in a specific future period.

[0125] The differentiated strategy generation module is connected to the future situation prediction and early warning module. The differentiated strategy generation module is used to generate differentiated business carrying capacity optimization strategies.

[0126] The multimodal data perception and fusion module includes a data acquisition unit, a data cleaning and standardization unit, and a data fusion and storage unit. The data acquisition unit is configured to automatically collect internal power grid business data and external environmental data. The internal power grid business data includes, but is not limited to, SCADA system data, electricity consumption information collection system data, asset management system data, and customer service work order data. The external multimodal environmental data includes, but is not limited to, meteorological data, geographic information data, macroeconomic data, demographic data, and third-party commercial platform data. The data cleaning and standardization unit is configured to clean, align, and standardize the collected data. The data fusion and storage unit is configured to integrate and store the processed data in a unified data resource pool.

[0127] The dynamic indicator diagnosis and management module includes an indicator library management unit, a feature importance analysis unit, and a model interpretability analysis unit. The indicator library management unit is configured to store and manage a preset bearing capacity assessment indicator system, which includes three primary indicators: supporting force (S), applied pressure (P), and regulating force (A), as well as eight secondary and 34 tertiary indicators belonging to the above primary indicators. The feature importance analysis unit selects key influencing indicators from the indicator library based on an ensemble learning algorithm. The model interpretability analysis unit uses the SHAP framework to analyze the influence mechanism of key indicators.

[0128] The intelligent load-bearing capacity assessment module includes a combined weighting unit and an assessment calculation unit; the combined weighting unit determines the weights of the indicators through a combination of subjective and objective weighting methods; the assessment calculation unit calculates the sub-item evaluation values ​​of each dimension and the comprehensive load-bearing capacity evaluation value through linear and nonlinear models.

[0129] The future situation simulation and early warning module includes a load prediction unit, a situation simulation unit, and a risk early warning unit; the load prediction unit predicts the future carrying capacity K through a prediction model.

[0130] The differentiated strategy generation module includes a strategy knowledge base and a strategy matching and generation unit. The strategy knowledge base is configured to store "IF-THEN" type optimization strategy rules, which define the mapping relationship between different carrying capacity states, key weakness indicators, future risk levels, and specific optimization action plans. The strategy matching and generation unit is configured to receive the current K, S, P, and A values, future K value trends, and key influencing indicator information, and use them as comprehensive input conditions to perform multi-condition matching and logical reasoning in the strategy knowledge base, automatically assembling and generating a structured optimization strategy report.

[0131] Specific limitations regarding the rural power supply station business capacity assessment and optimization system can be found in the above-mentioned limitations on the rural power supply station business capacity assessment and optimization methods, and will not be repeated here. Each module in the aforementioned rural power supply station business capacity assessment and optimization system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0132] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0133] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for assessing and optimizing the business carrying capacity of rural power supply stations, characterized in that, include: Construct a hierarchical indicator library, including a top level, a bottom level, and at least one intermediate level between the top and bottom levels; wherein, the top level includes indicators of rural power supply support capacity, rural power consumption capacity, and rural development regulation capacity; the bottom level includes quantifiable indicators corresponding to rural power supply support capacity, rural power consumption capacity, and rural development regulation capacity, respectively. Obtain the structural weight, objective weight, and subjective weight of each quantifiable indicator in the lowest level; among them, the structural weight is used to quantify the structural importance of each indicator in the hierarchical indicator library; the objective weight is used to quantify the objective importance of each indicator based on the actual operation data of the power supply station; and the subjective weight is used to quantify the subjective importance of each indicator based on expert experience judgment. The comprehensive weight of each quantifiable indicator is determined from structural weight, objective weight and subjective weight by game theory, and the operational quantitative data of each quantifiable indicator is weighted and integrated based on the comprehensive weight to obtain the evaluation value of the intermediate level indicator to which the lowest level belongs. Based on the evaluation values ​​of intermediate-level indicators, the same game-theoretic weighting method as the lowest level is used to iterate and aggregate upwards level by level until the evaluation values ​​of each indicator in the highest level are obtained. Based on the evaluation values ​​of each indicator in the highest level, the business carrying capacity assessment value of the target rural power supply station is obtained; a power supply optimization strategy that matches the business carrying capacity assessment value is selected from the preset condition IF-rule THEN strategy knowledge base.

2. The method for assessing and optimizing the business carrying capacity of rural power supply stations according to claim 1, characterized in that, The methods for obtaining the objective weights of each quantifiable indicator in the lowest level include: The evaluation model is trained by taking the historical operational data of all quantifiable indicators of the target rural power supply station as input and the historical assessment value of the business carrying capacity of the target rural power supply station as output. Based on the completed evaluation model, the contribution of each quantifiable indicator to the historical evaluation value of business carrying capacity is determined, and key quantifiable indicators that match the target rural power supply station are selected from the quantifiable indicators according to the contribution. The entropy weight method is used to process the operational quantification data of key quantifiable indicators to obtain the objective weights of each key quantifiable indicator.

3. The method for assessing and optimizing the business carrying capacity of rural power supply stations according to claim 2, characterized in that, The process of training an evaluation model, which takes historical operational data of all quantifiable indicators of the target rural power supply station as input and historical assessment values ​​of the target rural power supply station's business carrying capacity as output, specifically includes: Historical operational data with quantifiable indicators are collected from data sources associated with the target rural power supply stations. After data cleaning, outlier processing, and normalization, standardized time-series data is obtained. Statistical and temporal features of normalized time-series data are extracted, and multi-source data alignment and feature fusion are performed based on a unified spatiotemporal benchmark to generate multimodal fusion features. The multimodal fusion features are labeled according to the preset service carrying capacity level standard, and the training set, validation set and test set are divided. Using random forest as the base learner, the model is trained using the training set, and hyperparameters are optimized and overfitting is suppressed based on the validation set. After the performance is verified on the test set to meet the preset accuracy threshold, the evaluation model is completed.

4. The method for assessing and optimizing the business carrying capacity of rural power supply stations according to claim 3, characterized in that, The selection of key quantifiable indicators that match the target rural power supply station from quantifiable indicators based on contribution specifically includes: The interpretability analysis of the trained evaluation model is carried out using the SHAP interpretation model based on game theory. The marginal contribution of each quantifiable indicator to the evaluation value of business carrying capacity is calculated using each quantifiable indicator as input feature. The global importance measure of each quantifiable indicator is calculated based on the marginal contribution, and the importance ranking of the indicators is generated. By combining the importance of built-in features in the evaluation model with global importance metrics, redundant indicators are eliminated, and indicators that have a significant impact on the business carrying capacity evaluation results are retained as key quantifiable indicators.

5. The method for assessing and optimizing the business carrying capacity of rural power supply stations according to claim 1, characterized in that, The subjective weights of all indicators in the hierarchical indicator library are obtained through the G1 subjective weighting method.

6. The method for assessing and optimizing the business carrying capacity of rural power supply stations according to claim 1, characterized in that, The structural weights of all indicators in the hierarchical indicator library are obtained through DEMATEL-ANP; specifically including: The influence relationships between indicators are determined using the DEMATEL method: a direct influence matrix between indicators in a hierarchical indicator library is constructed, and the direct influence matrix is ​​normalized to obtain a normalized influence matrix. A comprehensive influence matrix is ​​calculated based on the normalized influence matrix. The influence degree and the degree of being influenced by each indicator are calculated based on the comprehensive influence matrix, and then the centrality and causality of each indicator are determined. The causal relationship and importance of each indicator are analyzed based on the centrality and causality, and key driving indicators and key result indicators at each level are identified. A hierarchical network structure is constructed and weights are calculated based on the ANP method: a hierarchical network structure model of each level of indicators is constructed according to causal relationships, and the dependency and feedback relationships between each indicator are determined; an unweighted hypermatrix is ​​constructed based on the dependency and feedback relationships, and the unweighted hypermatrix is ​​weighted by combining centrality to obtain a weighted hypermatrix; a limit operation is performed on the weighted hypermatrix, and the limit hypermatrix is ​​obtained when the matrix converges; the feature vectors corresponding to each indicator in the limit hypermatrix are extracted as the structural weight values ​​of each indicator. The structural weight values ​​of each level are normalized so that the sum of the weights of all indicators at the same level is 1, thus obtaining the final structural weights of all indicators in the hierarchical indicator library.

7. The method for assessing and optimizing the business carrying capacity of rural power supply stations according to claim 1, characterized in that, The method of determining the comprehensive weight of each quantifiable indicator from structural weight, objective weight, and subjective weight using game theory specifically includes: The structural weight, objective weight, and subjective weight of the same quantifiable indicator are regarded as the strategy sets of the three game participants. Each participant aims to maximize the reflection of its own weight information in the comprehensive weight, while minimizing the deviation of the weights of other participants. Establish a game optimization model: construct three original weight vectors consisting of structural weight, objective weight, and subjective weight, set a comprehensive weight vector, and construct an objective function to minimize the sum of squared deviations between the comprehensive weight and the three original weight vectors; introduce combination coefficients to perform linear weighted aggregation of the three types of original weights, and set the constraint that the sum of the combination coefficients is one and each coefficient is positive. Solve the game optimization model to obtain the optimal combination coefficients that minimize the sum of squared deviations; Based on the optimal combination coefficients, the three original weight vectors are weighted and fused to obtain the preliminary game comprehensive weights of each quantifiable indicator. The preliminary game comprehensive weights are then normalized to obtain the comprehensive weights of each quantifiable indicator.

8. The method for assessing and optimizing the business carrying capacity of rural power supply stations according to claim 1, characterized in that, The evaluation values ​​based on intermediate-level indicators are iteratively aggregated upwards level by level using the same game-theoretic weighting method as the lowest level, until the evaluation values ​​of each indicator in the highest level are obtained. Specifically, this includes: The evaluation value of the intermediate level indicator to which the lowest level belongs is obtained based on the operational quantitative data of quantifiable indicators: in, This indicates that it belongs to the rural power supply support capacity index. The middle level Evaluation values ​​of each indicator; This indicates that it belongs to the rural electricity supply pressure index. The middle level Evaluation values ​​of each indicator; This indicates that it belongs to the indicators of rural development regulation and control. The middle level Evaluation values ​​of each indicator; This indicates that it belongs to the rural power supply support capacity index. The middle level The first indicator Historical quantitative data of each quantifiable indicator; This indicates that it belongs to the rural electricity supply pressure index. The middle level The first indicator Historical quantitative data of each quantifiable indicator; This indicates that it belongs to the indicators of rural development regulation and control. The middle level The first indicator Historical quantitative data of each quantifiable indicator; , and This indicates that the quantifiable indicators belong to the rural power supply support capacity indicators. Rural electricity supply pressure indicators Rural development regulation and control indicators The number of key quantifiable indicators; , and They represent , and The overall weight; The evaluation values ​​of each indicator in the highest level are obtained based on the evaluation values ​​of the intermediate-level indicators: in, Indicators of rural power supply support capacity Evaluation value; Indicators of rural electricity supply pressure Evaluation value; Indicators representing the regulatory capacity for rural development Evaluation value; , and They represent , and The overall weight; , and This indicates that they belong to the rural power supply support capacity indicators respectively. Rural electricity supply pressure indicators Rural development regulation and control indicators The number of key quantifiable indicators.

9. The method for assessing and optimizing the business carrying capacity of rural power supply stations according to claim 1, characterized in that, The assessment value of the business carrying capacity of the target rural power supply station, obtained based on the evaluation values ​​of each indicator at the highest level, is achieved through the following formula: Where K represents the current assessment value of the business carrying capacity of the target rural power supply station.

10. The method for assessing and optimizing the business carrying capacity of rural power supply stations according to claim 1, characterized in that, The step of selecting a power supply optimization strategy that matches the service carrying capacity assessment value from a preset conditional IF-rule THEN strategy knowledge base specifically includes: A knowledge base for the IF-rule THEN strategy is constructed. The IF part of the IF-rule THEN strategy knowledge base includes the business carrying capacity assessment value level range, the evaluation value of the rural power supply support capacity index, the evaluation value of the rural electricity consumption capacity index, the evaluation value of the rural development regulation capacity index, and the carrying capacity change trend. The rule THEN part of the IF-rule THEN strategy knowledge base includes the corresponding current situation diagnosis conclusion, cause analysis, risk warning information, and actionable governance measures suggestions. Multiple quantifiable indicators of current operation data are collected from multiple data sources associated with the target rural power supply station. The multiple current operation data are then merged and input into a pre-trained long short-term memory network model to obtain the changing trend of the business carrying capacity assessment value of the target rural power supply station within a future preset assessment period. The changing trend of the business carrying capacity assessment value is combined with the evaluation values ​​of rural power supply support capacity indicators, rural electricity consumption capacity indicators, rural development regulation capacity indicators, and business carrying capacity assessment value level to form a multi-dimensional comprehensive diagnostic condition set. The multi-dimensional comprehensive diagnostic condition set is matched with the conditional IF-rule THEN strategy knowledge base, and a differentiated power supply optimization strategy is generated based on the matching results.

Images