Distribution network differential power supply method and system based on load grading

Abstract

The invention relates to the technical field of power distribution network power supply, and discloses a power distribution network differential power supply method and system based on load grading. The system comprises a data acquisition and processing module, a load grading evaluation module, a differentiated power supply strategy generation module, a power supply control execution module, an effect evaluation and strategy optimization module and a strategy management updating module. When power distribution network load grading evaluation is carried out, load grading is carried out by constructing a multi-dimensional power utilization characteristic index system and adopting an adaptive clustering algorithm, power utilization characteristics and importance of different loads can be accurately described, and the problem of classification deviation caused by incomplete characteristic extraction of a traditional method is solved; the accuracy and the reliability of a load grading result are ensured, and a solid basis is provided for formulating a subsequent differentiated power supply strategy.

Description

Technical Field

[0001] This invention relates to the field of power supply technology for power distribution networks, specifically to a differentiated power supply method and system for power distribution networks based on load grading. Background Technology

[0002] A distribution network refers to a power grid that receives electrical energy from the transmission network or regional power plants and distributes it locally or in stages according to voltage to various users through distribution facilities. It consists of overhead lines, cables, poles, distribution transformers, disconnect switches, reactive power compensators, and some auxiliary facilities, and plays an important role in distributing electrical energy within the power grid.

[0003] Currently, due to the complexity and diversity of load types and the dynamic changes in electricity consumption behavior in the distribution network, traditional methods are difficult to extract and quantify the electricity consumption characteristics of different loads in real time and accurately when conducting load classification assessment. When the characteristic parameters on which the classification is based are incomplete and the weight allocation is inappropriate, the load classification results will deviate from the actual importance of electricity consumption and cannot provide a reliable basis for differentiated power supply.

[0004] Therefore, a differentiated power supply method and system for distribution networks based on load grading is proposed to solve the above problems. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for differentiated power supply in distribution networks based on load grading, which solves the problem mentioned in the background that the load grading results deviate from the actual importance of electricity consumption and cannot provide a reliable basis for differentiated power supply.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a differentiated power supply method and system for distribution networks based on load grading, wherein the method includes the following steps: S1. Collect real-time and historical electricity consumption data of multiple load nodes in the distribution network, and generate a load node electricity consumption dataset. S2. Based on the load node electricity consumption dataset, perform load node electricity consumption feature extraction processing to generate a load node electricity consumption feature vector set; S3. Perform load node hierarchical evaluation processing based on the load node electricity consumption feature vector set to generate a load node hierarchical evaluation result set. S4. Based on the load node hierarchical evaluation result set and combined with the real-time operation status data of the distribution network, perform differentiated power supply strategy generation processing to generate a target differentiated power supply strategy set. S5. Perform power distribution network power supply control execution processing according to the target differentiated power supply strategy set, and generate power distribution network differentiated power supply execution result data; S6. Based on the power distribution network differentiated power supply execution result data, perform power supply effect evaluation and strategy optimization processing to generate an optimized differentiated power supply strategy set; S7. Update the optimized differentiated power supply strategy set to the power distribution network control system to realize dynamic optimization and adjustment of the power distribution network strategy.

[0007] Preferably, the data collection and dataset generation in step S1 includes the following steps: S11. Collect real-time electricity consumption data of multiple load nodes in the distribution network through smart meters, including real-time voltage, current, power factor and active power data; S12. Extract historical electricity consumption data of the multiple load nodes from the distribution network data management platform, including historical daily load curves, monthly electricity consumption and annual peak electricity consumption data; S13. Perform data fusion processing on the real-time electricity consumption data and the historical electricity consumption data to generate the load node electricity consumption dataset. : ; in Indicates the first A set of electricity consumption data for each load node. Indicates the index of the load node. This indicates the total number of load nodes.

[0008] Preferably, the electricity consumption feature extraction process in step S2 includes the following steps: S21. Data set of power consumption at the load nodes. Perform data preprocessing, including data cleaning, missing value imputation, and outlier handling; S22. Extract the power consumption characteristic parameters of each load node, including load factor characteristic parameters, peak-valley difference characteristic parameters, power consumption stability characteristic parameters, and importance characteristic parameters; S23. Based on principal component analysis, the extracted electricity consumption feature parameters are subjected to dimensionality reduction processing to generate the electricity consumption feature vector set of the load node. : ; in Indicates the first The electricity consumption characteristic vector of each load node, Indicates the index of the load node. This indicates the total number of load nodes.

[0009] Preferably, the graded evaluation process in S3 includes the following steps: S31. Construct a load grading assessment model. The model adopts an improved fuzzy C-means clustering algorithm, and its objective function is: ; in The objective function value of the fuzzy C-means clustering algorithm is... Represents the number of cluster centers. Indicates the first The load node belongs to the first Membership degree of a class For fuzzy weighting index, Indicates the first Cluster centers of classes For regularization parameters, This indicates the total number of load nodes. Indicates the first Electricity consumption characteristic vector of each load node; S32. Optimize the parameters of the load grading assessment model using an adaptive particle swarm optimization algorithm, including the number of clusters. and fuzzy weight index ; S33, Set the power consumption feature vector set of the load node Input the optimized load classification assessment model to generate the load node classification assessment result set. : ; in Indicates the first The classification results of each load node Indicates the index of the load node. This indicates the total number of load nodes.

[0010] Preferably, generating a differentiated power supply strategy in step S4 includes the following steps: S41. Obtain real-time operating status data of the distribution network, including grid frequency, node voltage level, line load rate and power supply reliability indicators; S42. Establish a multi-objective optimization model to generate differentiated power supply strategies. The optimization objectives include: Objective 1: Minimize the total operating cost of the distribution network. ; in Indicates power generation capacity. Indicates the cost of electricity generation. This represents the total operating cost of the power distribution network; Objective 2: Maximize power supply reliability: ; in Indicates the first Power supply reliability indicators for each load node Indicates the weighting coefficient. This indicates the overall power supply reliability index; Objective 3: Maximize load satisfaction. ; in Indicates the first Satisfaction index of each load node This represents the load importance coefficient. This represents the overall load satisfaction index; S43. The improved non-dominated sorting genetic algorithm is used to solve the multi-objective optimization model and generate the Pareto optimal solution set. S44. Based on the entropy-weighted TOPSIS method, select the optimal solution from the Pareto optimal solution set to generate the target differentiated power supply strategy set. .

[0011] Preferably, the power supply control execution process in S5 includes the following steps: S51. Analyze the target differentiated power supply strategy set. Generate a specific set of power supply control instructions; S52. Execute power supply control commands through the distribution network intelligent control system, including load control commands, switch operation commands, and generation dispatch commands; S53. Monitor the power supply control execution process in real time and collect power supply execution status data; S54. Generate differentiated power supply execution result data E for the power supply execution status data of the distribution network, including actual power supply, power supply quality and user satisfaction data.

[0012] Preferably, the effect evaluation and strategy optimization process in S6 includes the following steps: S61. Establish a power supply performance evaluation index system, including technical indicators, economic indicators and social indicators; S62. Use Data Envelopment Analysis to evaluate the power supply effect and calculate the relative efficiency value of each power supply strategy; S63. Based on the evaluation results, a deep reinforcement learning algorithm is used to optimize the policy and update the policy parameters; S64. Generate the optimized differentiated power supply strategy set. : ; in Indicates the first An optimized power supply strategy, Indicates the strategy index, This indicates the total number of strategies.

[0013] Preferably, the update strategy and dynamic optimization adjustment in S7 includes the following steps: S71, The optimized differentiated power supply strategy set Compare and analyze with existing power supply strategies; S72. Calculate the benefit assessment value for strategy update. When the benefit assessment value exceeds the preset threshold, execute strategy update. S73. Update the power supply strategy through the strategy management module of the power distribution network control system; S74. Establish a strategy update log to record the content, timing, and expected effects of strategy changes.

[0014] Preferably, updating the power supply policy through the policy management module in step S73 includes the following steps: S731: Receives external command data from the distribution network dispatch center and internal early warning signals from the early warning processing module in real time, and generates a set of policy update trigger events; S732. Parse the set of policy update triggering events, call the preset rule base based on the event type and level, and generate a preliminary policy adjustment plan. The rule base defines the mapping relationship between different event types and differentiated power supply policy adjustment parameters. S733. The preliminary policy adjustment scheme is input into a pre-trained deep Q-network reinforcement learning model, which uses minimizing load loss and maximizing system stability as the reward function and outputs an optimized policy adjustment action sequence. S734. Adjust the action sequence based on the strategy, perform simulation evaluation in a virtual simulation environment, and calculate multiple performance indicators of the simulation results; S735. When the overall performance index of the simulation result is better than the current operating strategy, a strategy update execution instruction is generated and automatically sent to the power supply control execution module. When the simulation result does not meet the requirements, it is fed back to the effect evaluation and optimization module to start a new round of optimization process.

[0015] Preferably, the system includes: The data acquisition and processing module collects real-time and historical electricity consumption data from multiple load nodes in the distribution network, performs data fusion processing, and generates a load node electricity consumption dataset. The load classification and assessment module receives the load node electricity consumption dataset, performs load node electricity consumption feature extraction processing and load node classification and assessment processing, and generates a load node electricity consumption feature vector set and a load node classification and assessment result set. The differentiated power supply strategy generation module receives the load node hierarchical evaluation result set, combines it with the real-time operation status data of the distribution network, performs differentiated power supply strategy generation processing, and generates a target differentiated power supply strategy set. The power supply control execution module receives the target differentiated power supply strategy set, performs power distribution network power supply control command parsing and distribution processing, and generates power distribution network differentiated power supply execution result data. The effect evaluation and strategy optimization module receives the differentiated power supply execution result data of the distribution network, performs power supply effect evaluation and strategy optimization processing, and generates an optimized differentiated power supply strategy set. The strategy management and update module receives the optimized differentiated power supply strategy set, performs strategy benefit evaluation and dynamic update processing, and realizes dynamic optimization and adjustment of the power supply strategy of the distribution network. Beneficial effects

[0016] Compared with existing technologies, the present invention provides a differentiated power supply method and system for distribution networks based on load grading, which has the following beneficial effects: 1. In this invention, when conducting load classification assessment of distribution networks, a multi-dimensional electricity consumption characteristic index system is constructed and an adaptive clustering algorithm is used for load classification. This can accurately characterize the electricity consumption characteristics and importance of different loads, solve the classification bias problem caused by incomplete feature extraction in traditional methods, ensure the accuracy and reliability of load classification results, and provide a solid basis for the formulation of subsequent differentiated power supply strategies.

[0017] 2. In this invention, when formulating differentiated power supply strategies, a multi-objective dynamic optimization model that integrates real-time operating status and load classification results is established to achieve accurate matching and dynamic adjustment between power supply strategies and actual grid operating conditions. This solves the problem of disconnect between strategy generation and real-time status, and can generate the optimal power supply scheme under multiple objectives such as economy, power supply reliability and user satisfaction, thereby improving the refinement and adaptability of power supply decisions.

[0018] 3. In this invention, during the execution of the power supply strategy, a strategy evaluation and adaptive optimization mechanism based on closed-loop feedback and reinforcement learning is constructed. This mechanism can perceive the strategy execution effect in real time and dynamically adjust the strategy parameters, solving the problem of the lack of feedback adjustment capability in traditional methods. This ensures that the system can quickly self-correct when the external environment changes or when deviations occur in the execution, thus guaranteeing the stability of the power supply quality and the overall intelligence level of the system. Attached Figure Description

[0019] Figure 1 This is a flowchart of the differentiated power supply method for distribution networks based on load grading according to the present invention; Figure 2 This is a diagram illustrating the architecture of the differentiated power supply system for distribution networks based on load grading, as described in this invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] For specific implementation examples, please refer to: Figure 1-2 A differentiated power supply method and system for distribution networks based on load grading, the method including the following steps: S1. Collect real-time and historical electricity consumption data of multiple load nodes in the distribution network, and generate a load node electricity consumption dataset. S2. Based on the load node electricity consumption dataset, perform load node electricity consumption feature extraction processing to generate a load node electricity consumption feature vector set; S3. Perform load node hierarchical evaluation processing based on the load node electricity consumption characteristic vector set, and generate a load node hierarchical evaluation result set. S4. Based on the load node hierarchical evaluation result set and combined with the real-time operation status data of the distribution network, a differentiated power supply strategy generation process is performed to generate a target differentiated power supply strategy set. S5. Perform power supply control execution processing on the distribution network according to the target differentiated power supply strategy set, and generate power supply execution result data of the distribution network; S6. Based on the differentiated power supply execution result data of the distribution network, evaluate the power supply effect and optimize the strategy to generate an optimized differentiated power supply strategy set; S7. Update the optimized differentiated power supply strategy set to the power distribution network control system to realize the dynamic optimization and adjustment of the power distribution network strategy.

[0022] The steps involved in collecting data and generating a dataset in S1 are as follows: S11. Collect real-time electricity consumption data of multiple load nodes in the distribution network through smart meters, including real-time voltage, current, power factor and active power data; S12. Extract historical electricity consumption data from multiple load nodes from the distribution network data management platform, including historical daily load curves, monthly electricity consumption, and annual peak electricity consumption data; S13. Perform data fusion processing on real-time electricity consumption data and historical electricity consumption data to generate a load node electricity consumption dataset. : ; in Indicates the first A set of electricity consumption data for each load node. Indicates the index of the load node. Indicates the total number of load nodes; Data fusion processing specifically includes the following steps: S131. Perform timestamp alignment processing on real-time electricity consumption data and historical electricity consumption data to generate a raw data queue with consistent time series. S132. The original data queue is denoised using a wavelet transform algorithm to remove abnormal fluctuation data points. The wavelet transform denoising process includes processing the wavelet coefficients using a threshold function, and the calculation formula is as follows: ; in These are the original wavelet coefficients. These are the processed coefficients. For threshold function, For threshold; S133. Based on the Kalman filter algorithm, perform state estimation and data smoothing on the denoised data to generate a load node power consumption dataset. The core recursive equations of the Kalman filter algorithm include: State prediction equation: ; in This represents the prior state estimate. Represents the state transition matrix. This represents the posterior state estimate. Indicates a time index; ; in This indicates a priori estimation of the covariance. This indicates the posterior estimate of the covariance. Represents the state transition matrix transpose, Represents the process noise covariance matrix. Indicates a time index; State update equation: ; in Indicates Kalman gain, This indicates a priori estimation of the covariance. Represents the observation matrix. Represents the observation matrix transpose, Represents the measurement noise covariance matrix. Indicates a time index; ; in This represents the posterior state estimate. This represents the actual measured value. Indicates the predicted observation value. Indicates measurement residuals, Indicates a time index; ; in This indicates the posterior estimate of the covariance. Represents the identity matrix. Indicates a time index; The electricity consumption feature extraction process in S2 includes the following steps: S21. Data set of power consumption at load nodes Perform data preprocessing, including data cleaning, missing value imputation, and outlier handling; S22. Extract the power consumption characteristic parameters of each load node, including load factor characteristic parameters, peak-valley difference characteristic parameters, power consumption stability characteristic parameters, and importance characteristic parameters; S23. Based on principal component analysis, the extracted electricity consumption characteristic parameters are subjected to dimensionality reduction processing to generate a set of electricity consumption characteristic vectors for load nodes. Principal component analysis includes the following steps: S231. Calculate the covariance matrix of the electricity consumption characteristic parameters, and solve for the eigenvalues ​​and eigenvectors of the matrix; S232. Sort the eigenvalues ​​from largest to smallest, and select the eigenvectors corresponding to the first k eigenvalues ​​to form a projection matrix; S233. Multiply the original feature parameter data with the projection matrix to generate the dimensionality-reduced load node electricity consumption feature vector set. : ; in Indicates the first The electricity consumption characteristic vector of each load node, Indicates the index of the load node. This indicates the total number of load nodes.

[0023] The graded assessment process in S3 includes the following steps: S31. Construct a load grading assessment model. The model adopts an improved fuzzy C-means clustering algorithm, and its objective function is: ; in The objective function value of the fuzzy C-means clustering algorithm is... Represents the number of cluster centers. Indicates the first The load node belongs to the first Membership degree of a class For fuzzy weighting index, Indicates the first Cluster centers of classes For regularization parameters, This indicates the total number of load nodes. Indicates the first Electricity consumption characteristic vector of each load node; S32. Optimize the parameters of the load grading assessment model using an adaptive particle swarm optimization algorithm, including the number of clusters. and fuzzy weight index Specifically, it includes the following steps: S321. Initialize the particle swarm's position and velocity; position represents the cluster number. and fuzzy weight index The possible combinations of values; S322. Calculate the fitness value of each particle and use the fitness function to evaluate the clustering effect. S323. Dynamically adjust the particle's velocity and position based on the particle's historical optimal position and global optimal position; S324. When the maximum number of iterations and fitness value converge, output the optimal number of clusters. and fuzzy weight index ; S33, Set the power consumption characteristic vector of the load node Input the optimized load classification assessment model to generate a load node classification assessment result set. : ; in Indicates the first The classification results of each load node Indicates the index of the load node. This indicates the total number of load nodes.

[0024] Generating a differentiated power supply strategy in S4 includes the following steps: S41. Obtain real-time operating status data of the distribution network, including grid frequency, node voltage level, line load rate and power supply reliability indicators; S42. Establish a multi-objective optimization model to generate differentiated power supply strategies. The optimization objectives include: Objective 1: Minimize the total operating cost of the distribution network. ; in Indicates power generation capacity. Indicates the cost of electricity generation. This represents the total operating cost of the power distribution network; Objective 2: Maximize power supply reliability: ; in Indicates the first Power supply reliability indicators for each load node Indicates the weighting coefficient. This indicates the overall power supply reliability index; Objective 3: Maximize load satisfaction. ; in Indicates the first Satisfaction index of each load node This represents the load importance coefficient. This represents the overall load satisfaction index; S43. An improved non-dominated sorting genetic algorithm is used to solve the multi-objective optimization model and generate the Pareto optimal solution set. This includes the following steps: S431. Initialize the population, with each individual representing a power supply strategy scheme; S432. Calculate multiple objective function values ​​for each individual in the population and perform non-dominated sorting; S433. Use the crowding comparison operator to maintain the distribution of the solution set; S434. Generate a new generation of population by using tournament selection operator, simulated binary crossover operator and polynomial mutation operator; S435. When the maximum number of generations is reached, output the Pareto optimal solution set; S44. Based on the entropy-weighted TOPSIS method, select the optimal solution from the Pareto optimal solution set to generate the target differentiated power supply strategy set. The entropy weight-TOPSIS method is implemented through the following steps: S441. Calculate the entropy weight of each objective function and determine the weight coefficient of each objective. S442. Construct a weighted normalized decision matrix; S443. Determine the positive ideal solution and the negative ideal solution; S444. Calculate the distance from each scheme to the positive and negative ideal solutions; S445. Calculate the relative closeness of each scheme and select the scheme with the highest relative closeness as the optimal solution.

[0025] The power supply control execution process in S5 includes the following steps: S51, Analyzing the Target Differentiated Power Supply Strategy Set Generate a specific set of power supply control instructions; S52. Execute power supply control commands through the distribution network intelligent control system, including load control commands, switch operation commands, and generation dispatch commands; S53. Monitor the power supply control execution process in real time and collect power supply execution status data; S54. Generate differentiated power supply execution result data E for the distribution network based on the power supply execution status data, including actual power supply, power supply quality and user satisfaction data.

[0026] The effectiveness evaluation and strategy optimization process in S6 includes the following steps: S61. Establish a power supply performance evaluation index system, including technical indicators, economic indicators and social indicators; S62. Evaluate the power supply effect using Data Envelopment Analysis (DEA) and calculate the relative efficiency values ​​of each power supply strategy, including the following steps: S621. Treat each power supply strategy as a decision-making unit and determine its input and output indicators; S622. Construct an input-output data matrix for all decision-making units; S623. Select a suitable DEA model, construct a linear programming equation, and calculate the relative efficiency value of each decision unit; S624. Solve the linear programming equations to obtain the relative efficiency values ​​and corresponding weight vectors of each power supply strategy. S63. Based on the evaluation results, a deep reinforcement learning algorithm is used to optimize the policy and update the policy parameters. The deep reinforcement learning algorithm is implemented through the following steps: S631. Construct a deep Q-network, whose input is power supply execution state data and output is the Q value of each power supply strategy; S632. Use an experience replay mechanism to store state transition samples; S633. Update network parameters by minimizing time difference error; S634. Use the target network to improve algorithm stability; S635. When the Q-value function converges, output the optimized policy parameters; S64. Generate the optimized differentiated power supply strategy set. : ; in Indicates the first An optimized power supply strategy, Indicates the strategy index, This indicates the total number of strategies.

[0027] Updating strategies and implementing dynamic optimization adjustments in S7 includes the following steps: S71, Optimize the differentiated power supply strategy set Compare and analyze with existing power supply strategies; S72. Calculate the strategy update benefit assessment value. When the benefit assessment value exceeds a preset threshold, execute the strategy update. The strategy update benefit assessment value is calculated using the following formula: ; in and The operating costs before and after the update are respectively. and These are the power supply reliability indicators before and after the update. and These are the load satisfaction indicators before and after the update. , , These are the weighting coefficients, and ; S73. Update the power supply strategy through the strategy management module of the power distribution network control system; S74. Establish a strategy update log to record the content, timing, and expected effects of strategy changes.

[0028] Updating the power supply policy through the policy management module in S73 includes the following steps: S731: Receives external command data from the distribution network dispatch center and internal early warning signals from the early warning processing module in real time, and generates a set of policy update trigger events; S732. Analyze the set of events triggered by the strategy update, call the preset rule base based on the event type and level, and generate a preliminary strategy adjustment plan. The rule base defines the mapping relationship between different event types and the adjustment parameters of the differentiated power supply strategy. S733. The initial policy adjustment scheme is input into the pre-trained deep Q-network reinforcement learning model. This model uses minimizing the load loss and maximizing the system stability as the reward function, and outputs the optimized policy adjustment action sequence. The deep Q-network reinforcement learning model is constructed in the following way: S7331. Construct a convolutional neural network to extract power supply status features; S7332, uses a fully connected layer to output the Q value of each adjustment action; S7333, the reward function is defined as follows: ; in For network loss power, Total power supply For voltage stability indicators, Rated voltage, For power supply reliability indicators, For maximum reliability, , , These are the weighting coefficients; S734. Based on the strategy, adjust the action sequence, perform simulation evaluation in a virtual simulation environment, and calculate multiple performance indicators of the simulation results. The virtual simulation environment is constructed in the following way: S7341. Constructing a digital twin model based on the distribution network topology; S7342, uses real-time data-driven simulation operation; S7343. Calculate various performance indicators of the simulation results using the power system analysis toolkit; S7344, performance indicators include: voltage deviation rate, frequency qualification rate, power supply reliability, and network loss rate; S735. When the overall performance index of the simulation result is better than the current operating strategy, a strategy update execution instruction is generated and automatically sent to the power supply control execution module. When the simulation result does not meet the requirements, it is fed back to the effect evaluation and optimization module to start a new round of optimization process.

[0029] The system includes: The data acquisition and processing module collects real-time and historical electricity consumption data from multiple load nodes in the distribution network, performs data fusion processing, and generates a load node electricity consumption dataset. The load classification and assessment module receives the load node electricity consumption dataset, performs load node electricity consumption feature extraction and load node classification and assessment, and generates a load node electricity consumption feature vector set and a load node classification and assessment result set. The differentiated power supply strategy generation module receives the load node hierarchical evaluation result set, combines it with the real-time operation status data of the distribution network, performs differentiated power supply strategy generation processing, and generates the target differentiated power supply strategy set. The power supply control execution module receives the target differentiated power supply strategy set, performs power supply control command parsing and distribution processing, and generates differentiated power supply execution result data for the power distribution network. The effect evaluation and strategy optimization module receives the execution result data of differentiated power supply in the distribution network, performs power supply effect evaluation and strategy optimization processing, and generates an optimized set of differentiated power supply strategies. The strategy management and update module receives the optimized set of differentiated power supply strategies, performs strategy benefit evaluation and dynamic update processing, and realizes dynamic optimization and adjustment of power supply strategies for the distribution network.

[0030] The operation steps of the load-level-based differentiated power supply method and system for distribution networks are as follows: I. Multi-source power consumption data acquisition and fusion processing The system uses smart meters and a data platform deployed in the power distribution network to synchronously collect real-time and historical electricity consumption data from each load node. Subsequently, these two types of heterogeneous time-series data undergo timestamp alignment, noise filtering, and data smoothing and fusion to generate a load node electricity consumption dataset, providing a reliable data foundation for subsequent analysis.

[0031] II. Multidimensional Electricity Consumption Feature Extraction and Dimensionality Reduction The system performs in-depth mining on the fused electricity consumption dataset to extract key feature parameters that characterize load characteristics, including load factor, peak-to-valley difference, stability, and importance. Based on this, principal component analysis is used to reduce the dimensionality of the high-dimensional features, refining core feature indicators to generate a feature vector set that characterizes the electricity consumption behavior of each load node.

[0032] III. Load Classification Assessment Based on Adaptive Optimization The system employs an improved fuzzy C-means clustering algorithm as the load grading model, with a regularization term introduced into its objective function to enhance the model's generalization ability. Subsequently, an adaptive particle swarm optimization algorithm is used to autonomously optimize key model parameters such as the number of clusters and fuzzy weights to obtain the best classification results. Finally, the load feature vector is input into the optimized model, outputting the grading evaluation results for each load node, achieving accurate and adaptive load level classification.

[0033] IV. Generation of Differentiated Power Supply Strategies under Multi-Objective Constraints After obtaining the load classification results, the system combines real-time power grid operation data on frequency, voltage, and load rate to construct a multi-objective optimization model centered on "minimizing operating costs," "maximizing power supply reliability," and "maximizing load satisfaction." An improved non-dominated sorting genetic algorithm is used to solve this model, yielding a set of Pareto optimal solutions. Finally, the entropy-weighted TOPSIS method is used to select the comprehensively optimal differentiated power supply strategy from the solution set, thereby formulating a scientific power supply scheme under multiple constraints.

[0034] V. Intelligent Execution and Effect Feedback of Power Supply Strategies The generated optimization strategy is parsed into specific control commands, which are then used by the distribution network intelligent control system to control loads, switches, and generation units. The system monitors the entire command execution process, collects actual power supply, power quality, and user feedback data, and generates strategy execution result reports, thus forming a closed loop from strategy formulation to implementation.

[0035] VI. Dynamic Optimization and Update of Strategies Based on Closed-Loop Learning The system evaluates the relative efficiency of implemented strategies using Data Envelopment Analysis (DEA) and employs deep reinforcement learning algorithms to self-optimize strategy parameters. The optimized new strategy is then compared with the old strategy in terms of effectiveness. After verification in a virtual simulation environment, performance improvements are automatically updated to the operating system, thereby achieving continuous self-evolution and dynamic adjustment of the power supply strategy and ensuring that the distribution network is always in an optimal operating state.

[0036] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0037] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A differentiated power supply method for distribution networks based on load grading, characterized in that: The method includes the following steps: S1. Collect real-time and historical electricity consumption data of multiple load nodes in the distribution network, and generate a load node electricity consumption dataset. S2. Based on the load node electricity consumption dataset, perform load node electricity consumption feature extraction processing to generate a load node electricity consumption feature vector set; S3. Perform load node hierarchical evaluation processing based on the load node electricity consumption feature vector set to generate a load node hierarchical evaluation result set. S4. Based on the load node hierarchical evaluation result set and combined with the real-time operation status data of the distribution network, perform differentiated power supply strategy generation processing to generate a target differentiated power supply strategy set. S5. Perform power distribution network power supply control execution processing according to the target differentiated power supply strategy set, and generate power distribution network differentiated power supply execution result data; S6. Based on the power distribution network differentiated power supply execution result data, perform power supply effect evaluation and strategy optimization processing to generate an optimized differentiated power supply strategy set; S7. Update the optimized differentiated power supply strategy set to the power distribution network control system to realize dynamic optimization and adjustment of the power distribution network strategy.

2. The differentiated power supply method for distribution networks based on load grading according to claim 1, characterized in that: The steps involved in collecting data and generating a dataset in S1 are as follows: S11. Collect real-time electricity consumption data of multiple load nodes in the distribution network through smart meters, including real-time voltage, current, power factor and active power data; S12. Extract historical electricity consumption data of the multiple load nodes from the distribution network data management platform, including historical daily load curves, monthly electricity consumption and annual peak electricity consumption data; S13. Perform data fusion processing on the real-time electricity consumption data and the historical electricity consumption data to generate the load node electricity consumption dataset. : ； in Indicates the first A set of electricity consumption data for each load node. Indicates the index of the load node. This indicates the total number of load nodes.

3. The method for differentiated power supply in a distribution network based on load grading according to claim 1, characterized in that: The electricity consumption feature extraction process in step S2 includes the following steps: S21. Data set of power consumption at the load nodes. Perform data preprocessing, including data cleaning, missing value imputation, and outlier handling; S22. Extract the power consumption characteristic parameters of each load node, including load factor characteristic parameters, peak-valley difference characteristic parameters, power consumption stability characteristic parameters, and importance characteristic parameters; S23. Based on principal component analysis, the extracted electricity consumption feature parameters are subjected to dimensionality reduction processing to generate the electricity consumption feature vector set of the load node. : ； in Indicates the first The electricity consumption characteristic vector of each load node, Indicates the index of the load node. This indicates the total number of load nodes.

4. The differentiated power supply method for distribution networks based on load grading according to claim 1, characterized in that: The graded evaluation process in S3 includes the following steps: S31. Construct a load grading assessment model. The model adopts an improved fuzzy C-means clustering algorithm, and its objective function is: ； in The objective function value of the fuzzy C-means clustering algorithm is... Represents the number of cluster centers. Indicates the first The load node belongs to the first Membership degree of a class For fuzzy weighting index, Indicates the first Cluster centers of classes For regularization parameters, This indicates the total number of load nodes. Indicates the first Electricity consumption characteristic vector of each load node; S32. Optimize the parameters of the load grading assessment model using an adaptive particle swarm optimization algorithm, including the number of clusters. and fuzzy weight index ; S33, Set the power consumption feature vector set of the load node Input the optimized load classification assessment model to generate the load node classification assessment result set. : ； in Indicates the first The classification results of each load node Indicates the index of the load node. This indicates the total number of load nodes.

5. The method for differentiated power supply in a distribution network based on load grading according to claim 1, characterized in that: The process of generating a differentiated power supply strategy in S4 includes the following steps: S41. Obtain real-time operating status data of the distribution network, including grid frequency, node voltage level, line load rate and power supply reliability indicators; S42. Establish a multi-objective optimization model to generate differentiated power supply strategies. The optimization objectives include: Objective 1: Minimize the total operating cost of the distribution network. ； in Indicates power generation capacity. Indicates the cost of electricity generation. This represents the total operating cost of the power distribution network; Objective 2: Maximize power supply reliability: ； in Indicates the first Power supply reliability indicators for each load node Indicates the weighting coefficient. This indicates the overall power supply reliability index; Objective 3: Maximize load satisfaction. ； in Indicates the first Satisfaction index of each load node This represents the load importance coefficient. This represents the overall load satisfaction index; S43. The improved non-dominated sorting genetic algorithm is used to solve the multi-objective optimization model and generate the Pareto optimal solution set. S44. Based on the entropy-weighted TOPSIS method, select the optimal solution from the Pareto optimal solution set to generate the target differentiated power supply strategy set. .

6. The method for differentiated power supply in a distribution network based on load grading according to claim 1, characterized in that: The power supply control execution process in S5 includes the following steps: S51. Analyze the target differentiated power supply strategy set. Generate a specific set of power supply control instructions; S52. Execute power supply control commands through the distribution network intelligent control system, including load control commands, switch operation commands, and generation dispatch commands; S53. Monitor the power supply control execution process in real time and collect power supply execution status data; S54. Generate differentiated power supply execution result data E for the power supply execution status data of the distribution network, including actual power supply, power supply quality and user satisfaction data.

7. The method for differentiated power supply in a distribution network based on load grading according to claim 1, characterized in that: The effect evaluation and strategy optimization process in S6 includes the following steps: S61. Establish a power supply performance evaluation index system, including technical indicators, economic indicators and social indicators; S62. Use Data Envelopment Analysis to evaluate the power supply effect and calculate the relative efficiency value of each power supply strategy; S63. Based on the evaluation results, a deep reinforcement learning algorithm is used to optimize the policy and update the policy parameters; S64. Generate the optimized differentiated power supply strategy set. : ； in Indicates the first An optimized power supply strategy, Indicates the strategy index, This indicates the total number of strategies.

8. The method for differentiated power supply in a distribution network based on load grading according to claim 1, characterized in that: The update strategy and dynamic optimization adjustment in S7 include the following steps: S71, The optimized differentiated power supply strategy set Compare and analyze with existing power supply strategies; S72. Calculate the benefit assessment value for strategy update. When the benefit assessment value exceeds the preset threshold, execute strategy update. S73. Update the power supply strategy through the strategy management module of the power distribution network control system; S74. Establish a strategy update log to record the content, timing, and expected effects of strategy changes.

9. The method for differentiated power supply in a distribution network based on load grading according to claim 8, characterized in that: The steps involved in updating the power supply policy via the policy management module in S73 are as follows: S731: Receives external command data from the distribution network dispatch center and internal early warning signals from the early warning processing module in real time, and generates a set of policy update trigger events; S732. Parse the set of policy update triggering events, call the preset rule base based on the event type and level, and generate a preliminary policy adjustment plan. The rule base defines the mapping relationship between different event types and differentiated power supply policy adjustment parameters. S733. The preliminary policy adjustment scheme is input into a pre-trained deep Q-network reinforcement learning model, which uses minimizing load loss and maximizing system stability as the reward function and outputs an optimized policy adjustment action sequence. S734. Adjust the action sequence based on the strategy, perform simulation evaluation in a virtual simulation environment, and calculate multiple performance indicators of the simulation results; S735. When the overall performance index of the simulation result is better than the current operating strategy, a strategy update execution instruction is generated and automatically sent to the power supply control execution module. When the simulation result does not meet the requirements, it is fed back to the effect evaluation and optimization module to start a new round of optimization process.

10. A load-level-based differentiated power supply system for distribution networks, used to implement the load-level-based differentiated power supply method for distribution networks as described in any one of claims 1-9, characterized in that: The system includes: The data acquisition and processing module collects real-time and historical electricity consumption data from multiple load nodes in the distribution network, performs data fusion processing, and generates a load node electricity consumption dataset. The load classification and assessment module receives the load node electricity consumption dataset, performs load node electricity consumption feature extraction processing and load node classification and assessment processing, and generates a load node electricity consumption feature vector set and a load node classification and assessment result set. The differentiated power supply strategy generation module receives the load node hierarchical evaluation result set, combines it with the real-time operation status data of the distribution network, performs differentiated power supply strategy generation processing, and generates a target differentiated power supply strategy set. The power supply control execution module receives the target differentiated power supply strategy set, performs power distribution network power supply control command parsing and distribution processing, and generates power distribution network differentiated power supply execution result data. The effect evaluation and strategy optimization module receives the differentiated power supply execution result data of the distribution network, performs power supply effect evaluation and strategy optimization processing, and generates an optimized differentiated power supply strategy set. The strategy management and update module receives the optimized differentiated power supply strategy set, performs strategy benefit evaluation and dynamic update processing, and realizes dynamic optimization and adjustment of the power supply strategy of the distribution network.

Images