Substation site selection method and system based on artificial intelligence

Abstract

The invention discloses a substation site selection method and system based on artificial intelligence, and the method comprises the steps: carrying out the rasterization of a target region, obtaining a plurality of grid units, and calculating a load density feature value of each grid unit based on the expansion planning data of the target region; inputting the weather forecast data and the load density characteristic value of each grid unit into a load prediction model to obtain a load prediction data set; clustering the load prediction data set to obtain a plurality of clustering areas, and calculating the load density of each clustering area; constructing a Voronoi diagram of the target region based on the clustering region and the load density, and obtaining initial candidate points based on the Voronoi diagram; and constructing a multi-objective optimization model of the substation site selection, and inputting the initial candidate points into the multi-objective optimization model for solving to obtain the substation site selection of the target area. According to the substation site selection method, the power supply reliability of the substation is improved.

Description

Technical Field

[0001] This application relates to the technical field of power system planning, and in particular, to a substation site selection method and system based on artificial intelligence. Background Art

[0002] In the expanding urban areas, the distribution network framework planning faces complex challenges in substation site selection.

[0003] Traditional substation site selection mostly relies on simple load statistical data and only focuses on the average load size of the region. For example, in the planning of some small and medium-sized cities, the substation location is often determined simply based on the historical total electricity consumption of the administrative region. This method ignores the impact of industrial layout differences, uneven population density distribution, and meteorological factors on the load.

[0004] Therefore, how to improve the existing substation site selection method to improve the power supply reliability of the substation has become an urgent technical problem for those skilled in the art. Summary of the Invention

[0005] This application provides a substation site selection method and system based on artificial intelligence to solve the technical problem of how to improve the existing substation site selection method, and achieve the effect of improving the power supply reliability of the substation.

[0006] To solve the above technical problem, an embodiment of this application provides a substation site selection method based on artificial intelligence, including: Perform grid processing on the target area to obtain a certain number of grid cells, and calculate the load density characteristic value of each grid cell based on the obtained expansion planning data of the target area; Input the obtained meteorological forecast data and load density characteristic values of each grid cell into a pre-constructed load prediction model for prediction to obtain the load prediction data set of the target area; Cluster the load prediction data set to obtain a certain number of cluster areas, and calculate the load density of each cluster area; Construct a Voronoi diagram of the target area based on the cluster areas and the load density, and obtain the initial candidate points for substation site selection in the target area based on the Voronoi diagram; Construct a multi-objective optimization model for substation site selection in the target area, and input at least the initial candidate points into the multi-objective optimization model for solution to obtain the substation site selection in the target area.

[0007] As a preferred solution, the expansion planning data includes industrial layout data, population density data, and economic index data; Calculating the load density eigenvalue of each grid cell based on the obtained expansion planning data of the target area includes: According to the obtained industrial layout data, population density data, and economic index data of each grid cell, calculate the load density eigenvalue corresponding to each grid cell, expressed as: Wherein, The load density eigenvalue of the th grid cell, Is the quantization value of the industrial layout data, Is the weight of the industrial layout data, Is the quantization value of the population density data, Is the weight of the population density data, Is the quantization value of the economic index data, Is the weight of the economic index data.

[0008] As one of the preferred solutions, before inputting the obtained weather forecast data and load density eigenvalues of each grid cell into the pre-constructed load prediction model for prediction, it further includes: Construct an initial load prediction model based on LSTM; Train the initial load prediction model based on the obtained historical weather data, historical load data, and historical load density eigenvalue data. During the training process, calculate the loss function according to the load prediction value obtained through forward propagation and the historical load data, and update the parameters of the initial load prediction model according to the value of the loss function to obtain a trained load prediction model.

[0009] As one of the preferred solutions, the clustering of the load prediction data set to obtain a certain number of clustering regions includes: Cluster the load prediction data set based on the improved K-means algorithm, which includes randomly selecting several sample data in the load prediction data set as the clustering centers; During the calculation process, calculate the distance from each sample data in the load prediction data set to the clustering center, and divide the sample data into the clustering cluster where the nearest clustering center is located according to the calculation result, and recalculate the mean value of the sample data in each clustering cluster as the new clustering center, where the distance includes the spatial distance and the feature distance; Repeat the calculation process until the clustering centers no longer change, and obtain a certain number of initial clustering regions; Calculate the silhouette coefficient of each sample data, evaluate the clustering quality of the initial clustering regions according to the silhouette coefficient, perform a secondary division on the initial clustering regions according to the result of the clustering quality evaluation, and obtain the clustering regions according to the result of the secondary division.

[0010] As one of the preferred solutions, the calculating the load density of each clustering region includes: Obtain the total load and the area of the region of all grid cells within the clustering region; Calculate the initial load density of the clustering region according to the total load and the area of the region; Introduce a load distribution non-uniformity index to correct the initial load density to obtain the load density of the clustering region, where the non-uniformity index is the standard deviation or coefficient of variation of the load within the region.

[0011] As one of the preferred solutions, the constructing the Voronoi diagram of the target region based on the clustering region and the load density includes: Use the centroid of the clustering region as the initial seed point to construct an initial Voronoi diagram; Adjust the boundary surface of the initial Voronoi diagram according to the load density, and the adjustment is designed to make the boundary surface shift towards the clustering region with a lower load density; Optimize the position of the initial seed points in the adjusted initial Voronoi diagram based on a genetic algorithm to obtain the Voronoi diagram of the target region, and the optimization is designed to make the load intensity of each Voronoi polygon balanced.

[0012] As one of the preferred solutions, the obtaining the initial candidate points for the substation location in the target region based on the Voronoi diagram includes: Calculate the deviation rate of the area of each Voronoi polygon in the Voronoi diagram from the average area; If the deviation rate exceeds the preset deviation rate range, re-optimize the position of the initial seed points based on a genetic algorithm; if the deviation rate is within the deviation rate range, use the optimized seed point position as the initial candidate point.

[0013] As one of the preferred solutions, the constructing the multi-objective optimization model for the substation location in the target region is expressed as: Among them, are the respective weight coefficients, is the maximum allowable value of electromagnetic radiation, is the maximum allowable value of noise, is the sum of the absolute differences between the actual power flow and the ideal power flow, is the total construction cost of the substation, is the total operating cost of the substation, is the average power outage time, is the average power outage frequency, is the total length of all transmission lines between substations, is the comprehensive optimization objective, is the economic objective, is the reliability objective, is the environmental impact objective, is the power grid structure objective.

[0014] As one of the preferred solutions, at least inputting the initial candidate points into the multi-objective optimization model for solution to obtain the substation site selection in the target area includes: Taking the initial candidate points as the initial population individuals and inputting them into the multi-objective optimization model for solution based on the improved genetic algorithm to obtain a candidate solution set; wherein the candidate solution set corresponds to the set of position coordinates of the candidate substations; Sorting the candidate solution set based on the TOPSIS method and screening according to the sorting result to obtain the substation site selection in the target area.

[0015] Another embodiment of the present application provides an artificial intelligence-based substation site selection system, including: A calculation module, configured to perform rasterization processing on the target area to obtain a certain number of raster units, and calculate the load density characteristic value of each raster unit based on the obtained expansion planning data of the target area; A prediction module, configured to input the obtained meteorological forecast data and load density characteristic values of each raster unit into a pre-constructed load prediction model for prediction to obtain a load prediction data set of the target area; A clustering module, configured to cluster the load prediction data set to obtain a certain number of clustering regions, and calculate the load density of each clustering region; A construction module, configured to construct a Voronoi diagram of the target area based on the clustering area and the load density, and obtain initial candidate points for the substation location in the target area based on the Voronoi diagram; A generation module, configured to construct a multi-objective optimization model for the substation location in the target area, and input at least the initial candidate points into the multi-objective optimization model for solution to obtain the substation location in the target area.

[0016] Compared with the prior art, the beneficial effects of the embodiments of the present application are at least one of the following: (1) In the present application, by clustering the load prediction data set, different clustering areas are divided, and the load density of each area is calculated. On this basis, a Voronoi diagram is constructed. This diagram can naturally divide the target area into multiple sub-areas according to the load distribution characteristics. The distance from the points in each sub-area to the nearest clustering center (i.e., the potential load concentration point) is the shortest. The initial candidate points obtained based on the Voronoi diagram are naturally distributed in and around the load concentration area. Compared with randomly selecting candidate points, the search space in the subsequent optimization process is greatly reduced, the site selection efficiency is improved, and the substation location that meets the requirements can be found faster.

[0017] (2) The entire site selection process of the present application is based on a large amount of data processing and analysis, from the expansion planning data, meteorological data to the load prediction data of the target area, etc., providing a solid data basis for decision-making. Through scientific models and algorithms to process these data, the obtained substation location is more scientific and reasonable. A reasonable substation location helps to optimize the power grid layout, make the power supply more balanced and efficient, reduce the operation risk of the power grid, ensure the long-term stable operation of the power grid, and provide reliable power guarantee for the regional economic development and residents' lives. Description of the Drawings

[0018] Figure 1 It is a schematic flowchart of a substation location method in one embodiment of the present application; Figure 2 It is a schematic diagram of a substation location system based on artificial intelligence in one embodiment of the present application. Detailed Embodiments

[0019] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present application.

[0020] In the description of this application, the terms "first", "second", "third", etc. are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first", "second", "third", etc. may explicitly or implicitly include one or more of such features. In the description of this application, unless otherwise stated, the meaning of "a plurality" is two or more.

[0021] In the description of this application, it should be noted that unless otherwise clearly specified and defined, the terms "installed", "connected", "connected to" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. The terms "vertical", "horizontal", "left", "right", "up", "down" and similar expressions used herein are only for the purpose of illustration and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be construed as a limitation to this application. The term "and / or" used herein includes any and all combinations of one or more of the related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0022] In the description of this application, it should be noted that unless otherwise defined, all technical and scientific terms used in this application have the same meaning as those commonly understood by those skilled in the technical field to which this application belongs. The terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0023] An embodiment of this application provides a substation site selection method based on artificial intelligence. Specifically, please refer to Figure 1 , Figure 1 which shows a schematic flow diagram of the substation site selection method in one of the embodiments of this application, including S1 - S5: S1: Perform grid processing on the target area to obtain a number of grid units, and calculate the load density characteristic value of each grid unit based on the obtained expansion planning data of the target area; Preferably, in an embodiment of this application, the expansion planning data includes industrial layout data, population density data, and economic indicator data; Calculating the load density characteristic value of each grid unit based on the obtained expansion planning data of the target area includes: According to the industrial layout data, population density data, and economic indicator data obtained for each grid cell, calculate the load density eigenvalue corresponding to each grid cell, expressed as: Wherein, The load density eigenvalue of the nth grid cell, is the quantization value of the industrial layout data, is the weight of the industrial layout data, is the quantization value of the population density data, is the weight of the population density data, is the quantization value of the economic indicator data,

[0024] Among them, the industrial layout data is obtained through channels such as relevant government planning documents and industrial park statistical materials. These data detail the distribution of various industries in the target area. For example, a grid cell is an industrial park, containing different industrial types such as electronics manufacturing and machining; or it is a commercial area, concentrated with various commercial stores, office buildings, etc. The population density data can be obtained from census materials and population distribution statistical data of urban planning departments. For each grid cell, the number of resident population is clarified, and the population density is calculated based on the area of the grid cell. The economic indicator data is sourced from local economic statistical yearbooks, reports of economic development departments, etc. These indicators may include gross regional product, total industrial output value, commercial sales volume, etc. For each grid cell, the corresponding economic indicator value is obtained.

[0025] In order to uniformly incorporate different types of data into the model for calculating the load density eigenvalue, it is necessary to perform quantization processing on the industrial layout data, population density data, and economic indicator data. The setting of weights determines the relative importance of the industrial layout data, population density data, and economic indicator data in calculating the load density eigenvalue. The determination of weights usually requires comprehensive consideration of the actual situation of the target area and expert experience.

[0026] After completing data acquisition, quantization, and weight determination, calculate the load density eigenvalue of each grid cell according to the formula.

[0027] S2: Input the obtained meteorological forecast data and load density eigenvalues of each grid cell into a pre-constructed load forecasting model for prediction to obtain the load forecasting dataset of the target area; Preferably, in an embodiment of the present application, before inputting the obtained meteorological forecast data and load density eigenvalues of each grid cell into a pre-constructed load forecasting model for prediction, it further includes: Construct an initial load prediction model based on LSTM; Train the initial load prediction model based on the obtained historical meteorological data, historical load data, and historical load density eigenvalue data. During the training process, calculate the loss function according to the load prediction value obtained through forward propagation and the historical load data, and update the parameters of the initial load prediction model according to the value of the loss function to obtain a trained load prediction model.

[0028] Among them, the long short-term memory network (LSTM) is a special recurrent neural network (RNN) that can effectively handle the long-term dependence problem in time series data and is very suitable for the load prediction task.

[0029] After constructing the initial load prediction model, obtain the meteorological data of the target area in the past period from channels such as the database of the meteorological department and professional meteorological data platforms, including information such as daily temperature, humidity, rainfall, and sunshine duration. Obtain the corresponding historical power load data of the target area from the power consumption information collection system of the power company. These data record the actual power consumption load values at different time points (such as every hour, every day). Based on the data of the industrial layout, population density, economic indicators, etc. of the target area in the historical period, calculate the load density eigenvalue data of each grid unit at the historical time point according to the calculation method introduced before.

[0030] Preprocess the historical data. The preprocessing specifically includes data cleaning, data normalization, and data partitioning. Before inputting the training data into the initial load prediction model, initialize the parameters in the model (such as the weight matrix and bias vector in the LSTM unit).

[0031] Input the historical meteorological data and historical load density eigenvalue data in the training set into the initialized LSTM initial load prediction model in sequence. The data enters from the input layer in chronological order and is processed by the LSTM units in the hidden layer. Each LSTM unit updates the unit state and outputs the processed feature information through the control of the forget gate, input gate, and output gate according to the input data and the state of the previous moment, and finally obtains the load prediction value at the output layer.

[0032] Compare the load prediction value obtained through forward propagation with the corresponding historical load data in the training set, and calculate the loss function. The commonly used loss function is the mean square error (MSE). According to the value of the loss function, calculate the gradient of the loss function with respect to the model parameters through the backpropagation algorithm. The gradient represents the degree of influence of parameter changes on the value of the loss function. By continuously repeating the processes of forward propagation, calculating the loss function, backpropagation, and parameter update, gradually adjust the model parameters to continuously reduce the value of the loss function and improve the prediction ability of the model.

[0033] S3: Cluster the load forecasting dataset to obtain a number of clustering regions, and calculate the load density of each clustering region. Preferably, in an embodiment of the present application, clustering the load forecasting dataset to obtain a number of clustering regions includes: Cluster the load forecasting dataset based on an improved K-means algorithm, which includes randomly selecting several sample data in the load forecasting dataset as clustering centers; During the calculation process, calculate the distance from each sample data in the load forecasting dataset to the clustering center, and divide the sample data into the clustering cluster where the nearest clustering center is located according to the calculation result, and recalculate the mean value of the sample data in each clustering cluster as the new clustering center. Here, the distance includes spatial distance and feature distance; Repeat the calculation process until the clustering center no longer changes, and obtain a number of initial clustering regions; Calculate the silhouette coefficient of each sample data, evaluate the clustering quality of the initial clustering regions according to the silhouette coefficient, perform a secondary division on the initial clustering regions according to the result of the clustering quality evaluation, and obtain the clustering regions according to the result of the secondary division.

[0034] Randomly select several sample data from the load forecasting dataset as the initial clustering centers. The determination of the number of clustering centers (K value) needs to comprehensively consider the data distribution characteristics and actual application requirements. For example, if there are several regions with significantly different electricity consumption characteristics in the target area (such as industrial areas, commercial areas, residential areas). Suppose the dataset contains the load forecasting values of each grid cell at different time points and related features (such as a multi-dimensional data formed by a combination of meteorological data, load density characteristic values, etc.), and the randomly selected sample data are used as clustering centers, which represent the initial cores of different categories in the data space.

[0035] For each sample data in the load forecasting dataset, calculate its distance to each clustering center. The distance here includes spatial distance and feature distance, and is measured by comprehensively considering the positional relationship and feature similarity in the multi-dimensional data space. For example, the Euclidean distance formula is used to calculate the spatial distance; for the feature distance, weights can be set according to the importance of the data features, and the differences in different feature dimensions are weighted and summed to obtain the comprehensive feature distance. According to the calculated distance, each sample data is divided into the clustering cluster where the nearest clustering center is located.

[0036] After all the sample data are partitioned, recalculate the mean of the sample data within each cluster and use it as the new cluster center. By continuously repeating the process of calculating distances, partitioning clusters, and updating the cluster centers, the cluster centers will gradually move to locations that better represent the core positions of the clusters. Continue the above calculation process until the cluster centers no longer change. At this time, the obtained several clusters form the initial clustering regions.

[0037] For each sample data, calculate its silhouette coefficient. The silhouette coefficient is used to measure the tightness of the sample data within its belonging cluster and the degree of separation from other clusters. Calculate the average silhouette coefficient of all sample data within each initial clustering region to evaluate the clustering quality of the entire clustering region. The higher the average silhouette coefficient, the better the consistency of the sample data within the clustering region and the better the clustering effect. According to the results of the clustering quality evaluation, perform a secondary partition on the initial clustering regions. For clustering regions with a relatively low average silhouette coefficient, there may be a situation of mixed sample data and further subdivision is required. For example, if a clustering region contains sample data that should originally belong to different electricity consumption patterns (such as a mixture of industrial electricity and residential electricity), resulting in a low average silhouette coefficient, the clustering parameters can be readjusted (such as increasing the number of cluster centers), and the improved K-means algorithm can be used again to cluster the data within this region and divide it into more reasonable clustering regions. Through the secondary partition, the final clustering regions are obtained, and these clustering regions can more accurately reflect the regional division of different electricity consumption characteristics within the target region.

[0038] Preferably, in an embodiment of the present application, calculating the load density of each clustering region includes: Obtain the total load and the area of the region for all grid cells within the clustering region; Calculate the initial load density of the clustering region based on the total load and the area; Introduce a load distribution non-uniformity index to correct the initial load density to obtain the load density of the clustering region, where the non-uniformity index is the standard deviation or coefficient of variation of the load within the region.

[0039] For each clustering region, determine all the grid cells it contains. Through the previous load prediction data set and the rasterization processing information, summarize the load prediction values of all grid cells within this clustering region to obtain the total load. At the same time, calculate the area of the clustering region based on the area information of the grid cells and the number of grid cells covered by the clustering region. Calculate the initial load density of the clustering region according to the total load and the area.

[0040] In one embodiment of the present application, in order to more accurately reflect the actual situation of the load distribution within the clustering region, a load distribution non-uniformity index is introduced to correct the initial load density. Among them, the standard deviation reflects the degree of dispersion of the data. The larger the standard deviation, the more uneven the load distribution.

[0041] S4: Construct a Voronoi diagram of the target region based on the clustering region and the load density, and obtain the initial candidate points for the substation location in the target region based on the Voronoi diagram; Preferably, in one embodiment of the present application, constructing a Voronoi diagram of the target region based on the clustering region and the load density includes: Using the centroid of the clustering region as the initial seed point to construct an initial Voronoi diagram; Adjust the boundary surface of the initial Voronoi diagram according to the load density, and the adjustment is designed to make the boundary surface shift towards the clustering region with a lower load density; Optimize the positions of the initial seed points in the adjusted initial Voronoi diagram based on the genetic algorithm to obtain the Voronoi diagram of the target region, and the optimization is designed to balance the load intensity of each Voronoi polygon.

[0042] After completing the clustering region division and load density calculation, for each clustering region, its centroid needs to be calculated. The calculation method of the centroid will vary according to the geometric shape of the clustering region and the characteristics of the grid cells it contains. If the clustering region consists of multiple regular grid cells, the geometric centers of all grid cells can be used as the centroid of the clustering region. For example, for a clustering region composed of square grid cells, the horizontal and vertical coordinates of these grid cells in the plane rectangular coordinate system are averaged respectively, and the obtained coordinate point is the centroid of the clustering region. Using the centroids of these clustering regions as the initial seed points, an initial Voronoi diagram is constructed using the Voronoi diagram generation algorithm. The Voronoi diagram generation algorithm will divide the plane space into multiple polygon regions according to the given seed points. The distance from any point within each polygon region to the corresponding seed point of that region is less than the distance to other seed points. In this process, the centroid of each clustering region becomes the core point of the corresponding Voronoi polygon, and the spatial division range of each clustering region is preliminarily determined.

[0043] Considering the importance of load density in substation location, it is necessary to adjust the boundary surface of the initial Voronoi diagram to make it more conform to the load distribution situation. The specific approach is to make the boundary surface shift towards the clustering region with a lower load density. For two adjacent clustering regions A and B, there is an interface between their corresponding Voronoi polygons. Let the load density of region A be DA and the load density of region B be DB. If DA > DB, then move this interface a certain distance in the direction of region B. The determination of the moving distance can be calculated according to the degree of difference in load density.

[0044] The genetic algorithm is an optimization algorithm that simulates the biological evolution process and is used to further optimize the positions of the initial seed points to balance the load intensity of each Voronoi polygon.

[0045] Preferably, in an embodiment of the present application, obtaining the initial candidate points for the substation site selection in the target area based on the Voronoi diagram includes: Calculating the deviation rate of the area of each Voronoi polygon in the Voronoi diagram from the average area; If the deviation rate exceeds the preset deviation rate range, then re-optimize the position of the initial seed point based on the genetic algorithm; if the deviation rate is within the deviation rate range, then use the optimized seed point position as the initial candidate point.

[0046] For the constructed Voronoi diagram, calculate the area of each Voronoi polygon. The deviation rate reflects the degree of deviation of the area of each Voronoi polygon from the average area. Preset a deviation rate range, such as [0, 0.1]. If the deviation rate of a certain Voronoi polygon exceeds the preset range, it indicates that the area of this polygon is quite different from the average area, which may lead to uneven distribution of substation site selection. At this time, re-optimize the position of the initial seed point based on the genetic algorithm. The optimization process is similar to the previous one. Through operations such as coding, fitness function design, selection, crossover, and mutation, adjust the position of the seed point, reconstruct the Voronoi diagram, and calculate the area and deviation rate of each Voronoi polygon again until the deviation rates of all Voronoi polygons are within the preset range.

[0047] When the deviation rates of all Voronoi polygons are within the deviation rate range, it indicates that the area distribution of the Voronoi diagram polygons is relatively uniform. At this time, use the optimized seed point position as the initial candidate point. These initial candidate points are relatively evenly distributed in space and consider the load density factor, providing a reasonable starting point for the multi-objective optimization of the subsequent substation site selection.

[0048] S5: Construct a multi-objective optimization model for the substation site selection in the target area, and at least input the initial candidate points into the multi-objective optimization model for solution to obtain the substation site selection in the target area.

[0049] Preferably, in an embodiment of the present application, the multi-objective optimization model for the substation site selection in the target area is expressed as: Among them, are the respective weight coefficients, is the maximum allowable value of electromagnetic radiation, is the maximum allowable value of noise, is the sum of the absolute differences between the actual power flow and the ideal power flow, is the total construction cost of the substation, is the total operation cost of the substation, is the average power outage time, is the average power outage frequency, is the total length of all transmission lines between substations, is the comprehensive optimization goal, is the economic goal, is the reliability goal, is the environmental impact goal, is the power grid structure goal.

[0050] Specifically, the multi-objective optimization model comprehensively considers multiple key factors to determine the optimal substation site selection plan. Economic goal (C): It covers the total construction cost of the substation (including equipment procurement, land acquisition, construction costs, etc.) and the total operation cost (such as energy loss, equipment maintenance costs, etc.). The construction cost can be obtained by calculating different equipment prices, land market prices, and construction cost standards; the operation cost needs to consider factors such as the service life of the equipment, energy price fluctuations, and maintenance cycles and costs. Reliability goal (R): It is measured by the average power outage time (SAIDI) and the average power outage frequency (SAIFI). The average power outage time reflects the average duration of power outages for users within the statistical period, and the average power outage frequency reflects the number of power outages within the statistical period. Reducing these two indicators can improve the reliability of power supply and ensure normal power consumption for users. Environmental impact objective (E): It is reflected by restricting the electromagnetic radiation intensity not to exceed the allowable maximum value and the noise not to exceed the allowable maximum value. When selecting the site of the substation, the sensitivity of the surrounding environment needs to be considered, such as areas sensitive to electromagnetic radiation and noise like residential areas, schools, hospitals, etc., to ensure that the impact of the substation operation on the environment and residents' lives is within an acceptable range. Power grid structure objective (S): The sum of the absolute differences between the actual power flow and the ideal power flow and the total length of all transmission lines between substations are used as measurement indicators. The difference between the actual power flow and the ideal power flow reflects the stability of the power grid operation and the rationality of load distribution; the total length of the transmission lines is closely related to construction costs, power transmission losses, etc. A shorter total length of the transmission lines helps to reduce costs and losses.

[0051] Preferably, in an embodiment of the present application, at least the initial candidate points are input into the multi-objective optimization model for solution to obtain the substation site selection in the target area, including: Taking the initial candidate points as the initial population individuals, input them into the multi-objective optimization model for solution based on the improved genetic algorithm to obtain a candidate solution set; where the candidate solution set corresponds to the set of position coordinates of the candidate substations; Sort the candidate solution set based on the TOPSIS method, and perform screening according to the sorting result to obtain the substation site selection in the target area.

[0052] Taking the initial candidate points obtained based on the Voronoi diagram as the initial population individuals and input them into the multi-objective optimization model. Each individual is composed of a set of position coordinates of the candidate substation, and these coordinates represent the potential substation site selection positions.

[0053] The improved genetic algorithm operations include selection operation, crossover operation, and mutation operation.

[0054] Selection operation: Select individuals with higher fitness from the current population as parents for breeding the next generation. In multi-objective optimization, the evaluation of fitness cannot be based solely on a single objective function, but multiple objectives need to be considered comprehensively. Here, a selection strategy based on Pareto dominance relationship and crowding distance is adopted. The Pareto dominance relationship is used to determine the superiority and inferiority relationship of individuals in multiple objectives. If individual A is not worse than individual B in all objectives and is superior to individual B in at least one objective, then it is said that individual A dominates individual B. The crowding distance measures the degree of crowding of an individual on its Pareto front. The larger the distance, the sparser the distribution of solutions around the individual, indicating better diversity. Through this selection strategy, individuals that perform well in multiple objectives and are evenly distributed are preferentially selected as parents.

[0055] Crossover operation: Perform crossover operation on the selected parent individuals to generate new offspring individuals. For individuals representing the location coordinates of substations, methods such as arithmetic crossover can be used.

[0056] Mutation operation: Mutate the genes (i.e., the location coordinates of substations) of individuals with a certain probability. Gaussian mutation and other methods can be used for the mutation operation to introduce new genes into the population, increase the diversity of the population, and prevent the algorithm from falling into a local optimal solution.

[0057] Iteration termination condition: Set the iteration termination condition, such as reaching the maximum number of iterations (e.g., 1000 times) or the optimal solution of the population not improving significantly for several consecutive generations (e.g., 50 generations). When the termination condition is met, the algorithm stops iterating and obtains a candidate solution set, which contains a series of substation site selection schemes that achieve a better balance on multiple objectives.

[0058] The TOPSIS method (Technique for Order Preference by Similarity to an Ideal Solution) is the Technique for Order Preference by Similarity to an Ideal Solution. It ranks the candidate solutions by calculating the distances between each solution and the ideal solution and the negative ideal solution. The ideal solution is the solution that achieves the best in all objectives, and the negative ideal solution is the solution that achieves the worst in all objectives.

[0059] The calculation steps of the TOPSIS method include standardizing the objective values of each solution in the candidate solution set to eliminate the influence of different objective dimensions; for each objective, finding the optimal value and the worst value in the candidate solution set; calculating the Euclidean distances between each solution and the ideal solution and the negative ideal solution; and calculating the relative closeness degree of each solution according to the Euclidean distance.

[0060] Rank the candidate solution set according to the relative closeness degree calculated by the TOPSIS method, and select the solutions with higher relative closeness degrees as the preferred locations of the substations. For example, select the top 3 solutions in terms of relative closeness degree. After further on-site investigation, feasibility analysis, etc., finally determine the substation site selection in the target area. These locations achieve a better balance on multiple objectives such as economy, reliability, environmental impact, and power grid structure, providing a strong basis for the scientific site selection of substations.

[0061] Another embodiment of the present application provides an artificial intelligence-based substation site selection system. Specifically, please refer to Figure 2 , Figure 2 which is shown as the schematic diagram of the artificial intelligence-based substation site selection system in one of the embodiments of the present application, and it includes: A calculation module 11, configured to perform rasterization processing on a target area to obtain a number of raster cells, and calculate a load density eigenvalue of each raster cell based on the obtained expansion planning data of the target area; A prediction module 12, configured to input the obtained meteorological forecast data and load density eigenvalues of each raster cell into a pre-constructed load prediction model for prediction to obtain a load prediction data set of the target area; A clustering module 13, configured to cluster the load prediction data set to obtain a number of clustering regions, and calculate the load density of each clustering region; A construction module 14, configured to construct a Voronoi diagram of the target area based on the clustering regions and the load density, and obtain an initial candidate point for substation site selection in the target area based on the Voronoi diagram; A generation module 15, configured to construct a multi-objective optimization model for substation site selection in the target area, and input at least the initial candidate point into the multi-objective optimization model for solution to obtain the substation site selection in the target area.

[0062] Compared with the prior art, the beneficial effects of the embodiments of the present application are at least one of the following: (1) In the present application, by clustering the load prediction data set, different clustering regions are divided, and the load density of each region is calculated. On this basis, a Voronoi diagram is constructed. This diagram can naturally divide the target area into multiple sub-regions according to the load distribution characteristics. The distance from each point in each sub-region to the nearest clustering center (i.e., the potential load concentration point) is the shortest. The initial candidate points obtained based on the Voronoi diagram are naturally distributed in and around the load concentration areas. Compared with randomly selecting candidate points, the search space in the subsequent optimization process is greatly reduced, the site selection efficiency is improved, and a substation location that meets the requirements can be found faster.

[0063] (2) The entire site selection process of the present application is based on a large amount of data processing and analysis, from the expansion planning data, meteorological data to the load prediction data of the target area, providing a solid data basis for decision-making. By processing these data through scientific models and algorithms, the obtained substation site selection is more scientific and reasonable. A reasonable substation site selection helps to optimize the power grid layout, make the power supply more balanced and efficient, reduce the operation risk of the power grid, ensure the long-term stable operation of the power grid, and provide reliable power guarantee for regional economic development and residents' lives.

[0064] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A substation site selection method based on artificial intelligence, characterized in that: include: Performing rasterization processing on the target area to obtain a number of grid cells, and calculating the load density characteristic value of each grid cell based on the acquired expansion planning data of the target area; Inputting the obtained weather forecast data of each grid unit and the load density characteristic value into a pre-built load prediction model for prediction, thereby obtaining a load prediction data set for the target area; Clustering the load forecasting data set to obtain a number of clustering regions, and calculating the load density of each clustering region; Constructing a Voronoi diagram of the target area based on the cluster area and the load density, and obtaining initial candidate points for substation site selection in the target area based on the Voronoi diagram; A multi-objective optimization model for substation site selection in the target area is constructed, and at least the initial candidate points are input into the multi-objective optimization model for solving to obtain the substation site selection in the target area.

2. The method for selecting a substation site based on artificial intelligence according to claim 1, characterized in that: The expansion planning data include industrial layout data, population density data and economic indicator data; The calculating of the load density characteristic value of each grid unit based on the acquired expansion planning data of the target area includes: According to the acquired industrial layout data, population density data and economic indicator data of each grid unit, the load density characteristic value corresponding to each grid unit is calculated, which is expressed as: in, No. The load density characteristic value of each grid cell is is the quantitative value of industrial layout data, The weight of the industry layout data, is the quantitative value of population density data, is the weight of population density data, is the quantitative value of economic indicator data, is the weight of economic indicator data.

3. The method for selecting a substation site based on artificial intelligence according to claim 1, characterized in that: Before inputting the acquired weather forecast data of each grid unit and the load density characteristic value into the pre-built load prediction model for prediction, the method further includes: Construct an initial load forecasting model based on LSTM; The initial load forecasting model is trained based on the acquired historical meteorological data, historical load data and historical load density characteristic value data. During the training process, the loss function is calculated based on the load prediction value obtained by forward propagation calculation and the historical load data, and the parameters of the initial load forecasting model are updated according to the value of the loss function to obtain a trained load forecasting model.

4. The method for selecting a substation site based on artificial intelligence according to claim 1, characterized in that: The load forecasting data set is clustered to obtain a number of clustering regions, including: Clustering the load forecasting data set based on an improved K-means algorithm, including randomly selecting a number of sample data in the load forecasting data set as clustering centers; During the calculation process, the distance from each sample data in the load forecast data set to the cluster center is calculated, and the sample data is divided into clusters where the nearest cluster center is located according to the calculation result, and the mean value of the sample data of each cluster is recalculated as a new cluster center, wherein the distance includes spatial distance and characteristic distance; Repeat the calculation process until the cluster center does not change, and obtain a certain number of initial cluster regions; Calculate the silhouette coefficient of each sample data, perform clustering quality assessment on the initial clustering area according to the silhouette coefficient, perform secondary division on the initial clustering area according to the result of the clustering quality assessment, and obtain the clustering area according to the result of the secondary division.

5. The method for selecting a substation site based on artificial intelligence according to claim 1, characterized in that: The calculating the load density of each clustering area includes: Obtaining the total load and area of ​​all grid cells in the clustering area; Calculating the initial load density of the cluster region according to the total load and the area of ​​the region; The load distribution unevenness index is introduced to correct the initial load density to obtain the load density of the cluster area, wherein the unevenness index is the standard deviation or coefficient of variation of the load in the area.

6. The method for selecting a substation site based on artificial intelligence according to claim 1, characterized in that: The constructing the Voronoi diagram of the target area based on the clustering area and the load density includes: Taking the centroid of the clustering area as an initial seed point, constructing an initial Voronoi diagram; Adjusting the boundary surface of the initial Voronoi diagram according to the load density, wherein the adjustment is designed to shift the boundary surface toward a clustering area with a lower load density; The positions of initial seed points in the adjusted initial Voronoi diagram are optimized based on a genetic algorithm to obtain the Voronoi diagram of the target area. The optimization is designed to balance the load intensity of each Voronoi polygon.

7. The method for selecting a substation site based on artificial intelligence according to claim 1, characterized in that: The obtaining of initial candidate points for the substation site selection in the target area based on the Voronoi diagram includes: Calculating the deviation rate between the area of ​​each Voronoi polygon in the Voronoi diagram and the average area; If the deviation rate exceeds a preset deviation rate range, the initial seed point position is optimized again based on a genetic algorithm; if the deviation rate is within the deviation rate range, the optimized seed point position is used as the initial candidate point.

8. The method for selecting a substation site based on artificial intelligence according to claim 1, characterized in that: The multi-objective optimization model for substation site selection in the target area is constructed as follows: in, are the weight coefficients, is the maximum value allowed for electromagnetic radiation, is the maximum value allowed for noise, is the sum of the absolute differences between the actual tidal current and the ideal tidal current, is the total cost of substation construction, is the total operating cost of the substation, is the average power outage time, is the average power outage frequency, is the total length of all transmission lines between substations, For comprehensive optimization goals, For economic purposes, For reliability goals, For environmental impact targets, For the grid structure target.

9. The method for selecting a substation site based on artificial intelligence according to claim 1, characterized in that: The step of inputting at least the initial candidate points into the multi-objective optimization model for solving to obtain the substation site selection in the target area includes: Input the initial candidate points as initial population individuals into the multi-objective optimization model for solving based on an improved genetic algorithm to obtain a candidate solution set; wherein the candidate solution set corresponds to a location coordinate set of candidate substations; The candidate solution set is sorted based on the TOPSIS method, and is screened according to the sorting result to obtain the substation site selection in the target area.

10. A substation site selection system based on artificial intelligence, characterized in that: include: A calculation module, used for performing a gridding process on the target area to obtain a number of grid cells, and calculating a load density characteristic value of each grid cell based on the acquired expansion planning data of the target area; A prediction module, used for inputting the acquired weather forecast data of each grid unit and the load density characteristic value into a pre-built load prediction model for prediction, so as to obtain a load prediction data set of the target area; A clustering module, used for clustering the load forecasting data set to obtain a number of clustering areas, and calculating the load density of each clustering area; A construction module, used to construct a Voronoi diagram of the target area based on the cluster area and the load density, and obtain an initial candidate point for the substation site selection in the target area based on the Voronoi diagram; A generation module is used to construct a multi-objective optimization model for substation site selection in the target area, and at least input the initial candidate points into the multi-objective optimization model for solving to obtain the substation site selection in the target area.

Images