Charging pile construction planning method and system based on artificial intelligence

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a charging pile construction planning method and system based on artificial intelligence. The method comprises the following steps: acquiring existing power infrastructure data and multi-source heterogeneous data, wherein the multi-source heterogeneous data comprises traffic data, geographic data, vehicle data and charging behavior data; performing time sequence feature extraction on the multi-source heterogeneous data, and performing cross-time scale fusion on different data source features by using a preset hierarchical attention network model to obtain factor weight data; and constructing a long-term charging demand prediction model through a sequence-to-sequence model according to the factor weight data and the multi-source heterogeneous data. According to the site selection method based on the risk minimum set coverage problem, initial site selection scheme data is generated by combining multi-objective optimization of the coverage rate, the service radius and the accessibility, and more scientific charging pile layout is achieved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a charging pile construction planning method and system based on artificial intelligence. Background Art

[0002] The development of electric vehicles has put forward higher requirements for charging infrastructure, requiring more efficient and intelligent charging pile layout and management systems. Artificial Intelligence (AI) is a science that studies and develops theories, methods, technologies and application systems for simulating, expanding and extending human intelligence. Through sub-fields such as machine learning, natural language processing, and computer vision, AI technology enables computers to perform tasks that usually require human intelligence, such as recognizing images, understanding natural language, and making complex decisions. The AI-based charging pile construction planning method came into being against the backdrop of the rapid popularization of electric vehicles, accelerated energy transformation, and the construction of smart cities.

[0003] When planning the construction of charging piles, the following problems are often encountered: traditional methods often rely on historical data and simple models to predict charging demand, and are unable to accurately predict future charging demand, especially in the face of a rapidly changing market and technological environment; the construction of charging piles needs to consider the carrying capacity and load distribution of the power grid. Traditional methods are difficult to achieve accurate simulation and optimization when conducting capacity planning and grid impact assessment. Summary of the invention

[0004] Based on this, it is necessary for the present invention to provide a charging pile construction planning method and system based on artificial intelligence to solve at least one of the above technical problems.

[0005] To achieve the above purpose, a charging pile construction planning method based on artificial intelligence includes the following steps: Step S1: obtaining existing power infrastructure data and multi-source heterogeneous data, wherein the multi-source heterogeneous data includes traffic data, geographic data, vehicle data, and charging behavior data; extracting time series features from the multi-source heterogeneous data, and using a preset hierarchical attention network model to fuse features from different data sources across time scales, thereby obtaining factor weight data; Step S2: constructing a long-term charging demand prediction model based on factor weight data and multi-source heterogeneous data through a sequence-to-sequence model; performing long-term prediction of future charging demand based on the long-term charging demand prediction model, and performing uncertainty interval evaluation to obtain charging demand prediction data; Step S3: Perform charging pile site selection based on the risk minimum set coverage problem according to the charging demand forecast data and the power infrastructure data, and perform target optimization based on coverage, service radius and accessibility to obtain preliminary site selection plan data; Step S4: predicting the load curve of the site selection point according to the charging demand forecast data and the preliminary site selection plan data, thereby obtaining the charging pile configuration data of each site selection point; generating capacity planning data according to the charging pile configuration data and the existing energy storage conditions; and evaluating the impact of the new load on the power grid according to the power infrastructure data on the capacity planning data, thereby obtaining the power grid impact assessment data; Step S5: Generate grid constraint data based on the grid impact assessment data, and make balance corrections to the preliminary site selection plan data and charging pile configuration data based on charging demand and grid stability according to the grid constraint data, so as to obtain planning plan data for charging pile construction.

[0006] By acquiring and fusing multi-source heterogeneous data, the present invention can comprehensively utilize information such as traffic data, geographic data, vehicle data, and charging behavior data to provide a more comprehensive and multi-dimensional data foundation. Time series feature extraction can capture the trend and pattern of data changes over time, revealing the characteristics of data such as periodicity, trend and correlation. Through the preset hierarchical attention network model, the features of different time scales can be integrated to better understand the long-term and short-term impacts of the data and provide more accurate factor weight data. Through the sequence-to-sequence model, a charging demand prediction model can be established based on factor weight data and multi-source heterogeneous data to provide an accurate prediction of future charging demand. By evaluating the uncertainty interval, a credibility estimate of the charging demand prediction results can be provided to help decision makers understand the scope and possible deviations of the prediction. Combining charging demand prediction data and power infrastructure data, the risk minimum set coverage problem can be comprehensively considered on the premise of meeting charging demand to ensure the accuracy and feasibility of the site selection plan. Based on the optimization of coverage, service radius and accessibility, different goals can be weighed in the site selection process to improve the service effect and coverage of charging piles. Through the prediction of the load curve of the site selection point, the charging demand and load characteristics of each site selection point can be estimated, providing a basis for the configuration and capacity planning of charging piles. Based on the configuration data of the charging piles and the existing energy storage conditions, capacity planning can be carried out to evaluate the impact of the new load on the power grid. This helps to understand the impact of charging pile construction on the stability and reliability of the power grid and provide a reasonable power grid transformation plan. Based on the power grid impact assessment data, power grid constraint data can be generated to comprehensively consider the charging demand and power grid stability to ensure the coordination and balance between the construction of charging piles and the power grid. Based on the balance between charging demand and power grid stability, the preliminary site selection plan and charging pile configuration are revised to meet the power grid constraints and provide reliable charging services. In summary, the effects of the above steps include data synthesis, feature extraction, prediction accuracy, uncertainty assessment, comprehensive consideration, target optimization, load curve prediction, capacity planning and impact assessment, power grid constraint consideration, and balance correction. These effects help to improve the accuracy and reliability of charging demand prediction, optimize the site selection and configuration plan of charging piles, and ensure the feasibility of charging services and the stability of the power grid.

[0007] The present invention also provides an artificial intelligence-based charging pile construction planning system, which is used to execute the above-mentioned artificial intelligence-based charging pile construction planning method. The artificial intelligence-based charging pile construction planning system includes: The multi-source data fusion module is used to obtain existing power infrastructure data and multi-source heterogeneous data, including traffic data, geographic data, vehicle data, and charging behavior data; extract time series features from multi-source heterogeneous data, and use the preset hierarchical attention network model to fuse the features of different data sources across time scales to obtain factor weight data; The long-term demand forecasting module is used to build a long-term charging demand forecasting model based on factor weight data and multi-source heterogeneous data through a sequence-to-sequence model; based on the long-term charging demand forecasting model, long-term forecasts of future charging demand are made, and uncertainty intervals are evaluated to obtain charging demand forecast data; The risk minimization site selection module is used to select charging pile sites based on the risk minimum set coverage problem according to the charging demand forecast data and power infrastructure data, and to optimize the targets based on coverage, service radius and accessibility to obtain preliminary site selection plan data; The load impact assessment module is used to predict the load curve of the site selection point based on the charging demand forecast data and the preliminary site selection plan data, so as to obtain the charging pile configuration data of each site selection point; generate capacity planning data based on the charging pile configuration data and the existing energy storage conditions; and evaluate the impact of the new load on the power grid based on the capacity planning data according to the power infrastructure data, so as to obtain the power grid impact assessment data; The planning scheme optimization module is used to generate grid constraint data based on the grid impact assessment data, and to make balance corrections to the preliminary site selection scheme data and charging pile configuration data based on the charging demand and grid stability according to the grid constraint data, so as to obtain the planning scheme data for the charging pile construction.

[0008] The present invention can comprehensively utilize information from various data sources to support charging pile construction planning by acquiring existing power infrastructure data and multi-source heterogeneous data (including traffic data, geographic data, vehicle data, and charging behavior data). By extracting time series features from multi-source heterogeneous data, the temporal variation pattern of the data can be captured, providing valuable feature information for subsequent demand forecasting and site selection decisions. By using a preset hierarchical attention network model to fuse the features of different data sources across time scales, the information of multi-source data can be effectively integrated to improve the data characterization capability. By fusing multi-source data and extracting factor weight data, the contribution of different data sources to charging demand can be quantified, providing an important reference for subsequent demand forecasting and site selection decisions. Based on factor weight data and multi-source heterogeneous data, a sequence-to-sequence model is constructed for long-term charging demand forecasting; this can help planners understand the trend and change of future charging demand and provide long-term planning and decision-making basis for charging pile construction. By performing uncertainty interval evaluation, the confidence interval of charging demand forecast data can be provided to help planners understand the reliability and uncertainty of the forecast results. Combined with charging demand forecast data and power infrastructure data, the method based on the risk minimum set covering problem is used to select the location of charging piles, which can ensure that the layout of charging piles fully covers the charging demand and minimizes the risk of power grid, such as reducing the potential risk of power system overload. Optimization based on coverage, service radius and accessibility can make the site selection plan more reasonable and effective, and improve the availability and convenience of charging services. By predicting the load curve of the charging demand forecast data and the preliminary site selection plan data, the load situation at the site selection point can be evaluated to provide a reference for the configuration of charging piles; generating capacity planning data, combined with existing energy storage conditions, can ensure that the energy storage capacity in the charging pile construction plan meets the demand and optimize energy utilization efficiency. According to the power infrastructure data, the impact of the new load on the power grid is evaluated on the capacity planning data, which can evaluate the impact of the construction of charging piles on the power grid, discover potential power grid problems in advance, and take corresponding measures to adjust and optimize. Generate power grid constraint data based on the power grid impact assessment data to ensure that the charging pile construction plan meets the reliability, stability and safety requirements of the power grid. Based on the balance correction of charging demand and power grid stability, the preliminary site selection plan data and charging pile configuration data are optimized to obtain the planning plan data for charging pile construction that meets the power grid constraints. By optimizing the planning scheme, we can comprehensively consider charging demand, grid constraints and other factors, improve the rationality and feasibility of charging pile construction planning, and achieve the optimal layout of charging services. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings: Figure 1This is a schematic diagram of the steps of the charging pile construction planning method based on artificial intelligence of the present invention; Figure 2 for Figure 1 Detailed step flow diagram of step S1; Figure 3 for Figure 1 Detailed step flow chart of step S2 in FIG. DETAILED DESCRIPTION

[0010] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0011] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0012] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0013] To achieve this, please refer to Figures 1 to 3 The present invention provides a charging pile construction planning method based on artificial intelligence, the method comprising the following steps: Step S1: obtaining existing power infrastructure data and multi-source heterogeneous data, wherein the multi-source heterogeneous data includes traffic data, geographic data, vehicle data, and charging behavior data; extracting time series features from the multi-source heterogeneous data, and using a preset hierarchical attention network model to fuse features from different data sources across time scales, thereby obtaining factor weight data; Step S2: constructing a long-term charging demand prediction model based on factor weight data and multi-source heterogeneous data through a sequence-to-sequence model; performing long-term prediction of future charging demand based on the long-term charging demand prediction model, and performing uncertainty interval evaluation to obtain charging demand prediction data; Step S3: Perform charging pile site selection based on the risk minimum set coverage problem according to the charging demand forecast data and the power infrastructure data, and perform target optimization based on coverage, service radius and accessibility to obtain preliminary site selection plan data; Step S4: predicting the load curve of the site selection point according to the charging demand forecast data and the preliminary site selection plan data, thereby obtaining the charging pile configuration data of each site selection point; generating capacity planning data according to the charging pile configuration data and the existing energy storage conditions; and evaluating the impact of the new load on the power grid according to the power infrastructure data on the capacity planning data, thereby obtaining the power grid impact assessment data; Step S5: Generate grid constraint data based on the grid impact assessment data, and make balance corrections to the preliminary site selection plan data and charging pile configuration data based on charging demand and grid stability according to the grid constraint data, so as to obtain planning plan data for charging pile construction.

[0014] In the embodiment of the present invention, reference Figure 1 The above is a schematic diagram of the steps of a charging pile construction planning method based on artificial intelligence of the present invention. In this example, the charging pile construction planning method based on artificial intelligence includes the following steps: Step S1: obtaining existing power infrastructure data and multi-source heterogeneous data, wherein the multi-source heterogeneous data includes traffic data, geographic data, vehicle data, and charging behavior data; extracting time series features from the multi-source heterogeneous data, and using a preset hierarchical attention network model to fuse features from different data sources across time scales, thereby obtaining factor weight data; The embodiments of the present invention collect existing power infrastructure data and multi-source heterogeneous data, including traffic data, geographic data, vehicle data and charging behavior data; perform time series feature extraction on the multi-source heterogeneous data to extract features related to charging demand, such as time, location, traffic flow, charging behavior, etc.; use a preset hierarchical attention network model to fuse features of different data sources across time scales to obtain factor weight data.

[0015] Step S2: constructing a long-term charging demand prediction model based on factor weight data and multi-source heterogeneous data through a sequence-to-sequence model; performing long-term prediction of future charging demand based on the long-term charging demand prediction model, and performing uncertainty interval evaluation to obtain charging demand prediction data; The embodiment of the present invention uses a sequence-to-sequence model, combines factor weight data and multi-source heterogeneous data, and constructs a long-term charging demand prediction model; performs long-term prediction of future charging demand based on the model, and evaluates the uncertainty interval of the prediction result to provide charging demand prediction data.

[0016] Step S3: Perform charging pile site selection based on the risk minimum set coverage problem according to the charging demand forecast data and the power infrastructure data, and perform target optimization based on coverage, service radius and accessibility to obtain preliminary site selection plan data; The embodiment of the present invention combines charging demand forecast data and power infrastructure data, and applies a method based on the risk minimum set covering problem to select charging pile sites; optimization is performed based on objectives such as coverage, service radius and accessibility to obtain preliminary site selection plan data.

[0017] Step S4: predicting the load curve of the site selection point according to the charging demand forecast data and the preliminary site selection plan data, thereby obtaining the charging pile configuration data of each site selection point; generating capacity planning data according to the charging pile configuration data and the existing energy storage conditions; and evaluating the impact of the new load on the power grid according to the power infrastructure data on the capacity planning data, thereby obtaining the power grid impact assessment data; The embodiment of the present invention uses charging demand forecast data and preliminary site selection plan data to predict the charging load curve of each site selection point to determine the configuration data of the charging pile; generates capacity planning data based on existing energy storage conditions to ensure that the power supply capacity of the charging pile meets the demand; evaluates the impact of the new load on the power grid based on the power infrastructure data, and generates power grid impact assessment data.

[0018] Step S5: Generate grid constraint data based on the grid impact assessment data, and make balance corrections to the preliminary site selection plan data and charging pile configuration data based on charging demand and grid stability according to the grid constraint data, so as to obtain planning plan data for charging pile construction.

[0019] The embodiment of the present invention generates grid constraint data based on grid impact assessment data, including constraints in terms of grid stability, load balancing, etc.; performs balance correction on preliminary site selection plan data and charging pile configuration data based on the grid constraint data to achieve a balance between charging demand and grid stability; and obtains planning plan data for charging pile construction, which meets charging demand and takes into account the impact and constraints of the grid.

[0020] By acquiring and fusing multi-source heterogeneous data, the present invention can comprehensively utilize information such as traffic data, geographic data, vehicle data, and charging behavior data to provide a more comprehensive and multi-dimensional data foundation. Time series feature extraction can capture the trend and pattern of data changes over time, revealing the characteristics of data such as periodicity, trend and correlation. Through the preset hierarchical attention network model, the features of different time scales can be integrated to better understand the long-term and short-term impacts of the data and provide more accurate factor weight data. Through the sequence-to-sequence model, a charging demand prediction model can be established based on factor weight data and multi-source heterogeneous data to provide an accurate prediction of future charging demand. By evaluating the uncertainty interval, a credibility estimate of the charging demand prediction results can be provided to help decision makers understand the scope and possible deviations of the prediction. Combining charging demand prediction data and power infrastructure data, the risk minimum set coverage problem can be comprehensively considered on the premise of meeting charging demand to ensure the accuracy and feasibility of the site selection plan. Based on the optimization of coverage, service radius and accessibility, different goals can be weighed in the site selection process to improve the service effect and coverage of charging piles. Through the prediction of the load curve of the site selection point, the charging demand and load characteristics of each site selection point can be estimated, providing a basis for the configuration and capacity planning of charging piles. Based on the configuration data of the charging piles and the existing energy storage conditions, capacity planning can be carried out to evaluate the impact of the new load on the power grid. This helps to understand the impact of charging pile construction on the stability and reliability of the power grid and provide a reasonable power grid transformation plan. Based on the power grid impact assessment data, power grid constraint data can be generated to comprehensively consider the charging demand and power grid stability to ensure the coordination and balance between the construction of charging piles and the power grid. Based on the balance between charging demand and power grid stability, the preliminary site selection plan and charging pile configuration are revised to meet the power grid constraints and provide reliable charging services. In summary, the effects of the above steps include data synthesis, feature extraction, prediction accuracy, uncertainty assessment, comprehensive consideration, target optimization, load curve prediction, capacity planning and impact assessment, power grid constraint consideration, and balance correction. These effects help to improve the accuracy and reliability of charging demand prediction, optimize the site selection and configuration plan of charging piles, and ensure the feasibility of charging services and the stability of the power grid.

[0021] Preferably, step S1 comprises the following steps: Step S11: Acquire existing power infrastructure data, including substation data, distribution network data, and line capacity information; Step S12: Collect road network, traffic flow and congestion information to obtain traffic data; obtain geographic data, including terrain information data, land use data and urban planning data; obtain vehicle data, including new energy vehicle distribution data, mileage data and charging frequency data; obtain charging behavior data, including charging time data, charging amount data and charging location data; Step S13: preprocessing the traffic data, geographic data, vehicle data, and charging behavior data, and performing time alignment to obtain multi-source heterogeneous data; Step S14: extract time series features from multi-source heterogeneous data, and use a preset hierarchical attention network model to fuse features from different data sources across time scales to obtain factor weight data.

[0022] As an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 Detailed step flow diagram of step S1 in the embodiment of the present invention, step S1 includes the following steps: Step S11: Acquire existing power infrastructure data, including substation data, distribution network data, and line capacity information; The embodiment of the present invention obtains the location, capacity, operation status and other information of the substation by contacting the power supply company or relevant institutions, and can use data query tools or exchange data with relevant institutions; obtains the topology structure, line connection relationship and capacity information of the distribution network. These data can come from the database of the power supply company or relevant institutions, or be obtained through on-site investigation and measurement; by contacting the power supply company or relevant institutions, the capacity information of the power line is obtained, which is usually recorded in the data of the substation and the distribution network.

[0023] Step S12: Collect road network, traffic flow and congestion information to obtain traffic data; obtain geographic data, including terrain information data, land use data and urban planning data; obtain vehicle data, including new energy vehicle distribution data, mileage data and charging frequency data; obtain charging behavior data, including charging time data, charging amount data and charging location data; The embodiments of the present invention may use traffic flow monitoring data, traffic camera data, GPS data, etc. provided by traffic management departments or relevant agencies. These data can be used to understand the topological structure, traffic flow, and congestion of the road network. Terrain information, land use data, and urban planning data can be obtained through a geographic information system (GIS) or a related geographic data service provider, which can provide information about the geographical environment, such as terrain height, land use, and urban planning layout. New energy vehicle distribution data, mileage data, and charging frequency data can be obtained through vehicle registration and management agencies, traffic management departments, or related agencies, which can help understand the number, usage, and charging needs of new energy vehicles. Charging time data, charging amount data, and charging location data can be obtained through charging pile operators, charging service providers, or related agencies, which can provide detailed information about charging behavior, such as charging time, charging amount, and the location of charging facilities.

[0024] Step S13: preprocessing the traffic data, geographic data, vehicle data, and charging behavior data, and performing time alignment to obtain multi-source heterogeneous data; The embodiment of the present invention cleans, denoises and formats the collected data to ensure the quality and consistency of the data, which may include processing missing values, outliers and duplicate values, as well as data type conversion and data standardization. Since the data sources may have different timestamps or time resolutions, the data needs to be time-aligned so that they can correspond to the same time point or time period in subsequent analysis. Time alignment operations can be performed using methods such as time interpolation, timestamp alignment or time period division.

[0025] Step S14: extract time series features from multi-source heterogeneous data, and use a preset hierarchical attention network model to fuse features from different data sources across time scales to obtain factor weight data.

[0026] The embodiments of the present invention extract time-related features from multi-source heterogeneous data. This may include basic statistical features (such as mean, variance, maximum, minimum), time series features (such as trend, periodicity, seasonality) and other domain-specific features. The features of different data sources are fused across time scales using a preset hierarchical attention network model. This model can perform weighted fusion at different time scales based on the importance and relevance of the features to obtain comprehensive factor weight data. The model can be implemented and trained using a deep learning framework (such as TensorFlow, PyTorch).

[0027] By acquiring power infrastructure data, the present invention can understand the location and capacity of substations, the topological structure of distribution networks, and the capacity information of lines, and provide basic data for subsequent charging pile site selection and planning; line capacity information can be used to evaluate the load carrying capacity of the power grid and determine whether power grid transformation is needed to support the construction of charging piles. Road network data can be used to analyze traffic flow, congestion, and vehicle driving paths, and help determine the site selection and layout strategy of charging piles; terrain information, land use data, and urban planning data in geographic data can assist in site selection analysis, considering the impact of geographical conditions and urban planning on the layout of charging piles; information such as the distribution of new energy vehicles, mileage, and charging frequency in vehicle data can provide the spatial distribution characteristics of vehicle demand, and provide a basis for charging pile site selection and capacity planning; information such as charging time, charging amount, and charging location in charging behavior data can reveal the user's charging mode and demand characteristics, and help optimize the layout and service strategy of charging piles. Through data preprocessing and time alignment, multi-source heterogeneous data can have a consistent time scale, which is convenient for subsequent feature extraction and fusion; data preprocessing can clean and process raw data, eliminate noise and outliers, and improve data quality and availability. Time series feature extraction can extract time-related features from multi-source heterogeneous data, revealing the periodicity, trend and correlation of the data, and providing a feature basis for subsequent charging demand forecasting and site selection analysis; through the preset hierarchical attention network model, features of different time scales can be fused, and long-term and short-term impacts can be comprehensively considered to obtain more accurate factor weight data, which helps to improve the reliability of charging demand forecasting and site selection plans.

[0028] Preferably, step S14 comprises the following steps: Step S141: divide the multi-source heterogeneous data into time windows, extract time series statistical features, and extract time-frequency domain features through wavelet transformation to obtain time series feature data; The embodiment of the present invention divides multi-source heterogeneous data into continuous time windows, each window contains a continuous time series data, and the size of the window can be selected according to the needs and the time resolution of the data. Statistical features are extracted from the data in each time window, including mean, variance, maximum value, minimum value, trend, etc. These statistical features can reflect the overall trend and change of the data. The time series data is subjected to wavelet transform to extract time-frequency domain features. Wavelet transform can reveal the frequency components and time domain changes of the data. Commonly used wavelet transform methods include discrete wavelet transform (DWT) and continuous wavelet transform (CWT).

[0029] Step S142: performing rasterization processing on multi-source heterogeneous data based on geographic space, and extracting spatial clustering features, thereby constructing a spatial adjacency matrix; The embodiment of the present invention converts geographic data into raster data, divides geographic space into regular raster units, and can use geographic information system (GIS) tools and algorithms to convert geographic data into raster data. Perform spatial clustering analysis on the rasterized data to extract spatial clustering features. Commonly used spatial clustering algorithms include K-means clustering, DBSCAN clustering, etc. These algorithms can divide raster units into spatial clusters with similar features.

[0030] Step S143: performing feature dimensionality reduction based on principal component analysis according to the spatial adjacency matrix and the time series feature data, and performing recursive feature elimination to obtain key feature data; The embodiment of the present invention performs principal component analysis on the time series feature data and the spatial clustering feature data, extracts key features and performs dimensionality reduction. PCA can reduce high-dimensional data to lower dimensions while retaining the most important feature information. Through the recursive feature elimination algorithm, the features are gradually eliminated to screen out the key features that have an important impact on problem solving. The algorithm can perform iterative selection based on the importance of the features and the performance of the model.

[0031] Step S144: designing a multi-level attention mechanism according to the key feature data, constructing a hierarchical network structure adapted to features of different scales, thereby obtaining a hierarchical attention network model; The embodiment of the present invention designs a network structure with a multi-level attention mechanism for adaptive weighting of features of different scales, which may include a convolutional neural network (CNN) or a recurrent neural network (RNN) based on the attention mechanism, etc. The attention mechanism can learn the importance weights of different features to achieve adaptive fusion of features. A hierarchical network structure is constructed to adapt to the extraction and fusion of features of different scales, which may include multiple layers of convolution or recurrent layers, as well as pooling or upsampling layers, etc. The hierarchical network structure can capture feature representations of different scales.

[0032] Step S145: Adaptively weighting the key feature data using a hierarchical attention network model, and fusing features across time scales to obtain multi-scale feature data; The embodiment of the present invention can use the hierarchical attention network model to adaptively weight the key feature data. Through the attention mechanism, the key feature data can be weighted according to the importance of the feature to highlight the features that contribute more to the model prediction. In addition, cross-time scale feature fusion can be performed to integrate features of different time scales to obtain multi-scale feature data, so that the feature information at different time scales can be comprehensively considered to improve the model's ability to represent data.

[0033] Step S146: Evaluate the impact weight of the charging demand according to the multi-scale feature data to obtain factor weight data.

[0034] The embodiment of the present invention uses multi-scale feature data to evaluate the impact weight of charging demand. By analyzing the correlation between the multi-scale feature data and the charging demand, the degree of influence of different features on the charging demand can be determined. In this way, factor weight data can be obtained for subsequent charging demand prediction or related decision-making.

[0035] The present invention can capture the periodicity, trend and correlation of data by dividing multi-source heterogeneous data into time windows and extracting time series statistical features, and provide feature information in the time dimension; through wavelet transform and other technologies, data can be converted from time domain to frequency domain, and feature information in the frequency domain can be extracted, providing a more comprehensive feature basis for subsequent analysis. By rasterizing multi-source heterogeneous data, data can be matched with geographic space, and feature information related to space, such as geographic location, distance, etc., can be extracted; by clustering the rasterized data, the aggregation pattern of data in geographic space can be found, spatial adjacency relationships can be extracted, and a spatial adjacency matrix can be constructed to provide feature information in the spatial dimension for subsequent analysis. Through methods such as principal component analysis, high-dimensional spatial adjacency matrices and time series feature data can be reduced to key feature data of lower dimensions, reducing data redundancy and dimensional complexity; through recursive feature elimination algorithms, features that have little impact on prediction results can be gradually eliminated, improving the effect and generalization ability of the model. By designing a multi-level attention mechanism, key feature data can be adaptively weighted to increase the attention to different features, so as to better capture feature information of different scales; by constructing a hierarchical network structure, feature information of different levels can be fully utilized to improve the model's expressiveness and prediction accuracy. Through the hierarchical attention network model, key feature data can be adaptively weighted, and weighted fusion can be performed according to the importance and influence of different features to improve the expressiveness and prediction performance of features; through cross-time scale feature fusion, feature information on different time scales can be integrated, so that the model can consider both short-term and long-term effects and improve the modeling ability of time changes. By analyzing and evaluating multi-scale feature data, the weight of each feature on charging demand can be determined, and the importance and contribution of different factors to charging demand can be understood; the obtained factor weight data can be used as a basis for decision support, help determine the key factors of charging demand, and guide the planning and resource allocation of charging facilities. In summary, the effects of the above steps include time series feature extraction, time-frequency domain feature extraction, spatial feature extraction, spatial clustering, feature dimension reduction, recursive feature elimination, attention mechanism, hierarchical network structure, adaptive weighting, feature fusion, influence weight evaluation and factor weight data, etc. These effects help to improve the accuracy, robustness and interpretability of data analysis and prediction models, thereby providing useful support for the research and decision-making of charging demand.

[0036] Preferably, step S2 comprises the following steps: Step S21: constructing a long-term charging demand prediction model based on factor weight data and multi-source heterogeneous data using a long short-term memory network sequence model, wherein the factor weight data is used as an additional input of the sequence model and the multi-source heterogeneous data is used as the main input of the sequence model when constructing the long-term charging demand prediction model; Step S22: performing a long-term prediction of future charging demand through a long-term charging demand prediction model, thereby obtaining charging demand prediction result data; Step S23: Generate multiple groups of prediction results according to the charging demand prediction result data by using the Monte Carlo simulation method, and estimate the upper and lower bounds of the prediction by using the quantile regression method, so as to obtain uncertainty interval data; Step S24: using the SHAP value to interpret the prediction results of the uncertainty interval data, thereby obtaining charging demand prediction data.

[0037] As an embodiment of the present invention, refer to Figure 3 As shown, Figure 1 Detailed step flow diagram of step S2 in the embodiment of the present invention, step S2 includes the following steps: Step S21: constructing a long-term charging demand prediction model based on factor weight data and multi-source heterogeneous data using a long short-term memory network sequence model, wherein the factor weight data is used as an additional input of the sequence model and the multi-source heterogeneous data is used as the main input of the sequence model when constructing the long-term charging demand prediction model; The embodiment of the present invention uses the Python programming language and a deep learning framework, such as TensorFlow, Keras, or PyTorch, to build a long short-term memory (LSTM) network model.

[0038] Data preparation: Prepare factor weight data and multi-source heterogeneous data as input to the model, and ensure that the data has undergone steps such as preprocessing, normalization, and feature engineering to meet the requirements of the LSTM model. Use the selected deep learning framework to build the LSTM model, taking the factor weight data as additional input and the multi-source heterogeneous data as the main input. The factor weight data can be processed by connecting this data as the input layer of the model or using an attention mechanism. Take the labeled charging demand data as the target variable, train the LSTM model with the prepared data, and adjust the weights and biases of the model through the backpropagation algorithm and optimizer (such as Adam or SGD) to enable the model to learn to predict long-term charging demand. Use a validation set or cross-validation method to evaluate the performance of the trained model. Various indicators such as root mean square error (RMSE) or mean absolute percentage error (MAPE) can be used to measure the accuracy and performance of the model.

[0039] Step S22: performing a long-term prediction of future charging demand through a long-term charging demand prediction model, thereby obtaining charging demand prediction result data; The embodiment of the present invention uses a trained long-term charging demand prediction model and provides future input data to the model to predict long-term charging demand; according to the requirements of the prediction model, the future factor weight data and multi-source heterogeneous data are sorted into a suitable input format, and data preparation steps such as preprocessing and normalization are performed. The prepared data is input into the prediction model, and the charging demand prediction result data generated by the model is obtained. These prediction results can cover a longer time range, such as a few days, weeks or months.

[0040] Step S23: Generate multiple groups of prediction results according to the charging demand prediction result data by using the Monte Carlo simulation method, and estimate the upper and lower bounds of the prediction by using the quantile regression method, so as to obtain uncertainty interval data; The embodiment of the present invention uses a Monte Carlo simulation method to generate multiple groups of prediction results based on the charging demand prediction result data. The Monte Carlo simulation can simulate uncertainty by randomly sampling from the probability distribution of the prediction results. For each Monte Carlo sampling, the required statistical indicators, such as the mean, median, etc., are calculated based on the generated prediction results. Based on the results of the Monte Carlo sampling, the quantile regression method is used to estimate the upper and lower bounds of the prediction results. Quantile regression can be used to estimate the upper and lower bounds of the prediction distribution at a given confidence level. Uncertainty interval data can be obtained by calculating different quantiles (e.g., upper and lower bounds).

[0041] Step S24: using the SHAP value to interpret the prediction results of the uncertainty interval data, thereby obtaining charging demand prediction data.

[0042] The embodiment of the present invention uses the SHAP (SHapley Additive exPlanations) value method to interpret the charging demand forecast result data. The SHAP value is a method for interpreting machine learning model predictions, which can quantify the contribution of each input feature to the forecast result. For each charging demand forecast result data point, the SHAP value of each input feature is calculated within the uncertainty interval, which will help determine which factors contribute most to the change in the forecast result. Analyze and interpret the SHAP value to gain insights into the charging demand forecast data. The interpretation results can be displayed by drawing visualization methods such as important feature graphs, summary graphs, or force-directed graphs of the SHAP value.

[0043] The present invention can capture the temporal dependency in the charging demand data and improve the modeling capability of long-term charging demand by using the LSTM network sequence model; taking the factor weight data as an additional input can better consider the importance and contribution of each factor to the charging demand, and improve the accuracy and interpretability of the prediction model; taking multi-source heterogeneous data as the main input can enrich the feature space of the charging demand prediction model, provide more comprehensive information, and improve the prediction performance. By using the pre-built long-term charging demand prediction model, the future charging demand can be predicted in the long term, and the overall trend and trend change of the charging demand in the future time period can be estimated; the obtained charging demand prediction result data can be used as a basis for decision support to help plan the capacity, resource allocation and operation strategy of charging facilities. Through the Monte Carlo simulation method, the uncertainty of the charging demand prediction results can be considered, and multiple groups of prediction results can be generated to reflect the prediction situation under different possibilities; using the quantile regression method, multiple groups of prediction results can be statistically analyzed, the upper and lower bounds of the prediction results can be estimated, and the confidence interval and uncertainty measurement of the charging demand prediction can be provided. By using the SHAP (Shapley Additive Explanations) value method, the charging demand forecast results can be explained, the contribution and influence of different factors on the forecast results can be understood, and an explanation and understanding of the forecast results can be provided; by analyzing the SHAP value, factors that have a positive or negative impact on the forecast results can be identified, helping decision makers understand the main driving factors of the charging demand forecast and providing guidance in the decision-making process.

[0044] Preferably, step S3 comprises the following steps: Step S31: Mapping charging demand forecast data and power infrastructure data to a geographic grid, thereby obtaining geographical demand distribution data and infrastructure distribution data; The embodiments of the present invention use geographic information system (GIS) software, such as ArcGIS, QGIS, or libraries such as GeoPandas and Folium in Python, to process geographic data and perform geographic information analysis. Prepare charging demand forecast data and power infrastructure data, and ensure that the data contains information related to the geographic location, such as longitude and latitude or address. Map the charging demand forecast data and power infrastructure data to a geographic grid, which can be a raster (such as a square or hexagonal grid) or a vector polygon boundary, ensuring that the data is associated with the grid cells or polygon boundaries to obtain geographical demand distribution data and infrastructure distribution data.

[0045] Step S32: performing hot spot analysis on the demand distribution data to determine high-demand areas, and screening available locations with land use restrictions based on the infrastructure distribution data to obtain a set of candidate charging pile locations; Embodiments of the present invention use hot spot analysis methods, such as Kernel Density Estimation or Getis-Ord Gi* statistics, to analyze demand distribution data to identify high-demand hot spots, which represent potential areas of concentrated charging demand. Use infrastructure distribution data and land use planning information to limit locations that can be used for charging pile installations. For example, land that has been occupied for other purposes, such as residential areas, nature reserves, or restricted areas, can be excluded. Combined with high-demand hot spots and land use restrictions, a set of qualified candidate charging pile locations is screened out, which should have high demand and suitable land use conditions.

[0046] Step S33: performing a location risk assessment based on natural disaster risks on the candidate charging pile location set, thereby obtaining location risk score data; The embodiment of the present invention prepares data related to natural disasters, such as historical records or probability data of disasters such as floods, earthquakes, storm surges, etc. These data can come from meteorological departments, geological survey institutions or related research institutions. Using risk assessment methods such as probability analysis, vulnerability assessment or multi-index assessment, a natural disaster risk assessment is performed on the candidate charging pile location set, and a risk score is calculated for each location based on the natural disaster data and location characteristics. A higher score indicates a location with a higher risk.

[0047] Step S34: Establishing a maximum coverage location problem model for the candidate charging pile location set according to the preset service radius and coverage rate index, thereby obtaining a location coverage matrix; The embodiment of the present invention determines the preset service radius of the charging pile, that is, the service range of the charging pile station, and at the same time, defines the coverage index, such as the demand volume covered within a specific range or the proportion of the population covered; according to the candidate charging pile location set, the service radius and the coverage index, a maximum coverage location problem model is established, the goal of which is to find the best combination of charging pile locations to achieve the maximum coverage demand or the maximum coverage population. Using a method for solving the maximum coverage location problem model (such as linear programming, integer programming or heuristic algorithm), a location coverage matrix is ​​calculated and generated, which shows the coverage range corresponding to each candidate charging pile location, as well as the demand volume or population covered.

[0048] Step S35: Calculate the road network distance and driving time according to the candidate charging pile location set and the road network map information, so as to obtain location accessibility data; The embodiment of the present invention obtains road network map data, including information such as the topological structure of the road, the road length and the connection relationship. This data can come from geographic information system (GIS) software, open map data or APIs of third-party map providers. Using the road network map data and the set of candidate charging pile locations, the road network distance and driving time between each candidate location are calculated. The shortest path algorithm (such as the Dijkstra algorithm or the A* algorithm) can be used to calculate the shortest path or the fastest path. Generate location accessibility data, which contains the road network distance and driving time between each candidate charging pile location and other locations. This data can be used to evaluate the accessibility and traffic convenience of the charging pile.

[0049] Step S36: establishing a multi-objective integer programming model according to the location risk score data, the location coverage matrix, and the location accessibility data, thereby obtaining a multi-objective optimization model; The embodiment of the present invention establishes a multi-objective integer programming model based on location risk score data, location coverage matrix and location accessibility data, the goal of which is to achieve a balance between minimizing location risk, maximizing location coverage and maximizing location accessibility. Define decision variables, such as charging pile site selection decision variables, and constraints, such as the number of charging piles in each area or the threshold constraint of the location risk score, and use a multi-objective optimization algorithm (such as NSGA-II, MOEA / D or SPEA2) to solve the multi-objective integer programming model to obtain a set of optimal or approximately optimal charging pile site selection solutions.

[0050] Step S37: Optimize the charging pile site selection for the candidate charging pile location set using the NSGA-II algorithm according to the multi-objective optimization model, thereby obtaining preliminary site selection plan data.

[0051] First, the embodiment of the present invention needs to define a multi-objective optimization model, which includes the objective function and constraints to be optimized. The objective function may include maximizing coverage, minimizing location risk score, etc., and the constraints may include service radius restrictions, land use restrictions, etc. Multiple objectives and constraints are determined based on actual needs and trade-offs.

[0052] NSGA-II algorithm: NSGA-II (non-dominated sorting genetic algorithm II) is a commonly used multi-objective optimization algorithm suitable for solving optimization problems with multiple objective functions. The algorithm searches and maintains a set of optimal solutions on the Pareto front by simulating the process of biological inheritance and evolution. According to the set of candidate charging pile locations, an initial population is initialized. Each individual in the population represents a charging pile location scheme. Binary coding, integer coding or floating point coding can be used to represent the location or selection state of the charging pile. For each individual in the population, its fitness value is calculated, that is, the target value of the individual is calculated according to the objective function of the multi-objective optimization model. The fitness value reflects the performance of the individual in the multi-objective optimization problem. The individuals in the population are non-dominated sorted and divided into different levels, where individuals with higher levels have better performance. Then, the crowding distance of the individuals is calculated to maintain the diversity of the population. According to the non-dominated sorting and crowding distance, a new generation of individuals is selected using a selection operation. Individuals with higher levels and individuals with greater crowding have a higher probability of being selected. Perform a crossover operation on the selected individuals to generate new individuals. The crossover operation can use single-point crossover, multi-point crossover or uniform crossover methods to produce offspring individuals with genetic characteristics. Apply mutation operations to the individuals after crossover to introduce random perturbations and increase the exploration ability of the search space; mutation operations can make small random changes to the position or state of individuals. Merge the individuals after crossover and mutation with the original population to form a new generation of population; ensure that the population size remains unchanged to maintain the diversity of the population until the predetermined number of iterations or convergence conditions are reached. Each iteration will gradually improve the charging pile site selection plan and approach the optimal solution on the Pareto front. Extract the Pareto front from the final population to obtain a set of optimal or near-optimal charging pile site selection plans, which represent the best solutions under different trade-offs.

[0053] The present invention can associate demand and infrastructure information with geographic space by mapping charging demand forecast data and power infrastructure data to geographic grids, so as to provide a more specific and accurate charging demand distribution map and infrastructure distribution map; these geographical data can help us better understand the spatial distribution of charging demand and the distribution of existing infrastructure. By performing hot spot analysis, high-demand areas, that is, areas where charging demand is more concentrated, can be determined, which can help us identify places where charging piles are most needed so as to provide better charging services in these areas. At the same time, based on the infrastructure distribution data, the available sites with land use restrictions can be screened, and some sites that are not suitable for the construction of charging piles, such as restricted land or areas that conflict with other facilities, can be excluded, which helps to narrow the scope of candidate charging pile locations and make the final site selection more feasible and reasonable. By conducting a location risk assessment based on natural disaster risks, the level of natural disaster risk faced by the candidate charging pile locations can be evaluated; this helps to avoid choosing to build charging piles in areas susceptible to natural disasters, thereby improving the reliability and stability of charging piles. By establishing a maximum coverage location problem model, it is possible to determine which locations in the candidate charging pile location set can best cover the charging needs of the surrounding area, which helps determine the best charging pile location to ensure the convenience and accessibility of charging services. By calculating the road network distance and travel time between the candidate charging pile location set and the road network map, the location accessibility of the charging pile can be evaluated, which helps determine whether the location of the charging pile can conveniently provide charging services to users, especially when the road network topology is complex or the traffic conditions are different. By establishing a multi-objective integer programming model, multiple factors such as location risk score, location coverage matrix and location accessibility can be comprehensively considered, which helps to find the best charging pile site selection plan to maximize the charging demand, reduce location risk and improve the accessibility of charging services. Through multi-objective optimization, an optimal solution or a set of optimal solutions can be found under different trade-offs and constraints. By using multi-objective optimization algorithms such as NSGA-II (non-dominated sorting genetic algorithm), the candidate charging pile location set can be optimized and selected to find a set of optimal charging pile site selection solutions, so that a balance can be found between different objectives, such as maximizing coverage, minimizing risk scores, etc. The preliminary site selection solution data obtained can be used as a basis for further evaluation and decision-making, and provide valuable reference for the final charging pile site selection. In general, the effects of the above steps include providing geographical demand and infrastructure distribution data, determining high-demand areas, excluding locations that are not suitable for charging pile construction, assessing location risks, determining maximum coverage locations, assessing location accessibility, multi-objective optimization site selection, and finally obtaining preliminary site selection solution data. The combined application of these steps can help decision makers conduct a comprehensive charging pile site selection analysis to meet charging needs and consider factors such as risk, accessibility, and coverage.

[0054] Preferably, step S37 includes the following steps: Step S371: Solve the multi-objective optimization model using the NSGA-II algorithm, and calculate the comprehensive score of each solution, thereby obtaining a screened optimal location solution candidate set; In step S37, the embodiment of the present invention has defined a multi-objective optimization model, including the objective function and constraints to be optimized. According to the actual situation, multiple objectives and constraints are determined, such as maximizing coverage, minimizing location risk score, etc. The multi-objective optimization model is solved using the NSGA-II algorithm. The NSGA-II algorithm can be implemented using a programming language (such as Python) and a corresponding optimization library or framework. The specific implementation method can refer to the NSGA-II algorithm example code provided above. For each solution (i.e., each candidate site selection scheme) obtained by the NSGA-II algorithm, a comprehensive score is calculated according to the objective function in the multi-objective optimization model; the comprehensive score can be calculated by weighted summation or other methods according to the importance and weight of the objective function. The higher the score, the better the comprehensive performance of the scheme. All candidate site selection schemes are sorted according to the comprehensive score, and the scheme with a higher score is selected as the optimal site selection scheme candidate set; the number of candidate schemes to be retained can be determined or the score threshold can be set according to the needs.

[0055] Step S372: Performing site condition evaluation based on the optimal site solution candidate set and geographic data to obtain site adaptability evaluation data; The embodiment of the present invention collects geographic data related to site selection, such as land use data, population distribution data, road network data, etc. These data can be obtained from relevant departments, map service providers or open data platforms. According to the site selection requirements and constraints, determine suitable site condition evaluation indicators, for example, indicators such as land type, planning restrictions, traffic convenience, and surrounding facilities can be considered. Based on the optimal site selection candidate set and geographic data, calculate the site adaptability evaluation data of each candidate solution, and perform quantitative or qualitative evaluation on each candidate solution based on the selected evaluation indicators.

[0056] Step S373: sorting the site selection priorities of the optimal site selection candidate set according to the site adaptability evaluation data, thereby obtaining a site selection candidate set; The embodiment of the present invention determines the site selection priority indexes according to the site selection requirements and the site adaptability evaluation data. These indexes may be some evaluation indexes in the site evaluation data or additional indexes, such as return on investment, environmental impact, etc. According to the site selection priority indexes, the optimal site selection candidate set is sorted. The sorting may be performed by weighted summation, hierarchical analysis method, fuzzy evaluation, etc., to obtain the site selection candidate set.

[0057] Step S374: Visually display the coverage and service area of ​​the sorted candidate set of site selection schemes, thereby obtaining visualization data of the site selection schemes; The embodiment of the present invention obtains corresponding geographic data, including geographic boundary data, road network data, map layers, etc. These data can be obtained from relevant departments, map service providers or open data platforms; selects a suitable map visualization tool to display the coverage and service area of ​​the site selection plan. According to the location and coverage of the site selection plan, the map visualization tool is used to mark and draw areas on the map. The API or library provided by the tool can be used to add marks, draw polygonal or circular areas, etc. The visualization data of the site selection plan is displayed in combination with the geographic data, and interactive functions can be added as needed, such as clicking on the mark to display detailed information, switching layers to display different information, etc.

[0058] Step S375: performing parameter sensitivity evaluation on the multi-objective optimization model according to the optimal location solution candidate set, thereby obtaining parameter sensitivity analysis data; The embodiment of the present invention first determines the possible value range of each parameter in the multi-objective optimization model, which can be estimated based on domain knowledge, experience or historical data. According to the parameter range, a set of experimental schemes are designed to cover different combinations of parameter space, which can use design experimental methods such as full factorial design, Latin hypercube design or random sampling; for each experimental scheme, a multi-objective optimization model is used to solve and obtain the corresponding optimal location scheme; for each experimental scheme, the performance indicators of the optimal location scheme, such as comprehensive score, cost, efficiency, etc., are analyzed, and statistical analysis methods such as analysis of variance (ANOVA) or regression analysis can be used to evaluate the degree of influence of the parameters. By comparing the performance indicators of the optimal location scheme under different parameter values, the sensitivity of each parameter to the optimal location scheme is determined, and sensitivity indicators such as the main effect, interaction effect or total effect of the parameter can be used to evaluate the importance of the parameter.

[0059] Step S376: Screening the site selection scheme visualization data based on the parameter sensitivity analysis data using a multi-criteria decision-making method to obtain preliminary site selection scheme data.

[0060] The embodiment of the present invention first determines a set of appropriate evaluation criteria according to the specific needs of the site selection project. These criteria may include aspects such as economy, feasibility, environmental impact, and social benefits; for each evaluation criterion, its relative importance, i.e., weight, is determined, which can be determined by expert judgment, questionnaire survey, analytic hierarchy process (AHP), etc.; for each scheme in the preliminary site selection scheme data, the determined evaluation criteria and weight are used for evaluation, and quantitative or qualitative evaluation may be required according to different criteria. According to the evaluation results of each site selection scheme, the weight is used to calculate the comprehensive score, which can be calculated by weighted summation or multi-attribute decision-making methods (such as TOPSIS, ELECTRE), etc. According to the comprehensive score, the site selection scheme with the highest score is selected as the most promising site selection scheme.

[0061] The present invention can obtain a set of optimal site selection candidate sets by solving the multi-objective optimization model using the NSGA-II algorithm. The NSGA-II algorithm is a commonly used multi-objective optimization algorithm that can find a set of balanced solutions between multiple objectives, rather than just a single optimal solution. By calculating the comprehensive score of each solution, its performance on each objective can be evaluated, and the best performing candidate site selection scheme can be screened out. This helps to narrow the site selection range, reduce the complexity of decision-making, and provide a set of competitive site selection schemes for further evaluation and decision-making. By evaluating the site conditions of the optimal site selection candidate set and geographic data, the site conditions involved in each candidate site selection scheme can be evaluated, such as land ownership, topography, environmental protection requirements, etc. This helps to identify whether the candidate site selection scheme has limitations or challenges in site conditions, and provides site adaptability evaluation data as a basis for further decision-making. By sorting the optimal site selection candidate set according to the site adaptability evaluation data, the priority of each candidate site selection scheme can be determined. This helps to rank the candidate solutions according to their adaptability and feasibility, so that decision makers can better understand which solutions have more advantages and select the most promising site solutions in a targeted manner. By visualizing the coverage and service area of ​​the ranked candidate set of site solutions, the coverage and service area of ​​each site solution can be presented in an intuitive way. This helps decision makers better understand the spatial distribution and coverage of each solution, as well as the degree to which it meets the charging demand. Visualized data can help decision makers make comparisons and analyses, and provide a more intuitive reference for the final site selection. By evaluating the parameter sensitivity of the optimal site solution candidate set, the sensitivity of the multi-objective optimization model to different parameters can be evaluated. This helps to understand the impact of different parameters on the site selection solution and helps decision makers understand the importance and adjustment range of the parameters. Parameter sensitivity analysis data can provide insights into the model behavior, so that decision makers can better understand the reliability and stability of the model results. By using a multi-criteria decision-making method, combined with parameter sensitivity analysis data and site selection solution visualization data, candidate site solutions can be screened and evaluated. This method can consider multiple decision criteria, such as economics, sustainability, social impact, etc., and weigh these criteria together to find the best preliminary site selection plan. Through this step, decision makers can analyze and compare various plans more systematically, and obtain preliminary site selection plan data to provide a basis for further decision-making and analysis. Combining the effects of the above steps, it can be seen that this process provides a series of powerful tools and data support for site selection decisions. By using multi-objective optimization models, site adaptability assessment, parameter sensitivity assessment, and screening based on multi-criteria decision-making methods, decision makers can more comprehensively and objectively evaluate and compare the advantages and disadvantages of different site selection plans, and finally determine the appropriate site selection plan. The use of visual data also makes the decision-making process more intuitive and understandable.The effects of the entire process include improving the scientific nature of decision-making, reducing the subjectivity of decision-making, lowering site selection risks, and providing a reliable basis for decision-making.

[0062] Preferably, step S4 comprises the following steps: Step S41: Allocate charging demand forecast data to each site selection point in the preliminary site selection plan data, and collect traffic, geographic and population information around the site selection point to obtain site selection point feature data; The embodiment of the present invention distributes charging demand forecast data according to the location of the site selection point. Geographic Information System (GIS) tools can be used to spatially match the forecast data with the site selection point; for each site selection point, the surrounding traffic, geographic and population information is collected. This can include traffic flow data, road network data, terrain data, census data, etc. These data can be obtained through public data sources, local government agencies or professional data providers.

[0063] Step S42: mapping the site selection point to the charging demand according to the site selection point characteristic data, thereby generating a load curve prediction model for the site selection point; The embodiment of the present invention cleans and processes the feature data of the site selection point, including missing value processing, outlier processing, etc.; based on the feature data of the site selection point, relevant features are extracted for the load curve prediction model; for example, the population density, traffic flow, road network conditions, etc. around the site selection point can be considered as features. Based on the feature data of the site selection point and the corresponding charging demand, a load curve prediction model is established, and various machine learning methods such as linear regression, decision tree, support vector regression, etc. can be used. Selecting an appropriate model requires evaluation and comparison based on the specific situation.

[0064] Step S43: Determine the type and quantity of charging piles for each site selection point according to the load curve prediction model, thereby obtaining charging pile configuration data; The embodiment of the present invention determines the type of charging pile suitable for each site selection point based on the output of the load curve prediction model, combined with factors such as the power and charging time of the charging pile. For example, a fast charging pile or a slow charging pile can be selected based on the peak demand of the load curve. The number of charging piles required for each site selection point is calculated based on the output of the load curve prediction model and the needs of the site selection point; the reasonable number of charging piles is determined by considering factors such as the charging demand, charging efficiency, and charging time of the site selection point.

[0065] Step S44: Evaluate the role of energy storage demand at each site in peak load shaving and valley filling based on the load curve prediction model, charging pile configuration data, and existing energy storage conditions, so as to obtain capacity planning data; The embodiment of the present invention uses a load curve prediction model to predict the load curve of each site selection point, and evaluates the energy storage demand of each site selection point based on the load curve and charging pile configuration data. Considering factors such as the charging and discharging efficiency and energy storage capacity of the energy storage equipment, the energy storage demand is determined to achieve the goal of peak shaving and valley filling.

[0066] Step S45: Evaluate the impact of the newly added load on the power grid based on the capacity planning data according to the power infrastructure data, thereby obtaining power grid impact evaluation data.

[0067] The embodiment of the present invention obtains power infrastructure data, including information on power transmission lines, substations, distribution equipment, etc. of the power grid, which can be obtained from power companies, energy management departments or related institutions; integrates capacity planning data with power infrastructure data, considering the impact of new loads on the power grid; and evaluates the impact of new loads on the power grid based on capacity planning data and power infrastructure data. This can include evaluations of power grid stability, voltage stability, line overload, etc., and power system simulation tools can be used to simulate and analyze power grid impacts, such as PSS / E, OpenDSS, etc.

[0068] The present invention can associate the charging demand with the specific site selection point by allocating the charging demand forecast data with each site selection point in the preliminary site selection plan data; at the same time, collecting the traffic, geographic and population information around the site selection point can provide the characteristic data of the site selection point, which can include the traffic accessibility of the site selection point, the surrounding population density, the distribution of commercial areas and other information. By obtaining the characteristic data of the site selection point, the potential charging demand and suitability of the site selection point can be better understood, providing a basis for subsequent decision-making and analysis. By mapping the site selection point with the charging demand according to the characteristic data of the site selection point, a load curve prediction model for the site selection point can be established; this model can predict the charging load of the future site selection point according to the characteristics of the site selection point and the charging demand data. The load curve prediction model can help understand the electricity consumption characteristics of the site selection point, predict the load level in different time periods, and provide a basis for subsequent charging pile configuration and capacity planning. By using the load curve prediction model, the type and number of charging piles at each site selection point can be configured and determined. According to the predicted load conditions, different types of charging piles (such as fast charging and slow charging) and their quantities can be reasonably arranged to meet the future charging needs of the site selection point. The charging pile configuration data can provide the layout and distribution of the charging piles, providing guidance for the construction of charging facilities and investment decisions. By combining the load curve prediction model, the charging pile configuration data and the existing energy storage conditions, the energy storage demand and role of each site selection point in peak shaving and valley filling can be evaluated. Energy storage can be used to balance the load fluctuations of the power system and improve the stability and reliability of the power system; by evaluating the energy storage demand and role of the site selection point, the energy storage capacity planning data can be determined to provide guidance for the configuration and operation of energy storage equipment and further optimize the operating efficiency of the power system. By evaluating the capacity planning data based on the power infrastructure data, the impact of the new load on the power grid can be analyzed, including evaluating the impact of the new charging demand on the power load, voltage stability, line load and other aspects of the power grid. The power grid impact assessment data can help understand the potential impact of the construction of charging facilities on the power system and provide an important reference for the planning and operation of the power system; through these evaluation data, necessary measures can be taken to solve potential power system bottlenecks and ensure the reliable power supply of charging facilities and the stable operation of the power system.

[0069] Preferably, step S45 includes the following steps: Step S451: Perform data fusion on the charging pile configuration data and the capacity planning data to obtain the newly added load data; compare the newly added load data with the load spectrum of the existing power grid, and identify the impact point to obtain the newly added load characteristic data; The embodiment of the present invention integrates and fuses the charging pile configuration data and the capacity planning data to obtain the newly added load data. This may include information such as the number of charging piles, power, and charging demand. The newly added load data is compared and analyzed with the load spectrum of the existing power grid to determine the impact of the newly added load on the power grid. This step can use power system simulation tools to compare and analyze the loads. Based on the comparative analysis, the points where the newly added load has a greater impact on the operation of the power grid are identified, such as significant load growth and large voltage fluctuations, which helps to further evaluate the characteristics and impact of the newly added load.

[0070] Step S452: constructing a power grid topology model describing the connection relationship between substations, transmission lines and transformer equipment according to the power infrastructure data; The embodiment of the present invention organizes the collected power infrastructure data, including the location, capacity, connection relationship and other information of equipment such as substations, transmission lines and transformers. Based on the power infrastructure data, a model describing the topological structure of the power grid is constructed, which can be represented by a graph structure in graph theory, where nodes represent substations or transformers, and edges represent transmission lines or connection relationships.

[0071] Step S453: Calculate the AC power flow according to the newly added load characteristic data and the power grid topology model, evaluate whether the line power flow and transformer capacity exceed the safety limit, identify risky lines and weak links, and obtain power flow safety assessment data; The embodiment of the present invention uses the power grid topology model and the newly added load characteristic data to perform AC power flow calculation, which can use power flow calculation software or power system simulation tools, such as PowerWorld, PSS / E, etc. According to the power flow calculation results, it is evaluated whether the line power flow and transformer capacity exceed their safety limits, and it is determined whether there are risky lines and weak links, that is, parts that are prone to overload or cause power grid stability problems.

[0072] Step S454: evaluating the voltage condition at each node of the power grid under the newly added load characteristic data on the power grid topology model, thereby obtaining voltage quality evaluation data, wherein the voltage condition evaluation is to detect whether there is a voltage violation condition of overvoltage or undervoltage; The embodiment of the present invention uses the power grid topology model and the newly added load characteristic data to calculate the voltage, and obtains the voltage condition at each node according to the result of the power flow calculation. The calculated voltage condition is evaluated to detect whether there is overvoltage or undervoltage, and the calculation result is compared with the voltage specification requirements of the power grid to determine the voltage violation condition to evaluate the voltage quality.

[0073] Step S455: Summarize the voltage quality assessment data and the power flow safety assessment data, identify the constraint points that have the greatest impact on the grid operation, and generate grid impact assessment data based on the constraint points.

[0074] The embodiment of the present invention aggregates the voltage quality assessment data and the power flow safety assessment data to comprehensively consider the voltage situation and power flow safety; based on the aggregated data, the constraint points that have the greatest impact on the operation of the power grid are identified. This can be nodes or lines with large voltage anomalies or power flows exceeding safety limits; based on the constraint points, power grid impact assessment data is generated to assess the impact of the new load on the operation of the power grid, which can include assessments of power grid stability, voltage control capability, line capacity, etc.

[0075] The present invention can obtain the newly added load data by fusing the charging pile configuration data and the capacity planning data. The newly added load data represents the additional load introduced by the construction of charging facilities and capacity planning. By comparing the newly added load data with the load spectrum of the existing power grid, the impact points of the newly added load on the power grid, that is, the locations and time periods where the load changes greatly, can be identified. In this way, the characteristic data of the newly added load can be obtained, including information such as load peak and volatility. These data are helpful to understand the impact of the newly added load on the operation of the power grid and provide a basis for subsequent power flow calculation and voltage evaluation. The power grid topology model describes the connection relationship between substations, transmission lines and transformer equipment in the power system. By constructing the power grid topology model based on the power infrastructure data, the connection and layout relationship between the various devices in the power system can be accurately represented. Such a model can serve as the basis for subsequent power flow calculation and voltage evaluation, and is used to analyze the power flow and voltage distribution in the power system. The newly added load characteristic data and the grid topology model can be used to perform AC power flow calculations. Power flow calculations can analyze the power flow in the power system, including line power flow and transformer capacity. By evaluating whether the line power flow and transformer capacity exceed the safety limit, risky lines and weak links can be identified. The power flow safety assessment data provides information on power flow distribution and equipment load in the power system, which helps to determine the safety and stability of the power grid. According to the grid topology model and the newly added load characteristic data, the voltage conditions at each node of the power grid can be evaluated. The voltage quality assessment data is used to detect whether the voltage is overvoltage or undervoltage, that is, voltage violation. By evaluating the voltage quality, the voltage stability problems in the power system can be determined, providing a basis for taking corresponding measures to ensure the reliability and quality of power supply. The voltage quality assessment data and power flow safety assessment data are summarized, and the impact of voltage stability and equipment load on the grid operation is comprehensively considered; the constraint points that have the greatest impact on the grid operation are identified, that is, the key locations where there may be voltage problems or equipment overload risks; the grid impact assessment data is generated based on the constraint points, providing the grid operation conditions at key locations, providing a decision-making basis for further optimizing the power system, planning equipment expansion or improving operation strategies.

[0076] Preferably, step S5 comprises the following steps: Step S51: extracting grid constraint information from the grid impact assessment data, and structuring the constraint data to obtain grid constraint data; The embodiment of the present invention is based on the grid impact assessment data generated in the aforementioned step S455, including voltage quality assessment data and flow safety assessment data; constraint information related to the construction of charging piles is extracted from the grid impact assessment data. This may include constraints such as voltage limitation, line capacity limitation, and flow limitation; the extracted grid constraint information is structured to form grid constraint data, which may be in the form of arranging the constraints into a table or data structure in a certain format to facilitate the input and processing of the subsequent optimization model.

[0077] Step S52: construct a two-layer optimization model, using the preliminary site selection plan data and the charging pile configuration data as optimization variables, wherein the upper layer is site selection optimization, the goal is to meet charging demand and minimize grid risk, and the lower layer is configuration optimization, the goal is to minimize the degree of violation of grid constraints; The embodiment of the present invention formulates preliminary charging pile site selection plan data according to demand and related factors, which may include the location, quantity, capacity and other information of candidate site selection points; determines the specific configuration plan of the charging piles, including the number, power, charging speed and other parameters of the charging piles; based on the above data, constructs a two-layer optimization model, the upper layer is a site selection optimization model, which considers the goal of minimizing charging demand and grid risk; the lower layer is a configuration optimization model, which considers the goal of minimizing the degree of violation of grid constraints.

[0078] Step S53: The preliminary site selection plan and charging pile configuration data are input as the initial solution into the two-layer optimization model for solution. During the optimization solution process, the site selection plan and the configuration plan are updated alternately and continuously revised until the constraints in the power grid constraint data are met, thereby obtaining the planning plan data for the charging pile construction that meets the power grid constraints.

[0079] The embodiment of the present invention inputs the preliminary site selection plan and the charging pile configuration data as the initial solution into the double-layer optimization model, and uses relevant optimization algorithms, such as genetic algorithm, particle swarm optimization, etc., to solve the double-layer optimization model; by alternately updating the site selection plan and the configuration plan, the initial solution is gradually corrected so that it gradually satisfies the constraints in the power grid constraint data; in the optimization solution process, the site selection plan and the configuration plan are corrected according to the solution results of the optimization model until the constraints in the power grid constraint data are met.

[0080] The present invention extracts constraint information related to grid operation by analyzing grid impact assessment data; performs structured processing on the extracted constraint information to make it conform to the format and specification of constraint data; obtains grid constraint data, which includes various grid constraints that affect the construction plan of charging piles, such as line capacity restrictions, voltage stability requirements, etc. By constructing a two-layer optimization model, the site selection plan and charging pile configuration are used as optimization variables to achieve comprehensive optimization of the charging pile construction plan; the upper optimization goal is to minimize the risk impact on the grid frame while meeting the charging demand, such as reducing dependence on weak links or avoiding risky lines; the lower optimization goal is to minimize the violation of grid constraints by optimizing the configuration of charging piles on the basis of the preliminary site selection plan, such as avoiding exceeding the line capacity or voltage stability range. The preliminary site selection plan and charging pile configuration data are input into the double-layer optimization model as the initial solution, and the charging pile construction planning plan that meets the grid constraints is obtained by solving the optimization model; in the optimization solution process, the site selection plan and configuration plan are updated alternately, and are continuously revised and optimized to gradually meet the various constraints in the grid constraint data; the final planning plan data will meet the grid constraints, effectively consider the comprehensive factors of charging demand, grid risk and constraints, and provide a scientific basis for the decision-making of charging pile construction. Combining the above steps, by extracting grid constraint information, constructing a double-layer optimization model and optimizing the solution process, the charging pile construction planning plan data that meets the grid constraints can be obtained. These steps help to fully consider the relationship between charging demand, grid risk and constraints, realize the optimization and rationality of charging pile construction planning, and ensure that the layout and configuration of charging piles meet the reliability, stability and safety requirements of the grid.

[0081] The present invention also provides an artificial intelligence-based charging pile construction planning system, which is used to execute the above-mentioned artificial intelligence-based charging pile construction planning method. The artificial intelligence-based charging pile construction planning system includes: The multi-source data fusion module is used to obtain existing power infrastructure data and multi-source heterogeneous data, including traffic data, geographic data, vehicle data, and charging behavior data; extract time series features from multi-source heterogeneous data, and use the preset hierarchical attention network model to fuse the features of different data sources across time scales to obtain factor weight data; The long-term demand forecasting module is used to build a long-term charging demand forecasting model based on factor weight data and multi-source heterogeneous data through a sequence-to-sequence model; based on the long-term charging demand forecasting model, long-term forecasts of future charging demand are made, and uncertainty intervals are evaluated to obtain charging demand forecast data; The risk minimization site selection module is used to select charging pile sites based on the risk minimum set coverage problem according to the charging demand forecast data and power infrastructure data, and to optimize the targets based on coverage, service radius and accessibility to obtain preliminary site selection plan data; The load impact assessment module is used to predict the load curve of the site selection point based on the charging demand forecast data and the preliminary site selection plan data, so as to obtain the charging pile configuration data of each site selection point; generate capacity planning data based on the charging pile configuration data and the existing energy storage conditions; and evaluate the impact of the new load on the power grid based on the capacity planning data according to the power infrastructure data, so as to obtain the power grid impact assessment data; The planning scheme optimization module is used to generate grid constraint data based on the grid impact assessment data, and to make balance corrections to the preliminary site selection scheme data and charging pile configuration data based on the charging demand and grid stability according to the grid constraint data, so as to obtain the planning scheme data for the charging pile construction.

[0082] The present invention can comprehensively utilize information from various data sources to support charging pile construction planning by acquiring existing power infrastructure data and multi-source heterogeneous data (including traffic data, geographic data, vehicle data, and charging behavior data). By extracting time series features from multi-source heterogeneous data, the temporal variation pattern of the data can be captured, providing valuable feature information for subsequent demand forecasting and site selection decisions. By using a preset hierarchical attention network model to fuse the features of different data sources across time scales, the information of multi-source data can be effectively integrated to improve the data characterization capability. By fusing multi-source data and extracting factor weight data, the contribution of different data sources to charging demand can be quantified, providing an important reference for subsequent demand forecasting and site selection decisions. Based on factor weight data and multi-source heterogeneous data, a sequence-to-sequence model is constructed for long-term charging demand forecasting; this can help planners understand the trend and change of future charging demand and provide long-term planning and decision-making basis for charging pile construction. By performing uncertainty interval evaluation, the confidence interval of charging demand forecast data can be provided to help planners understand the reliability and uncertainty of the forecast results. Combined with charging demand forecast data and power infrastructure data, the method based on the risk minimum set covering problem is used to select the location of charging piles, which can ensure that the layout of charging piles fully covers the charging demand and minimizes the risk of power grid, such as reducing the potential risk of power system overload. Optimization based on coverage, service radius and accessibility can make the site selection plan more reasonable and effective, and improve the availability and convenience of charging services. By predicting the load curve of the charging demand forecast data and the preliminary site selection plan data, the load situation at the site selection point can be evaluated to provide a reference for the configuration of charging piles; generating capacity planning data, combined with existing energy storage conditions, can ensure that the energy storage capacity in the charging pile construction plan meets the demand and optimize energy utilization efficiency. According to the power infrastructure data, the impact of the new load on the power grid is evaluated on the capacity planning data, which can evaluate the impact of the construction of charging piles on the power grid, discover potential power grid problems in advance, and take corresponding measures to adjust and optimize. Generate power grid constraint data based on the power grid impact assessment data to ensure that the charging pile construction plan meets the reliability, stability and safety requirements of the power grid. Based on the balance correction of charging demand and power grid stability, the preliminary site selection plan data and charging pile configuration data are optimized to obtain the planning plan data for charging pile construction that meets the power grid constraints. By optimizing the planning scheme, we can comprehensively consider charging demand, grid constraints and other factors, improve the rationality and feasibility of charging pile construction planning, and achieve the optimal layout of charging services.

[0083] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0084] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A charging pile construction planning method based on artificial intelligence, characterized in that: The following steps are involved: Step S1: Acquire existing power infrastructure data and multi-source heterogeneous data, wherein the multi-source heterogeneous data includes traffic data, geographic data, vehicle data, and charging behavior data; Extract time series features from multi-source heterogeneous data, and use the preset hierarchical attention network model to fuse features from different data sources across time scales to obtain factor weight data; Step S2: constructing a long-term charging demand prediction model based on factor weight data and multi-source heterogeneous data through a sequence-to-sequence model; performing long-term prediction of future charging demand based on the long-term charging demand prediction model, and performing uncertainty interval evaluation to obtain charging demand prediction data; Step S3: Perform charging pile site selection based on the risk minimum set coverage problem according to the charging demand forecast data and the power infrastructure data, and perform target optimization based on coverage, service radius and accessibility to obtain preliminary site selection plan data; Step S4: predicting the load curve of the site selection point according to the charging demand forecast data and the preliminary site selection plan data, thereby obtaining the charging pile configuration data of each site selection point; generating capacity planning data according to the charging pile configuration data and the existing energy storage conditions; and evaluating the impact of the new load on the power grid according to the power infrastructure data on the capacity planning data, thereby obtaining the power grid impact assessment data; Step S5: Generate grid constraint data based on the grid impact assessment data, and make balance corrections to the preliminary site selection plan data and charging pile configuration data based on charging demand and grid stability according to the grid constraint data, so as to obtain planning plan data for charging pile construction.

2. The method for planning charging pile construction based on artificial intelligence according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Acquire existing power infrastructure data, including substation data, distribution network data, and line capacity information; Step S12: Collect road network, traffic flow and congestion information to obtain traffic data; obtain geographic data, including terrain information data, land use data and urban planning data; obtain vehicle data, including new energy vehicle distribution data, mileage data and charging frequency data; obtain charging behavior data, including charging time data, charging amount data and charging location data; Step S13: preprocessing the traffic data, geographic data, vehicle data, and charging behavior data, and performing time alignment to obtain multi-source heterogeneous data; Step S14: extract time series features from multi-source heterogeneous data, and use a preset hierarchical attention network model to fuse features from different data sources across time scales to obtain factor weight data.

3. The method for planning charging pile construction based on artificial intelligence according to claim 2 is characterized in that: Step S14 includes the following steps: Step S141: divide the multi-source heterogeneous data into time windows, extract time series statistical features, and extract time-frequency domain features through wavelet transformation to obtain time series feature data; Step S142: performing rasterization processing on multi-source heterogeneous data based on geographic space, and extracting spatial clustering features, thereby constructing a spatial adjacency matrix; Step S143: performing feature dimensionality reduction based on principal component analysis according to the spatial adjacency matrix and the time series feature data, and performing recursive feature elimination to obtain key feature data; Step S144: designing a multi-level attention mechanism according to the key feature data, constructing a hierarchical network structure adapted to features of different scales, thereby obtaining a hierarchical attention network model; Step S145: Adaptively weighting the key feature data using a hierarchical attention network model, and fusing features across time scales to obtain multi-scale feature data; Step S146: Evaluate the impact weight of the charging demand according to the multi-scale feature data to obtain factor weight data.

4. The method for planning charging pile construction based on artificial intelligence according to claim 3 is characterized in that: Step S2 includes the following steps: Step S21: constructing a long-term charging demand prediction model based on factor weight data and multi-source heterogeneous data using a long short-term memory network sequence model, wherein the factor weight data is used as an additional input of the sequence model and the multi-source heterogeneous data is used as the main input of the sequence model when constructing the long-term charging demand prediction model; Step S22: performing a long-term prediction of future charging demand through a long-term charging demand prediction model, thereby obtaining charging demand prediction result data; Step S23: Generate multiple groups of prediction results according to the charging demand prediction result data by using the Monte Carlo simulation method, and estimate the upper and lower bounds of the prediction by using the quantile regression method, so as to obtain uncertainty interval data; Step S24: using the SHAP value to interpret the prediction results of the uncertainty interval data, thereby obtaining charging demand prediction data.

5. The method for planning charging pile construction based on artificial intelligence according to claim 4 is characterized in that: Step S3 includes the following steps: Step S31: Mapping charging demand forecast data and power infrastructure data to a geographic grid, thereby obtaining geographical demand distribution data and infrastructure distribution data; Step S32: performing hot spot analysis on the demand distribution data to determine high-demand areas, and screening available locations with land use restrictions based on the infrastructure distribution data to obtain a set of candidate charging pile locations; Step S33: performing a location risk assessment based on natural disaster risks on the candidate charging pile location set, thereby obtaining location risk score data; Step S34: Establishing a maximum coverage location problem model for the candidate charging pile location set according to the preset service radius and coverage rate index, thereby obtaining a location coverage matrix; Step S35: Calculate the road network distance and driving time according to the candidate charging pile location set and the road network map information, so as to obtain location accessibility data; Step S36: establishing a multi-objective integer programming model according to the location risk score data, the location coverage matrix, and the location accessibility data, thereby obtaining a multi-objective optimization model; Step S37: Optimize the charging pile site selection for the candidate charging pile location set using the NSGA-II algorithm according to the multi-objective optimization model, thereby obtaining preliminary site selection plan data.

6. The method for planning charging pile construction based on artificial intelligence according to claim 5 is characterized in that: Step S37 includes the following steps: Step S371: Solve the multi-objective optimization model using the NSGA-II algorithm, and calculate the comprehensive score of each solution, thereby obtaining a screened optimal location solution candidate set; Step S372: Performing site condition evaluation based on the optimal site solution candidate set and geographic data to obtain site adaptability evaluation data; Step S373: sorting the site selection priorities of the optimal site selection candidate set according to the site adaptability evaluation data, thereby obtaining a site selection candidate set; Step S374: Visually display the coverage and service area of ​​the sorted candidate set of site selection schemes, thereby obtaining visualization data of the site selection schemes; Step S375: performing parameter sensitivity evaluation on the multi-objective optimization model according to the optimal location solution candidate set, thereby obtaining parameter sensitivity analysis data; Step S376: Screening the site selection scheme visualization data based on the parameter sensitivity analysis data using a multi-criteria decision-making method to obtain preliminary site selection scheme data.

7. The method for planning charging pile construction based on artificial intelligence according to claim 6 is characterized in that: Step S4 includes the following steps: Step S41: Allocate charging demand forecast data to each site selection point in the preliminary site selection plan data, and collect traffic, geographic and population information around the site selection point to obtain site selection point feature data; Step S42: mapping the site selection point to the charging demand according to the site selection point characteristic data, thereby generating a load curve prediction model for the site selection point; Step S43: Determine the type and quantity of charging piles for each site selection point according to the load curve prediction model, thereby obtaining charging pile configuration data; Step S44: Evaluate the role of energy storage demand at each site in peak load shaving and valley filling based on the load curve prediction model, charging pile configuration data, and existing energy storage conditions, so as to obtain capacity planning data; Step S45: Evaluate the impact of the newly added load on the power grid based on the capacity planning data according to the power infrastructure data, thereby obtaining power grid impact evaluation data.

8. The method for planning charging pile construction based on artificial intelligence according to claim 7 is characterized in that: Step S45 includes the following steps: Step S451: Perform data fusion on the charging pile configuration data and the capacity planning data to obtain the newly added load data; compare the newly added load data with the load spectrum of the existing power grid, and identify the impact point to obtain the newly added load characteristic data; Step S452: constructing a power grid topology model describing the connection relationship between substations, transmission lines and transformer equipment according to the power infrastructure data; Step S453: Calculate the AC power flow according to the newly added load characteristic data and the power grid topology model, evaluate whether the line power flow and transformer capacity exceed the safety limit, identify risky lines and weak links, and obtain power flow safety assessment data; Step S454: evaluating the voltage condition at each node of the power grid under the newly added load characteristic data on the power grid topology model, thereby obtaining voltage quality evaluation data, wherein the voltage condition evaluation is to detect whether there is a voltage violation condition of overvoltage or undervoltage; Step S455: Summarize the voltage quality assessment data and the power flow safety assessment data, identify the constraint points that have the greatest impact on the grid operation, and generate grid impact assessment data based on the constraint points.

9. The method for planning charging pile construction based on artificial intelligence according to claim 8, characterized in that: Step S5 includes the following steps: Step S51: extracting grid constraint information from the grid impact assessment data, and structuring the constraint data to obtain grid constraint data; Step S52: construct a two-layer optimization model, using the preliminary site selection plan data and the charging pile configuration data as optimization variables, wherein the upper layer is site selection optimization, the goal is to meet charging demand and minimize grid risk, and the lower layer is configuration optimization, the goal is to minimize the degree of violation of grid constraints; Step S53: The preliminary site selection plan and charging pile configuration data are input as the initial solution into the two-layer optimization model for solution. During the optimization solution process, the site selection plan and the configuration plan are updated alternately and continuously revised until the constraints in the power grid constraint data are met, thereby obtaining the planning plan data for the charging pile construction that meets the power grid constraints.

10. A charging pile construction planning system based on artificial intelligence, characterized in that: Used to execute the artificial intelligence-based charging pile construction planning method according to claim 1, the artificial intelligence-based charging pile construction planning system comprises: The multi-source data fusion module is used to obtain existing power infrastructure data and multi-source heterogeneous data, including traffic data, geographic data, vehicle data, and charging behavior data; extract time series features from multi-source heterogeneous data, and use the preset hierarchical attention network model to fuse the features of different data sources across time scales to obtain factor weight data; The long-term demand forecasting module is used to build a long-term charging demand forecasting model based on factor weight data and multi-source heterogeneous data through a sequence-to-sequence model; based on the long-term charging demand forecasting model, long-term forecasts of future charging demand are made, and uncertainty intervals are evaluated to obtain charging demand forecast data; The risk minimization site selection module is used to select charging pile sites based on the risk minimum set coverage problem according to the charging demand forecast data and power infrastructure data, and to optimize the targets based on coverage, service radius and accessibility to obtain preliminary site selection plan data; The load impact assessment module is used to predict the load curve of the site selection point based on the charging demand forecast data and the preliminary site selection plan data, so as to obtain the charging pile configuration data of each site selection point; generate capacity planning data based on the charging pile configuration data and the existing energy storage conditions; and evaluate the impact of the new load on the power grid based on the capacity planning data according to the power infrastructure data, so as to obtain the power grid impact assessment data; The planning scheme optimization module is used to generate grid constraint data based on the grid impact assessment data, and to make balance corrections to the preliminary site selection scheme data and charging pile configuration data based on the charging demand and grid stability according to the grid constraint data, so as to obtain the planning scheme data for the charging pile construction.

Images