A power distribution network load characteristic analysis management method and system containing an ultra-fast charging station

Abstract

The application provides a power distribution network load characteristic analysis management method and system containing an ultra-fast charging station, comprising the following steps: clustering and load analysis of each power supply unit in the power distribution network by using the historical load data and land use properties of each power supply unit to obtain a general load scenario model of each power supply unit; based on the historical load data, performing load curve similarity analysis and clustering on the ultra-fast charging station, and performing typical load characteristic analysis on the clustering results to generate a specific load scenario model of each ultra-fast charging station; inputting the current operating parameters of each power supply unit and the corresponding general load scenario model into a load prediction model to generate a load prediction value, and similarly generating a load prediction value of the ultra-fast charging station, and generating a dispatching instruction according to the load prediction value; the scheme differentiates load scenario modeling, and inputs the load scenario model into the prediction model to provide prior knowledge for prediction, thereby improving the prediction accuracy and providing accurate data basis for dispatching analysis.

Description

Technical Field

[0001] This invention relates to the field of power distribution network load analysis technology, and specifically to a method and system for analyzing and managing the load characteristics of a power distribution network including an ultra-fast charging station. Background Technology

[0002] With the popularization of electric vehicles and the development of charging infrastructure, ultra-fast charging stations have been widely used as key facilities. Ultra-fast charging stations can provide high-power charging services for electric vehicles in a very short time, significantly reducing user waiting time and improving user experience. However, this high-power demand brings unprecedented challenges to the power distribution network, as ultra-fast charging stations release high power demand in a concentrated period of time, exhibiting significant temporal concentration and spatial distribution.

[0003] However, existing power distribution network systems still use traditional load modeling methods and have not established specialized load models for the special characteristics of ultra-fast charging stations, resulting in load forecasting and scheduling strategies being unable to effectively cope with this new type of load.

[0004] Traditional load modeling methods often employ linear regression, ARIMA (Auto-Regressive Integrated Moving Average Model), and support vector machines for load forecasting. These methods lack the ability to model the characteristics of ultra-fast charging loads, have limited modeling dimensions, and long update cycles. They are unable to handle high-dimensional, nonlinear, and time-varying load data, resulting in low prediction accuracy and slow response. This can lead to delayed grid dispatching strategies or misjudgments of load peaks, preventing timely load transfer or energy storage adjustments. Ultimately, this can cause safety issues such as sudden voltage drops, transformer overheating, and line overloads. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, this invention proposes a method for analyzing and managing the load characteristics of a distribution network including an ultra-fast charging station, comprising: By utilizing the historical load data and land use characteristics of each power supply unit in the distribution network, clustering and load analysis are performed on multiple power supply units to obtain a general load scenario model for each type of power supply unit; some of the power supply units contain ultra-fast charging stations; Based on the historical load data of each of the ultra-fast charging stations, the load curve similarity analysis and clustering of multiple ultra-fast charging stations are performed using dynamic time warping algorithm and spectral clustering algorithm. Then, the typical load characteristics of each type of ultra-fast charging station obtained by clustering are analyzed to generate a specific load scenario model for each type of ultra-fast charging station. The current operating parameters of each power supply unit and the corresponding general load scenario model are input into the load prediction model to generate the load prediction value of each power supply unit. The current operating parameters of each ultra-fast charging station and the corresponding specific load scenario model are input into the load prediction model to generate the load prediction value of each ultra-fast charging station. The scheduling instructions are generated based on the analysis of the load prediction values ​​of multiple power supply units and multiple ultra-fast charging stations.

[0006] Optionally, the step of using historical load data and land use characteristics of each power supply unit in the distribution network to perform clustering and load analysis on multiple power supply units to obtain a general load scenario model for each type of power supply unit includes: Load pattern characteristics and time sequence characteristics are extracted from the historical load data of each power supply unit, and the land use nature of each power supply unit is converted into numerical characteristics; Based on each power supply unit, the multi-source characteristics of each power supply unit are obtained by integrating the load pattern characteristics, timing characteristics and numerical characteristics of the power supply unit; Based on the multi-source characteristics of each power supply unit, the k-means clustering algorithm combined with the DBSCAN algorithm is used to cluster multiple power supply units to obtain multiple power supply unit clusters; Based on the average load curve and load characteristic parameters in the cluster of each power supply unit, a general load scenario model for each power supply unit is constructed.

[0007] Optionally, based on the historical load data of each of the ultra-fast charging stations, the load curve similarity analysis and clustering of multiple ultra-fast charging stations are performed using dynamic time warping and spectral clustering algorithms. Then, typical load characteristic analysis is performed on each type of ultra-fast charging station obtained from the clustering to generate a specific load scenario model for each type of ultra-fast charging station, including: A standard load curve for each ultra-fast charging station is generated based on the historical load data of each ultra-fast charging station. The similarity of the standard load curves of every two ultra-fast charging stations is calculated using the dynamic time warping algorithm, and a DTW distance matrix is ​​constructed based on the similarity of the standard load curves of every two ultra-fast charging stations. The standard load curve is treated as a node using a spectral clustering algorithm, and the connection strength between each pair of nodes is defined using the DTW distance matrix to construct a similarity graph. The similarity map is mapped to a low-dimensional space by calculating the Laplacian matrix eigenvector of the similarity map. The k-means algorithm is then used to cluster the standard load curves of multiple ultra-fast charging stations in the low-dimensional space to obtain clusters of various ultra-fast charging stations. Based on the median of the standard load curve, the peak and valley values, the occurrence time of the peak and valley values, and the curve shape parameters in the cluster of each ultra-fast charging station, a specific load scenario model for each ultra-fast charging station is constructed.

[0008] Optionally, the step of inputting the current operating parameters of each power supply unit and the corresponding general load scenario model into the load prediction model to generate the load prediction value for each power supply unit includes: Based on each of the power supply units: The current operating parameters of the power supply unit are standardized and then concatenated to form the feature vector of the power supply unit; the current operating parameters include electricity consumption data, geographical location data, and time dimension data; A spatial graph is constructed by taking the power supply unit as a node and the geographical relationship between the power supply unit and multiple other power supply units in the distribution network as edges, and a time series is constructed based on the feature vector of the power supply unit for a certain period of time. The time series is input into an LSTM to generate the temporal coupling characteristics of the power supply unit, and the spatial graph is input into a graph neural network. After coupling, the spatiotemporal coupling characteristics of the power supply unit are obtained. The spatiotemporal coupling characteristics of the power supply unit, the category of the corresponding general load scenario model, and the typical load characteristics are input into the load prediction model to perform load prediction, thereby obtaining the load prediction value of the power supply unit.

[0009] Optionally, the step of generating scheduling instructions based on load forecast values ​​of the multiple power supply units and the multiple ultra-fast charging stations includes: The load forecast values ​​of the multiple power supply units and the multiple ultra-fast charging stations, as well as the grid operation status, energy storage status and scheduling objectives are defined as the decision state space; The power adjustment strategies of the multiple power supply units and the multiple ultra-fast charging stations are defined as the action space; A reward function is constructed based on the load deviation, dispatching cost, and power supply stability of the aforementioned distribution network; Based on the real-time information in the decision state space, the reinforcement learning algorithm is driven to search for the optimal strategy in the action space under the guidance of the reward function, and outputs the scheduling instructions for each power supply unit and each ultra-fast charging station.

[0010] Optionally, after generating scheduling instructions based on the load forecast values ​​of each of the power supply units and the ultra-fast charging stations, the method further includes: According to a preset time interval, under the premise of satisfying preset constraints, the objective function of the distribution network optimization scheduling is solved by a genetic algorithm to obtain the global optimal scheduling scheme of the distribution network. After adjusting the operation of each power supply unit and the ultra-fast charging station according to the scheduling instructions, the operation of each power supply unit and the ultra-fast charging station is adjusted again using the global scheduling optimal scheme; The constraints include one or more of the following: power balance constraints, energy storage capacity constraints, and equipment operation constraints. The objective function is constructed to minimize the total scheduling cost and maximize power supply stability. The expression of the objective function is:

[0011] In the formula, This represents the scheduling cost coefficient of the i-th unit. The scheduling variable for the i-th unit, This indicates the number of units in the power distribution network, including power supply units and ultra-fast charging stations. Indicates the weighting coefficient. This indicates the loss of power supply stability in the distribution network.

[0012] Optionally, after generating scheduling instructions based on the load forecast values ​​of each of the power supply units and the ultra-fast charging stations, the method further includes: Obtain a multidimensional sensing dataset of the power distribution network, which includes electrical quantity data, complaint data, and external factor data; The load factor is calculated based on the electrical quantity data of each area in the distribution network, and a color map layer reflecting the distribution is generated based on the load factor using a color mapping function. A scatter plot layer reflecting the distribution of complaints is generated based on the location of the complaint data in the power distribution network. The color patch map layer and scatter plot layer are overlaid on the GIS platform to visualize the power grid operation status.

[0013] Optionally, after generating scheduling instructions based on the load forecast values ​​of each of the power supply units and the ultra-fast charging stations, the method further includes: Acquire historical power grid operation data, and extract basic indicator features from the historical power grid operation data. The basic indicator features include time-series features, statistical features, and frequency domain features. Fault features are extracted from the basic index features using a convolutional neural network, and the fault features are input into a random forest for state classification to obtain the fault category; The initial risk prediction model is trained based on historical power grid operation data containing fault categories to obtain a risk prediction model. The risk prediction model is then used to predict risks based on current power grid operation data and output the risk level. The alarm mechanism of the power distribution network is triggered when the risk level reaches the risk threshold.

[0014] Based on the same inventive concept, this invention proposes a power distribution network load characteristic analysis and management system including an ultra-fast charging station, comprising: A general load scenario construction module is used to cluster and analyze multiple power supply units by utilizing historical load data and land use characteristics of each power supply unit in the distribution network, and obtain a general load scenario model for each power supply unit; some of the power supply units contain ultra-fast charging stations; The specific load scenario construction module is used to perform load curve similarity analysis and clustering on multiple ultra-fast charging stations based on the historical load data of each of the ultra-fast charging stations, using dynamic time warping algorithm and spectral clustering algorithm, and to perform typical load characteristic analysis on each type of ultra-fast charging station obtained by clustering, so as to generate a specific load scenario model for each type of ultra-fast charging station. The scheduling instruction generation module is used to input the current operating parameters of each power supply unit and the corresponding general load scenario model into the load prediction model to generate the load prediction value of each power supply unit, and to input the current operating parameters of each ultra-fast charging station and the corresponding specific load scenario model into the load prediction model to generate the load prediction value of each ultra-fast charging station, and to generate scheduling instructions based on the load prediction values ​​of multiple power supply units and multiple ultra-fast charging stations.

[0015] Optionally, the general load scenario construction module is specifically used for: Load pattern characteristics and time sequence characteristics are extracted from the historical load data of each power supply unit, and the land use nature of each power supply unit is converted into numerical characteristics; Based on each power supply unit, the multi-source characteristics of each power supply unit are obtained by integrating the load pattern characteristics, timing characteristics and numerical characteristics of the power supply unit; Based on the multi-source characteristics of each power supply unit, the k-means clustering algorithm combined with the DBSCAN algorithm is used to cluster multiple power supply units to obtain multiple power supply unit clusters; Based on the average load curve and load characteristic parameters in the cluster of each power supply unit, a general load scenario model for each power supply unit is constructed.

[0016] Optionally, the specific load scenario construction module is specifically used for: A standard load curve for each ultra-fast charging station is generated based on the historical load data of each ultra-fast charging station. The similarity of the standard load curves of every two ultra-fast charging stations is calculated using the dynamic time warping algorithm, and a DTW distance matrix is ​​constructed based on the similarity of the standard load curves of every two ultra-fast charging stations. The standard load curve is treated as a node using a spectral clustering algorithm, and the connection strength between each pair of nodes is defined using the DTW distance matrix to construct a similarity graph. The similarity map is mapped to a low-dimensional space by calculating the Laplacian matrix eigenvector of the similarity map. The k-means algorithm is then used to cluster the standard load curves of multiple ultra-fast charging stations in the low-dimensional space to obtain clusters of various ultra-fast charging stations. Based on the median of the standard load curve, the peak and valley values, the occurrence time of the peak and valley values, and the curve shape parameters in the cluster of each ultra-fast charging station, a specific load scenario model for each ultra-fast charging station is constructed.

[0017] Optionally, the scheduling instruction generation module is specifically used for: Based on each of the power supply units: The current operating parameters of the power supply unit are standardized and then concatenated to form the feature vector of the power supply unit; the current operating parameters include electricity consumption data, geographical location data, and time dimension data; A spatial graph is constructed by taking the power supply unit as a node and the geographical relationship between the power supply unit and multiple other power supply units in the distribution network as edges, and a time series is constructed based on the feature vector of the power supply unit for a certain period of time. The time series is input into an LSTM to generate the temporal coupling characteristics of the power supply unit, and the spatial graph is input into a graph neural network. After coupling, the spatiotemporal coupling characteristics of the power supply unit are obtained. The spatiotemporal coupling characteristics of the power supply unit, the category of the corresponding general load scenario model, and the typical load characteristics are input into the load prediction model to perform load prediction, thereby obtaining the load prediction value of the power supply unit.

[0018] Optionally, the scheduling instruction generation module is specifically used for: The load forecast values ​​of the multiple power supply units and the multiple ultra-fast charging stations, as well as the grid operation status, energy storage status and scheduling objectives are defined as the decision state space; The power adjustment strategies of the multiple power supply units and the multiple ultra-fast charging stations are defined as the action space; A reward function is constructed based on the load deviation, dispatching cost, and power supply stability of the aforementioned distribution network; Based on the real-time information in the decision state space, the reinforcement learning algorithm is driven to search for the optimal strategy in the action space under the guidance of the reward function, and outputs the scheduling instructions for each power supply unit and each ultra-fast charging station.

[0019] Optionally, the system further includes a global optimization module, used for: According to a preset time interval, under the premise of satisfying preset constraints, the objective function of the distribution network optimization scheduling is solved by a genetic algorithm to obtain the global optimal scheduling scheme of the distribution network. After adjusting the operation of each power supply unit and the ultra-fast charging station according to the scheduling instructions, the operation of each power supply unit and the ultra-fast charging station is adjusted again using the global scheduling optimal scheme; The constraints include one or more of the following: power balance constraints, energy storage capacity constraints, and equipment operation constraints. The objective function is constructed to minimize the total scheduling cost and maximize power supply stability. The expression of the objective function is:

[0020] In the formula, This represents the scheduling cost coefficient of the i-th unit. The scheduling variable for the i-th unit, This indicates the number of units in the power distribution network, including power supply units and ultra-fast charging stations. Indicates the weighting coefficient. This indicates the loss of power supply stability in the distribution network.

[0021] Optionally, the system further includes a visualization module for: Obtain a multidimensional sensing dataset of the power distribution network, which includes electrical quantity data, complaint data, and external factor data; The load factor is calculated based on the electrical quantity data of each area in the distribution network, and a color map layer reflecting the distribution is generated based on the load factor using a color mapping function. A scatter plot layer reflecting the distribution of complaints is generated based on the location of the complaint data in the power distribution network. The color patch map layer and scatter plot layer are overlaid on the GIS platform to visualize the power grid operation status.

[0022] Optionally, the system further includes a risk warning module, used for: Acquire historical power grid operation data, and extract basic indicator features from the historical power grid operation data. The basic indicator features include time-series features, statistical features, and frequency domain features. Fault features are extracted from the basic index features using a convolutional neural network, and the fault features are input into a random forest for state classification to obtain the fault category; The initial risk prediction model is trained based on historical power grid operation data containing fault categories to obtain a risk prediction model. The risk prediction model is then used to predict risks based on current power grid operation data and output the risk level. The alarm mechanism of the power distribution network is triggered when the risk level reaches the risk threshold.

[0023] In another aspect, this application also provides an electronic device, comprising: at least one processor and a memory; the memory and the processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, a method for analyzing and managing the load characteristics of a distribution network including an ultra-fast charging station, as described above, is implemented.

[0024] In another aspect, this application also provides a computer-readable storage medium having an executable program stored thereon, which, when executed, implements the above-described method for analyzing and managing the load characteristics of a distribution network including an ultra-fast charging station.

[0025] Compared with the closest existing technology, the present invention has the following beneficial effects: This invention provides a method and system for analyzing and managing the load characteristics of a distribution network containing ultra-fast charging stations. The method includes: using historical load data and land use characteristics of each power supply unit in the distribution network, clustering and analyzing the load of multiple power supply units to obtain a general load scenario model for each power supply unit; some of the power supply units contain ultra-fast charging stations; based on the historical load data of each ultra-fast charging station, using dynamic time warping and spectral clustering algorithms, performing load curve similarity analysis and clustering on multiple ultra-fast charging stations, and performing typical load characteristic analysis on each type of ultra-fast charging station obtained from the clustering to generate a specific load scenario model for each type of ultra-fast charging station; inputting the current operating parameters of each power supply unit and the corresponding general load scenario model into a load prediction model to generate a load prediction value for each power supply unit, and inputting the current operating parameters of each ultra-fast charging station and... The corresponding specific load scenario model is input into the load prediction model to generate the load prediction value for each of the ultra-fast charging stations. Based on the load prediction values ​​of multiple power supply units and multiple ultra-fast charging stations, dispatch instructions are generated. In this scheme, different types of general load scenario models are constructed based on the land use nature and load data analysis of each power supply unit, and the load curves of different ultra-fast charging stations are modeled differently. This enables accurate load prediction analysis for different units (power supply units and ultra-fast charging stations). Inputting the general load scenario model of the power supply unit (including scenario model categories and typical load characteristics) and the specific load scenario model of the ultra-fast charging station into the prediction model provides strong physical constraints and prior knowledge, improving prediction accuracy and providing an accurate data foundation for subsequent dispatch analysis, thus avoiding power grid safety problems caused by improper dispatch. Attached Figure Description

[0026] Figure 1 A flowchart illustrating a method for analyzing and managing the load characteristics of a power distribution network including an ultra-fast charging station, provided by the present invention. Figure 2 A schematic diagram of the structure of a power distribution network load characteristic analysis and management system including an ultra-fast charging station provided by the present invention; Figure 3 This is a schematic diagram of the structure of an electronic device provided by the present invention. Detailed Implementation

[0027] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0028] Example 1 This invention provides a method for analyzing and managing the load characteristics of a distribution network including an ultra-fast charging station, such as... Figure 1 As shown, it includes: S1. Using the historical load data and land use characteristics of each power supply unit in the distribution network, clustering and load analysis are performed on multiple power supply units to obtain a general load scenario model for each power supply unit; some of the power supply units contain ultra-fast charging stations; S2. Based on the historical load data of each of the ultra-fast charging stations, the load curve similarity analysis and clustering of multiple ultra-fast charging stations are performed using dynamic time warping algorithm and spectral clustering algorithm. Typical load characteristics analysis is performed on each type of ultra-fast charging station obtained by clustering to generate a specific load scenario model for each type of ultra-fast charging station. S3. Input the current operating parameters of each power supply unit and the corresponding general load scenario model into the load prediction model to generate the load prediction value of each power supply unit, and input the current operating parameters of each ultra-fast charging station and the corresponding specific load scenario model into the load prediction model to generate the load prediction value of each ultra-fast charging station. Analyze and generate scheduling instructions based on the load prediction values ​​of multiple power supply units and multiple ultra-fast charging stations.

[0029] In step S1, 1-1, load pattern characteristics and time sequence characteristics are extracted from the historical load data of each power supply unit, and the land use nature of each power supply unit is converted into numerical characteristics; based on each power supply unit, the load pattern characteristics, time sequence characteristics and numerical characteristics of the power supply unit are fused to obtain the multi-source characteristics of each power supply unit.

[0030] This step involves normalizing the historical load curves of each power supply unit and integrating static labels such as land use type to construct a feature vector describing the load pattern and attributes.

[0031] Specifically, historical load data (such as a 96-point daily curve) is acquired for each power supply unit. The historical load data is then cleaned to remove missing and outlier values. To eliminate the influence of dimensions, the historical load data is normalized. In this embodiment, the normalization method is to divide the historical load data of each power supply unit by the maximum load of that unit, and then plot the load curve of each power supply unit based on the normalized value, focusing on the shape of the load curve rather than its absolute size.

[0032] Key statistical features are extracted from the normalized load curve, including: average daily load rate, minimum load rate, peak-valley difference rate, and load fluctuation rate, as the conformal characteristics of the power supply unit; and time-series features are extracted from the load curve, including the start time, peak value, and duration of the morning peak, noon peak, and evening peak.

[0033] The land use type (such as commercial, residential, or industrial) of each power supply unit is individually thermally encoded and converted into numerical features.

[0034] Based on each power supply unit, the multi-source characteristics of each power supply unit are obtained by integrating the load pattern characteristics, timing characteristics, and numerical characteristics of the power supply unit.

[0035] 1-2. Based on the multi-source characteristics of each power supply unit, the k-means clustering algorithm combined with the DBSCAN algorithm (Density-Based Spatial Clustering of Applications with Noise) is used to cluster multiple power supply units to obtain clusters of various power supply units.

[0036] The k-means algorithm is used for main clustering, which has the advantages of high efficiency, fast convergence, and suitability for large-scale data. The silhouette coefficient method is used to calculate the silhouette coefficient of the clustering results under different k values. The k value with the largest silhouette coefficient is selected as the optimal number of clusters to find the best number of clusters. The silhouette coefficient measures the degree of compactness between a sample point and other points in its own cluster, and the degree of separation from neighboring clusters. Simultaneously, the DBSCAN algorithm can be used as an auxiliary and verification method. DBSCAN can effectively identify anomalous units far from the main clusters (such as factories or data centers with special power consumption behaviors), ensuring the purity of the core clustering results.

[0037] Specifically, the k-means clustering algorithm is used to cluster multiple power supply units based on the multi-source features of each power supply unit, resulting in multiple power supply unit clusters; The DBSCAN algorithm is used to perform secondary analysis on each cluster to identify and isolate abnormal power supply units within each cluster. Abnormal power supply units refer to power supply units in misclassified clusters. Recalculate the distance (such as Euclidean distance or cosine similarity) between each abnormal power supply unit and the cluster center of each type of power supply unit, and assign them to the cluster with the closest distance and the highest similarity.

[0038] Thus, clusters of various power supply units were obtained. By combining the k-means clustering algorithm and the DBSCAN algorithm, the purity and accuracy of the final clusters were ensured while maintaining the efficiency and effectiveness of the core clustering.

[0039] 1-3. Based on the average load curve and load characteristic parameters in the cluster of each power supply unit, construct a general load scenario model for each power supply unit.

[0040] Specifically, based on the clustering of each type of power supply unit, the average curve of the load curves of all power supply units in the cluster is calculated to form a standard typical daily load curve. This average load curve represents the most core and most common time-series pattern of this type of load.

[0041] More than just providing an average curve, it's about building a parametric model library to extract and record the following load characteristic parameters for each general load scenario model: Basic parameters: key point values ​​on a typical daily load curve (such as peak values ​​and valley values); morphological parameters: peak-valley difference rate, load factor, and peak period range; fluctuation range: calculate the standard deviation of the curve within the cluster at each time point to form the confidence interval of the typical daily load curve, so as to characterize the volatility of this type of load.

[0042] Based on the average load curve and load characteristic parameters in the clusters of each power supply unit, a general load scenario model for each power supply unit is constructed. Each load scenario model corresponds to a typical power consumption pattern, which lays a solid foundation for subsequent accurate matching and efficient prediction.

[0043] Directly classifying by land use is a static and rigid classification based on administrative or planning labels, assuming that "commercial areas must have the same electricity consumption behavior." Clustering based on historical load data and land use characteristics, on the other hand, is a dynamic, data-driven discovery based on actual electricity consumption data, revealing more complex realities. For example, a commercial area located within a large office building complex (with two peak hours throughout the day) and a commercial area near a night market (with a single peak hour at night), despite having the same land use, will be identified as two different clusters by the clustering algorithm based on the differences in the shape of their actual load curves.

[0044] Therefore, clustering, based on land use characteristics, further reveals the refined electricity consumption patterns that actually exist within the land use characteristics due to differences in functions, user habits, etc., thereby constructing a more accurate load model.

[0045] In step S2, 2-1, a standard load curve for each ultra-fast charging station is generated based on the historical load data of each ultra-fast charging station.

[0046] Specifically, the historical load data of each ultra-fast charging station may be at non-uniform time intervals, and must be uniformly resampled to a fixed time interval (such as 15 minutes or 1 hour per point) to form a regular time series; Identifying and handling abnormal high / low load points and data loss caused by equipment failure and communication interruption, common methods include interpolation, filling in data before and after time intervals, or directly removing abnormal day data; Continuous time series data is divided into natural days (e.g., from 0:00 to 24:00) to form independent daily load curves. It is essential to ensure that all curves are on the same time base. In order to eliminate the differences in absolute power values ​​between different charging stations due to their scale and number of charging piles, and to make their curves comparable, it is necessary to normalize each daily load curve to obtain a standard load curve for each ultra-fast charging station.

[0047] 2-2. Calculate the similarity of the standard load curves of every two ultra-fast charging stations using the dynamic time warping algorithm. Construct a DTW (Dynamic Time Warping) distance matrix based on the similarity of the standard load curves of every two ultra-fast charging stations. Use a spectral clustering algorithm to treat each standard load curve as a node, and use the DTW distance matrix to define the connection strength between each pair of nodes to construct a similarity graph. 2-3. By calculating the Laplacian matrix eigenvector of the similarity map, the similarity map is mapped to a low-dimensional space. The k-means algorithm is used to cluster the standard load curves of multiple ultra-fast charging stations in the low-dimensional space to obtain clusters of various ultra-fast charging stations. 2-4. Based on the median of the standard load curve, the peak and valley values ​​of the load, the occurrence time of the peak and valley values, and the curve shape parameters in the cluster of each ultra-fast charging station, a specific load scenario model for each ultra-fast charging station is constructed.

[0048] Ultra-fast charging load curves exhibit significant scaling and shifts along the time axis (e.g., morning peaks may occur earlier or later on different dates), and traditional Euclidean distance cannot effectively measure this similarity. DTW can accurately capture similar patterns distorted along the time axis, while spectral clustering can discover cluster structures of arbitrary shapes. Combining the two can fundamentally identify load patterns such as "morning peak + evening peak" that are not strictly aligned in time but have essentially the same pattern, thus achieving more physically accurate classification.

[0049] In step S3, 3-1, the current operating parameters of each power supply unit and the corresponding general load scenario model are input into the load prediction model to generate the load prediction value of each power supply unit.

[0050] Specifically, based on each of the power supply units: The current operating parameters of the power supply unit are standardized and then concatenated to form the feature vector of the power supply unit; the current operating parameters include electricity consumption data, geographical location data, and time dimension data; A spatial graph is constructed by taking the power supply unit as a node and the geographical relationship between the power supply unit and multiple other power supply units in the distribution network as edges, and a time series is constructed based on the feature vector of the power supply unit for a certain period of time. The time series is input into an LSTM to generate the temporal coupling characteristics of the power supply unit, and the spatial graph is input into a graph neural network. After coupling, the spatiotemporal coupling characteristics of the power supply unit are obtained. The spatiotemporal coupling characteristics of the power supply unit, the category of the corresponding general load scenario model, and the typical load characteristics are input into the load prediction model to perform load prediction, thereby obtaining the load prediction value of the power supply unit.

[0051] 3-2. Input the current operating parameters of each ultra-fast charging station and the corresponding specific load scenario model into the load prediction model to generate the load prediction value of each ultra-fast charging station.

[0052] Specifically, based on each of the aforementioned ultra-fast charging stations: The current operating parameters of the ultra-fast charging station are standardized and then concatenated to form the feature vector of the ultra-fast charging station; the current operating parameters include electricity consumption data, geographical location data, and time dimension data; Using the ultra-fast charging station as a node, and the geographical relationship between the ultra-fast charging station and other ultra-fast charging stations in the power distribution network as edges, a spatial graph is constructed, and a time series is constructed based on the feature vector of the ultra-fast charging station for a certain period of time. The time series is input into an LSTM (Long Short-Term Memory) network to generate the temporal coupling characteristics of the ultrafast charging station, and the spatial graph is input into a graph neural network. After coupling, the spatiotemporal coupling characteristics of the ultrafast charging station are obtained. The spatiotemporal coupling characteristics of the ultrafast charging station, the category of the corresponding specific load scenario model, and typical load characteristics are input into the load prediction model to perform load prediction, thereby obtaining the load prediction value of the ultrafast charging station.

[0053] 3-3. Generate scheduling instructions based on the load forecast values ​​of the multiple power supply units and the multiple ultra-fast charging stations.

[0054] Based on reinforcement learning or genetic algorithms, we develop adaptive power allocation and scheduling strategies, enabling the system to dynamically adjust the power scheduling plans of each node according to the current load forecast results and the grid operating status.

[0055] Specifically, the load forecast values ​​of the multiple power supply units and the multiple ultra-fast charging stations, as well as the grid operating status, energy storage status, and scheduling objectives, are defined as the decision state space, and its expression is:

[0056] In the formula, Represents the decision state space. This represents the sum of the predicted load values ​​for the multiple power supply units and the multiple ultra-fast charging stations. Indicates the operating status of the power grid. Indicates the energy storage status. Indicates the scheduling target.

[0057] The power adjustment strategies of the multiple power supply units and the multiple ultra-fast charging stations are defined as the action space, and its expression is:

[0058] In the formula, Represents the action space; , , These represent the power adjustment amount for each unit, which includes the power supply unit and the ultra-fast charging station.

[0059] A reward function is constructed based on the load deviation, dispatching cost, and power supply stability of the distribution network, and its expression is as follows:

[0060] In the formula, , , All of these represent weighting coefficients, which can be set according to actual needs. This represents the instantaneous reward value at time t.

[0061] Based on real-time information in the decision state space, a reinforcement learning algorithm, guided by a reward function, searches for the optimal policy in the action space and outputs scheduling instructions for each power supply unit and each ultra-fast charging station. The reinforcement learning algorithm includes deep Q-networks or near-end policy optimization algorithms.

[0062] The operation of each power supply unit and the ultra-fast charging station is adjusted according to the scheduling instructions.

[0063] In summary, this application constructs different types of general load scenario models by analyzing the land use and load data of each power supply unit, and performs differentiated modeling of the load curves of different ultra-fast charging stations. This enables accurate load prediction analysis for different units (power supply units and ultra-fast charging stations). Inputting the general load scenario model of the power supply unit (including scenario model categories and typical load characteristics) and the specific load scenario model of the ultra-fast charging station into the prediction model provides strong physical constraints and prior knowledge, improving prediction accuracy and providing an accurate data foundation for subsequent scheduling analysis. At the same time, it can avoid power grid safety problems caused by improper scheduling.

[0064] After generating scheduling instructions based on the load forecast values ​​of each power supply unit and the ultra-fast charging station, this solution also includes: According to a preset time interval, under the premise of satisfying preset constraints, the objective function of the distribution network optimization scheduling is solved by a genetic algorithm to obtain the global optimal scheduling scheme of the distribution network. After adjusting the operation of each power supply unit and the ultra-fast charging station according to the scheduling instructions, the operation of each power supply unit and the ultra-fast charging station is adjusted again using the global scheduling optimal scheme; The constraints include one or more of the following: power balance constraints, energy storage capacity constraints, and equipment operation constraints. The objective function is constructed to minimize the total scheduling cost and maximize power supply stability. The expression of the objective function is:

[0065] In the formula, This represents the scheduling cost coefficient of the i-th unit. The scheduling variable for the i-th unit, This indicates the number of units in the power distribution network, including power supply units and ultra-fast charging stations. Indicates the weighting coefficient. This indicates the loss of power supply stability in the distribution network.

[0066] Furthermore, by deploying distributed energy storage devices at key nodes (ultra-fast charging stations), the sudden surge in power demand at these stations can severely impact the local power grid, leading to overload or voltage instability. Microgrids and energy storage provide local balancing and rapid support, instantly absorbing the impact and ensuring power supply security. Coordinated control through an intelligent management system enables a smooth load curve design, allowing the system to release stored energy during peak loads and charge during off-peak periods. This proactive adjustment and peak shaving of the load curve achieves the technical effects of improving grid stability, reducing operating costs, and extending equipment lifespan. This design enhances the flexibility of the distribution network and the utilization efficiency of energy storage resources, strengthening the system's adaptability to highly fluctuating loads.

[0067] After generating scheduling instructions based on the load forecast values ​​of each power supply unit and the ultra-fast charging station, the method further includes: Obtain a multidimensional sensing dataset of the power distribution network, which includes electrical quantity data, complaint data, and external factor data; The load factor is calculated based on the electrical quantity data of each area in the distribution network. Using a color mapping function, a color patch layer reflecting the distribution pattern is generated based on the load factor. The expression is as follows:

[0068] in, It is a color mapping function used to convert load rate into a corresponding color. The color indicates the load factor. The load factor for each zone is derived from the ratio of the actual load power to the rated power in the electrical quantity.

[0069] A scatter plot layer reflecting the distribution of complaints is generated based on the location of the complaint data in the power distribution network. The color patch map layer and scatter plot layer are overlaid on the GIS platform to visualize the power grid operation status.

[0070] By utilizing a GIS (Geographic Information System) platform to display color-coded maps and scatter plots, the system visually presents the regional load distribution and complaint hotspots. This enables the system to visualize the power grid's operating status and user feedback, achieving rapid identification of load distribution characteristics and problem areas. This results in technical effects that assist in operation and maintenance decision-making and improve service quality.

[0071] After generating scheduling instructions based on the load forecast values ​​of each power supply unit and the ultra-fast charging station, the method further includes: Acquire historical power grid operation data, and extract basic indicator features from the historical power grid operation data. The basic indicator features include time-series features, statistical features, and frequency domain features. Fault features are extracted from the basic index features using a convolutional neural network, and the fault features are input into a random forest for state classification to obtain the fault category; The initial risk prediction model is trained based on historical power grid operation data containing fault categories to obtain a risk prediction model. The risk prediction model is then used to predict risks based on current power grid operation data and output the risk level. When the risk level reaches the risk threshold, the alarm mechanism of the distribution network is triggered, that is, when the model output is a warning or a fault, the alarm mechanism is triggered; at the same time, information such as the fault type, the location of occurrence, and the possible scope of impact is recorded.

[0072] Example 2 Based on the same inventive concept, this invention also provides a power distribution network load characteristic analysis and management system including an ultra-fast charging station, such as... Figure 2 As shown, it includes: A general load scenario construction module is used to cluster and analyze multiple power supply units by utilizing historical load data and land use characteristics of each power supply unit in the distribution network, and obtain a general load scenario model for each power supply unit; some of the power supply units contain ultra-fast charging stations; The specific load scenario construction module is used to perform load curve similarity analysis and clustering on multiple ultra-fast charging stations based on the historical load data of each of the ultra-fast charging stations, using dynamic time warping algorithm and spectral clustering algorithm, and to perform typical load characteristic analysis on each type of ultra-fast charging station obtained by clustering, so as to generate a specific load scenario model for each type of ultra-fast charging station. The scheduling instruction generation module is used to input the current operating parameters of each power supply unit and the corresponding general load scenario model into the load prediction model to generate the load prediction value of each power supply unit, and to input the current operating parameters of each ultra-fast charging station and the corresponding specific load scenario model into the load prediction model to generate the load prediction value of each ultra-fast charging station, and to generate scheduling instructions based on the load prediction values ​​of multiple power supply units and multiple ultra-fast charging stations.

[0073] In one possible implementation, the aforementioned general load scenario construction module is specifically used for: Load pattern characteristics and time sequence characteristics are extracted from the historical load data of each power supply unit, and the land use nature of each power supply unit is converted into numerical characteristics; Based on each power supply unit, the multi-source characteristics of each power supply unit are obtained by integrating the load pattern characteristics, timing characteristics and numerical characteristics of the power supply unit; Based on the multi-source characteristics of each power supply unit, the k-means clustering algorithm combined with the DBSCAN algorithm is used to cluster multiple power supply units to obtain multiple power supply unit clusters; Based on the average load curve and load characteristic parameters in the cluster of each power supply unit, a general load scenario model for each power supply unit is constructed.

[0074] In one possible implementation, the aforementioned specific load scenario construction module is specifically used for: A standard load curve for each ultra-fast charging station is generated based on the historical load data of each ultra-fast charging station. The similarity of the standard load curves of every two ultra-fast charging stations is calculated using the dynamic time warping algorithm, and a DTW distance matrix is ​​constructed based on the similarity of the standard load curves of every two ultra-fast charging stations. The standard load curve is treated as a node using a spectral clustering algorithm, and the connection strength between each pair of nodes is defined using the DTW distance matrix to construct a similarity graph. The similarity map is mapped to a low-dimensional space by calculating the Laplacian matrix eigenvector of the similarity map. The k-means algorithm is then used to cluster the standard load curves of multiple ultra-fast charging stations in the low-dimensional space to obtain clusters of various ultra-fast charging stations. Based on the median of the standard load curve, the peak and valley values, the occurrence time of the peak and valley values, and the curve shape parameters in the cluster of each ultra-fast charging station, a specific load scenario model for each ultra-fast charging station is constructed.

[0075] In one possible implementation, the aforementioned scheduling instruction generation module is specifically used for: Based on each of the power supply units: The current operating parameters of the power supply unit are standardized and then concatenated to form the feature vector of the power supply unit; the current operating parameters include electricity consumption data, geographical location data, and time dimension data; A spatial graph is constructed by taking the power supply unit as a node and the geographical relationship between the power supply unit and multiple other power supply units in the distribution network as edges, and a time series is constructed based on the feature vector of the power supply unit for a certain period of time. The time series is input into an LSTM to generate the temporal coupling characteristics of the power supply unit, and the spatial graph is input into a graph neural network. After coupling, the spatiotemporal coupling characteristics of the power supply unit are obtained. The spatiotemporal coupling characteristics of the power supply unit, the category of the corresponding general load scenario model, and the typical load characteristics are input into the load prediction model to perform load prediction, thereby obtaining the load prediction value of the power supply unit.

[0076] In one possible implementation, the aforementioned scheduling instruction generation module is specifically used for: The load forecast values ​​of the multiple power supply units and the multiple ultra-fast charging stations, as well as the grid operation status, energy storage status and scheduling objectives are defined as the decision state space; The power adjustment strategies of the multiple power supply units and the multiple ultra-fast charging stations are defined as the action space; A reward function is constructed based on the load deviation, dispatching cost, and power supply stability of the aforementioned distribution network; Based on the real-time information in the decision state space, the reinforcement learning algorithm is driven to search for the optimal strategy in the action space under the guidance of the reward function, and outputs the scheduling instructions for each power supply unit and each ultra-fast charging station.

[0077] In one possible implementation, the system further includes a global optimization module for: According to a preset time interval, under the premise of satisfying preset constraints, the objective function of the distribution network optimization scheduling is solved by a genetic algorithm to obtain the global optimal scheduling scheme of the distribution network. After adjusting the operation of each power supply unit and the ultra-fast charging station according to the scheduling instructions, the operation of each power supply unit and the ultra-fast charging station is adjusted again using the global scheduling optimal scheme; The constraints include one or more of the following: power balance constraints, energy storage capacity constraints, and equipment operation constraints. The objective function is constructed to minimize the total scheduling cost and maximize power supply stability. The expression of the objective function is:

[0078] In the formula, This represents the scheduling cost coefficient of the i-th unit. The scheduling variable for the i-th unit, This indicates the number of units in the power distribution network, including power supply units and ultra-fast charging stations. Indicates the weighting coefficient. This indicates the loss of power supply stability in the distribution network.

[0079] In one possible implementation, the system further includes a visualization module for: Obtain a multidimensional sensing dataset of the power distribution network, which includes electrical quantity data, complaint data, and external factor data; The load factor is calculated based on the electrical quantity data of each area in the distribution network, and a color map layer reflecting the distribution is generated based on the load factor using a color mapping function. A scatter plot layer reflecting the distribution of complaints is generated based on the location of the complaint data in the power distribution network. The color patch map layer and scatter plot layer are overlaid on the GIS platform to visualize the power grid operation status.

[0080] In one possible implementation, the system further includes a risk warning module, used for: Acquire historical power grid operation data, and extract basic indicator features from the historical power grid operation data. The basic indicator features include time-series features, statistical features, and frequency domain features. Fault features are extracted from the basic index features using a convolutional neural network, and the fault features are input into a random forest for state classification to obtain the fault category; The initial risk prediction model is trained based on historical power grid operation data containing fault categories to obtain a risk prediction model. The risk prediction model is then used to predict risks based on current power grid operation data and output the risk level. The alarm mechanism of the power distribution network is triggered when the risk level reaches the risk threshold.

[0081] Example 3 like Figure 3 As shown, the present invention also provides an electronic device, which may be a computer device, a microcontroller device, a smart mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, processor, and transceiver component are connected via a bus; the memory can be used to store executable programs, and an exemplary executable program may include instructions; the processor is used to execute the instructions stored in the memory. The memory can also be used to store data, which can be accessed and / or modified when instructions are executed.

[0082] The processor may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, and it is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the storage medium to realize the corresponding method flow or corresponding function, so as to realize the steps of the distribution network load characteristic analysis and management method with ultra-fast charging station in the above embodiment.

[0083] Example 4 Based on the same inventive concept, this invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory). This readable storage medium is a memory device within an electronic device used to store programs and data. It is understood that the storage medium here can include both built-in storage media within the electronic device and extended storage media supported by the electronic device. The storage medium provides storage space, which stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more executable programs (including program code). It should be noted that the storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. Loading and executing one or more instructions stored in the storage medium by the processor can implement the steps of the distribution network load characteristic analysis and management method with an ultra-fast charging station described in the above embodiments.

[0084] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0085] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0086] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0087] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit its scope of protection. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading the present invention, they can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the application, but these changes, modifications or equivalent substitutions are all within the scope of protection of the claims pending approval.

Claims

1. A method for analyzing and managing the load characteristics of a distribution network including an ultra-fast charging station, characterized in that, include: By utilizing the historical load data and land use characteristics of each power supply unit in the distribution network, clustering and load analysis are performed on multiple power supply units to obtain a general load scenario model for each type of power supply unit; some of the power supply units contain ultra-fast charging stations; Based on the historical load data of each of the ultra-fast charging stations, the load curve similarity analysis and clustering of multiple ultra-fast charging stations are performed using dynamic time warping algorithm and spectral clustering algorithm. Then, the typical load characteristics of each type of ultra-fast charging station obtained by clustering are analyzed to generate a specific load scenario model for each type of ultra-fast charging station. The current operating parameters of each power supply unit and the corresponding general load scenario model are input into the load prediction model to generate the load prediction value of each power supply unit. The current operating parameters of each ultra-fast charging station and the corresponding specific load scenario model are input into the load prediction model to generate the load prediction value of each ultra-fast charging station. The scheduling instructions are generated based on the analysis of the load prediction values ​​of multiple power supply units and multiple ultra-fast charging stations.

2. The method as described in claim 1, characterized in that, The method involves using historical load data and land use characteristics of each power supply unit in the distribution network to perform clustering and load analysis on multiple power supply units, resulting in a general load scenario model for each type of power supply unit, including: Load pattern characteristics and time sequence characteristics are extracted from the historical load data of each power supply unit, and the land use nature of each power supply unit is converted into numerical characteristics; Based on each power supply unit, the multi-source characteristics of each power supply unit are obtained by integrating the load pattern characteristics, timing characteristics and numerical characteristics of the power supply unit; Based on the multi-source characteristics of each power supply unit, the k-means clustering algorithm combined with the DBSCAN algorithm is used to cluster multiple power supply units to obtain multiple power supply unit clusters; Based on the average load curve and load characteristic parameters in the cluster of each power supply unit, a general load scenario model for each power supply unit is constructed.

3. The method as described in claim 1, characterized in that, Based on the historical load data of each of the ultra-fast charging stations, the system utilizes dynamic time warping and spectral clustering algorithms to perform load curve similarity analysis and clustering on multiple ultra-fast charging stations. Furthermore, it analyzes typical load characteristics for each type of ultra-fast charging station obtained from the clustering, generating a specific load scenario model for each type of ultra-fast charging station, including: A standard load curve for each ultra-fast charging station is generated based on the historical load data of each ultra-fast charging station. The similarity of the standard load curves of every two ultra-fast charging stations is calculated using the dynamic time warping algorithm, and a DTW distance matrix is ​​constructed based on the similarity of the standard load curves of every two ultra-fast charging stations. The standard load curve is treated as a node using a spectral clustering algorithm, and the connection strength between each pair of nodes is defined using the DTW distance matrix to construct a similarity graph. The similarity map is mapped to a low-dimensional space by calculating the Laplacian matrix eigenvector of the similarity map. The k-means algorithm is then used to cluster the standard load curves of multiple ultra-fast charging stations in the low-dimensional space to obtain clusters of various ultra-fast charging stations. Based on the median of the standard load curve, the peak and valley values, the occurrence time of the peak and valley values, and the curve shape parameters in the cluster of each ultra-fast charging station, a specific load scenario model for each ultra-fast charging station is constructed.

4. The method as described in claim 1, characterized in that, The step of inputting the current operating parameters of each power supply unit and the corresponding general load scenario model into the load prediction model to generate the load prediction value for each power supply unit includes: Based on each of the power supply units: The current operating parameters of the power supply unit are standardized and then concatenated to form the feature vector of the power supply unit; the current operating parameters include electricity consumption data, geographical location data, and time dimension data; A spatial graph is constructed by taking the power supply unit as a node and the geographical relationship between the power supply unit and multiple other power supply units in the distribution network as edges, and a time series is constructed based on the feature vector of the power supply unit for a certain period of time. The time series is input into an LSTM to generate the temporal coupling characteristics of the power supply unit, and the spatial graph is input into a graph neural network. After coupling, the spatiotemporal coupling characteristics of the power supply unit are obtained. The spatiotemporal coupling characteristics of the power supply unit, the category of the corresponding general load scenario model, and the typical load characteristics are input into the load prediction model to perform load prediction, thereby obtaining the load prediction value of the power supply unit.

5. The method as described in claim 1, characterized in that, The step of generating scheduling instructions based on load forecast values ​​of multiple power supply units and multiple ultra-fast charging stations includes: The load forecast values ​​of the multiple power supply units and the multiple ultra-fast charging stations, as well as the grid operation status, energy storage status and scheduling objectives are defined as the decision state space; The power adjustment strategies of the multiple power supply units and the multiple ultra-fast charging stations are defined as the action space; A reward function is constructed based on the load deviation, dispatching cost, and power supply stability of the aforementioned distribution network; Based on the real-time information in the decision state space, the reinforcement learning algorithm is driven to search for the optimal strategy in the action space under the guidance of the reward function, and outputs the scheduling instructions for each power supply unit and each ultra-fast charging station.

6. The method as described in claim 1, characterized in that, After generating scheduling instructions based on the load forecast values ​​of each power supply unit and the ultra-fast charging station, the process further includes: According to a preset time interval, under the premise of satisfying preset constraints, the objective function of the distribution network optimization scheduling is solved by a genetic algorithm to obtain the global optimal scheduling scheme of the distribution network. After adjusting the operation of each power supply unit and the ultra-fast charging station according to the scheduling instructions, the operation of each power supply unit and the ultra-fast charging station is adjusted again using the global scheduling optimal scheme; The constraints include one or more of the following: power balance constraints, energy storage capacity constraints, and equipment operation constraints. The objective function is constructed to minimize the total scheduling cost and maximize power supply stability. The expression of the objective function is: In the formula, This represents the scheduling cost coefficient of the i-th unit. The scheduling variable for the i-th unit, This indicates the number of units in the power distribution network, including power supply units and ultra-fast charging stations. Indicates the weighting coefficient. This indicates the loss of power supply stability in the distribution network.

7. The method as described in claim 1 or 2, characterized in that, After generating scheduling instructions based on the load forecast values ​​of each power supply unit and the ultra-fast charging station, the process further includes: Obtain a multidimensional sensing dataset of the power distribution network, which includes electrical quantity data, complaint data, and external factor data; The load factor is calculated based on the electrical quantity data of each area in the distribution network, and a color map layer reflecting the distribution is generated based on the load factor using a color mapping function. A scatter plot layer reflecting the distribution of complaints is generated based on the location of the complaint data in the power distribution network. The color patch map layer and scatter plot layer are overlaid on the GIS platform to visualize the power grid operation status.

8. The method as described in claim 1 or 2, characterized in that, After generating scheduling instructions based on the load forecast values ​​of each power supply unit and the ultra-fast charging station, the process further includes: Acquire historical power grid operation data, and extract basic indicator features from the historical power grid operation data. The basic indicator features include time-series features, statistical features, and frequency domain features. Fault features are extracted from the basic index features using a convolutional neural network, and the fault features are input into a random forest for state classification to obtain the fault category; The initial risk prediction model is trained based on historical power grid operation data containing fault categories to obtain a risk prediction model. The risk prediction model is then used to predict risks based on current power grid operation data and output the risk level. The alarm mechanism of the power distribution network is triggered when the risk level reaches the risk threshold.

9. A distribution network load characteristic analysis and management system including an ultra-fast charging station, characterized in that, include: A general load scenario construction module is used to cluster and analyze multiple power supply units by utilizing historical load data and land use characteristics of each power supply unit in the distribution network, and obtain a general load scenario model for each power supply unit; some of the power supply units contain ultra-fast charging stations; The specific load scenario construction module is used to perform load curve similarity analysis and clustering on multiple ultra-fast charging stations based on the historical load data of each of the ultra-fast charging stations, using dynamic time warping algorithm and spectral clustering algorithm, and to perform typical load characteristic analysis on each type of ultra-fast charging station obtained by clustering, so as to generate a specific load scenario model for each type of ultra-fast charging station. The scheduling instruction generation module is used to input the current operating parameters of each power supply unit and the corresponding general load scenario model into the load prediction model to generate the load prediction value of each power supply unit, and to input the current operating parameters of each ultra-fast charging station and the corresponding specific load scenario model into the load prediction model to generate the load prediction value of each ultra-fast charging station, and to generate scheduling instructions based on the load prediction values ​​of multiple power supply units and multiple ultra-fast charging stations.

10. The system as described in claim 9, characterized in that, The general load scenario construction module is specifically used for: Load pattern characteristics and time sequence characteristics are extracted from the historical load data of each power supply unit, and the land use nature of each power supply unit is converted into numerical characteristics; Based on each power supply unit, the multi-source characteristics of each power supply unit are obtained by integrating the load pattern characteristics, timing characteristics and numerical characteristics of the power supply unit; Based on the multi-source characteristics of each power supply unit, the k-means clustering algorithm combined with the DBSCAN algorithm is used to cluster multiple power supply units to obtain multiple power supply unit clusters; Based on the average load curve and load characteristic parameters in the cluster of each power supply unit, a general load scenario model for each power supply unit is constructed.

11. The system as described in claim 9, characterized in that, The specific load scenario construction module is specifically used for: A standard load curve for each ultra-fast charging station is generated based on the historical load data of each ultra-fast charging station. The similarity of the standard load curves of every two ultra-fast charging stations is calculated using the dynamic time warping algorithm, and a DTW distance matrix is ​​constructed based on the similarity of the standard load curves of every two ultra-fast charging stations. The standard load curve is treated as a node using a spectral clustering algorithm, and the connection strength between each pair of nodes is defined using the DTW distance matrix to construct a similarity graph. The similarity map is mapped to a low-dimensional space by calculating the Laplacian matrix eigenvector of the similarity map. The k-means algorithm is then used to cluster the standard load curves of multiple ultra-fast charging stations in the low-dimensional space to obtain clusters of various ultra-fast charging stations. Based on the median of the standard load curve, the peak and valley values, the occurrence time of the peak and valley values, and the curve shape parameters in the cluster of each ultra-fast charging station, a specific load scenario model for each ultra-fast charging station is constructed.

12. The system as described in claim 9, characterized in that, The scheduling instruction generation module is specifically used for: Based on each of the power supply units: The current operating parameters of the power supply unit are standardized and then concatenated to form the feature vector of the power supply unit; the current operating parameters include electricity consumption data, geographical location data, and time dimension data; A spatial graph is constructed by taking the power supply unit as a node and the geographical relationship between the power supply unit and multiple other power supply units in the distribution network as edges, and a time series is constructed based on the feature vector of the power supply unit for a certain period of time. The time series is input into an LSTM to generate the temporal coupling characteristics of the power supply unit, and the spatial graph is input into a graph neural network. After coupling, the spatiotemporal coupling characteristics of the power supply unit are obtained. The spatiotemporal coupling characteristics of the power supply unit, the category of the corresponding general load scenario model, and the typical load characteristics are input into the load prediction model to perform load prediction, thereby obtaining the load prediction value of the power supply unit.

13. The system as described in claim 9, characterized in that, The scheduling instruction generation module is specifically used for: The load forecast values ​​of the multiple power supply units and the multiple ultra-fast charging stations, as well as the grid operation status, energy storage status and scheduling objectives are defined as the decision state space; The power adjustment strategies of the multiple power supply units and the multiple ultra-fast charging stations are defined as the action space; A reward function is constructed based on the load deviation, dispatching cost, and power supply stability of the aforementioned distribution network; Based on the real-time information in the decision state space, the reinforcement learning algorithm is driven to search for the optimal strategy in the action space under the guidance of the reward function, and outputs the scheduling instructions for each power supply unit and each ultra-fast charging station.

14. The system as described in claim 9, characterized in that, The system also includes a global optimization module, used for: According to a preset time interval, under the premise of satisfying preset constraints, the objective function of the distribution network optimization scheduling is solved by a genetic algorithm to obtain the global optimal scheduling scheme of the distribution network. After adjusting the operation of each power supply unit and the ultra-fast charging station according to the scheduling instructions, the operation of each power supply unit and the ultra-fast charging station is adjusted again using the global scheduling optimal scheme; The constraints include one or more of the following: power balance constraints, energy storage capacity constraints, and equipment operation constraints. The objective function is constructed to minimize the total scheduling cost and maximize power supply stability. The expression of the objective function is: In the formula, This represents the scheduling cost coefficient of the i-th unit. The scheduling variable for the i-th unit, This indicates the number of units in the power distribution network, including power supply units and ultra-fast charging stations. Indicates the weighting coefficient. This indicates the loss of power supply stability in the distribution network.

15. The system as described in claim 9, characterized in that, The system also includes a visualization module for: Obtain a multidimensional sensing dataset of the power distribution network, which includes electrical quantity data, complaint data, and external factor data; The load factor is calculated based on the electrical quantity data of each area in the distribution network, and a color map layer reflecting the distribution is generated based on the load factor using a color mapping function. A scatter plot layer reflecting the distribution of complaints is generated based on the location of the complaint data in the power distribution network. The color patch map layer and scatter plot layer are overlaid on the GIS platform to visualize the power grid operation status.

16. The system as described in claim 9, characterized in that, The system also includes a risk warning module, used for: Acquire historical power grid operation data, and extract basic indicator features from the historical power grid operation data. The basic indicator features include time-series features, statistical features, and frequency domain features. Fault features are extracted from the basic index features using a convolutional neural network, and the fault features are input into a random forest for state classification to obtain the fault category; The initial risk prediction model is trained based on historical power grid operation data containing fault categories to obtain a risk prediction model. The risk prediction model is then used to predict risks based on current power grid operation data and output the risk level. The alarm mechanism of the power distribution network is triggered when the risk level reaches the risk threshold.

17. An electronic device, characterized in that, include: At least one processor and memory; The memory and processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, a method for analyzing and managing the load characteristics of a distribution network including an ultra-fast charging station as described in any one of claims 1 to 8 is implemented.

18. A computing device readable storage medium, characterized in that, It contains an execution program, which, when executed, implements a method for analyzing and managing the load characteristics of a distribution network including an ultra-fast charging station as described in any one of claims 1 to 8.

Images