Source network load optimization scheduling analysis method and system for active power distribution network of electric vehicle

Abstract

The invention relates to the technical field of scheduling management, in particular to a source network load optimization scheduling analysis method and system for an active power distribution network of an electric vehicle. The method comprises the following steps: acquiring power supply side data, power grid side data and load side data in real time, and performing sliding filtering and standardization processing to obtain source network load standard data of the electric vehicle; historical charging data, traffic flow data, weather data and user appearing habit data are obtained, and load and power output prediction calculation is carried out to predict and output electric vehicle charging power distribution and distributed power generation power distribution; a corresponding multi-objective optimization function and a power distribution network operation constraint condition are constructed through an electric vehicle active power distribution network, a source network load optimization scheduling model is constructed, and a source network load optimization scheduling scheme is solved to adjust corresponding electric vehicle charging power and distributed power generation power in real time. The energy utilization efficiency and the power grid operation stability can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of dispatching management, and in particular to a source-grid-load optimization dispatching analysis method and system for an active distribution network for electric vehicles. Background Art

[0002] With the rapid increase in the number of electric vehicles (EVs), the impact of EV charging demand on the load of the power grid is becoming increasingly significant, especially in cities and high-density areas. The charging load of EVs will cause the power supply capacity of the distribution network to be overloaded, which in turn affects the safety and stability of the power grid. Therefore, how to effectively coordinate the charging demand of EVs with the power supply capacity of the distribution network to avoid overload and power waste in the power grid has become an important issue that needs to be solved in the power system. At present, most EV charging management methods adopt a passive load scheduling strategy, usually adjusting the charging time of EVs through simple peak-valley electricity prices or time division to reduce the impact on the power grid. However, this method mainly focuses on the regulation of traditional loads (such as industrial, commercial and residential loads), ignores the interactive relationship between EVs and distribution networks, and cannot achieve the coordinated optimization scheduling of source-grid-load (power source, distribution network and load), and it is difficult to meet the flexibility of EVs in charging time, thereby reducing the efficiency of coordinated scheduling between source-grid-load. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide a source-grid-load optimization scheduling analysis method and system for an active distribution network for electric vehicles to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a source-grid-load optimization scheduling analysis method for an electric vehicle active distribution network comprises the following steps: Step S1: by real-time collection of power supply side data, grid side data and load side data in the operation of the electric vehicle active distribution network, and sliding filtering and standardization processing of the power supply side data, grid side data and load side data, the electric vehicle source grid load standard data is obtained; Step S2: Obtain historical charging data, traffic flow data, weather data, and user habit data, and build an electric vehicle load prediction model in combination with a gated cycle unit to perform load prediction calculations on the electric vehicle source grid load standard data, so as to predict and output the electric vehicle charging power distribution corresponding to different time periods and different regions; obtain historical power generation data and numerical weather forecast data, and build a distributed power hybrid prediction model to perform power output prediction calculations on the electric vehicle source grid load standard data, so as to predict and output the distributed power generation power distribution corresponding to different time periods and different regions; Step S3: Construct a corresponding multi-objective optimization function and distribution network operation constraints through the electric vehicle active distribution network, and construct a source-network-load optimization scheduling model for the multi-objective optimization function based on the distribution network operation constraints to generate a source-network-load collaborative optimization scheduling model; Step S4: Set a corresponding source-network-load scheduling period, and read corresponding prediction data from the electric vehicle charging power distribution and distributed power generation power distribution corresponding to different time periods and regions based on the source-network-load scheduling period; input the corresponding prediction data into the source-network-load collaborative optimization scheduling model to solve for the corresponding source-network-load optimization scheduling plan, and adjust the corresponding electric vehicle charging power and distributed power generation power in real time according to the source-network-load optimization scheduling plan.

[0005] Further, step S1 includes the following steps: Step S11: Real-time collect the power supply side data during the operation of the electric vehicle active distribution network, including the power generation power, remaining capacity, and real-time operation status data of the distributed power source; Step S12: Real-time collect the grid side data during the operation of the electric vehicle active distribution network, including the voltage of each node of the distribution network, line current, network topology, and transformer capacity limit operation parameters; Step S13: Real-time collect the load side data during the operation of the electric vehicle active distribution network, including the real-time power demand of traditional fixed loads, electric vehicle locations, remaining battery power, estimated driving time, and charging demand data; Step S14: Set the sliding window size at 3 - 5 time points, and perform sliding filtering processing on the power supply side data, grid side data, and load side data based on the sliding window size to obtain the electric vehicle source-network-load sliding filtered data; Step S15: Perform interpolation filling and standardization processing on the electric vehicle source-network-load sliding filtered data to fill the corresponding missing data using median interpolation and eliminate the dimension between different format data to obtain the electric vehicle source-network-load standard data.

[0006] Further, step S2 includes the following steps: Step S21: Obtain historical charging data, specifically the electric vehicle charging records within the past 1 - 2 years, including charging time, charging duration, and charging power; Step S22: Obtain traffic flow data, including the number of vehicle flows of electric vehicles at different time periods and regions; Step S23: Obtain weather data and user travel habit data, where the weather data includes temperature, humidity, and light, and the user travel habit data includes daily travel time, routes, and destinations collected through questionnaires or mobile applications; Step S24: Based on historical charging data, traffic flow data, weather data, and user appearance habit data, and combined with a gated recurrent unit, construct an electric vehicle load prediction model to perform load prediction calculations on the electric vehicle source-network-load standard data, so as to predict and output the electric vehicle charging power distribution corresponding to different time periods and different regions; Step S25: Obtain historical power generation data and numerical weather forecast data, and construct a distributed power hybrid prediction model to perform power output prediction calculations on the electric vehicle source-network-load standard data, so as to predict and output the distributed power generation power distribution corresponding to different time periods and different regions.

[0007] Furthermore, step S24 includes the following steps: Step S241: Obtain the charging time characteristics of electric vehicles corresponding to hours, days, weeks, months, and seasons through historical charging data; Step S242: Obtain the vehicle traffic characteristics of electric vehicles in different regions through traffic flow data, including vehicle traffic density and vehicle congestion index; Step S243: Obtain the weather influence characteristics corresponding to electric vehicles through weather data, including the change rates corresponding to temperature, humidity, and light intensity; Step S244: Obtain the corresponding user travel characteristics through user travel habit data, including travel distance, travel purpose, and charging frequency; obtain the potential influence characteristics of the source-network-load state on electric vehicle charging through the analysis of electric vehicle source-network-load standard data; Step S245: Construct an electric vehicle load prediction model by selecting a gated recurrent unit (GRU), and use the charging time characteristics, vehicle traffic characteristics, weather influence characteristics, user travel characteristics, and the potential influence characteristics of the source-network-load state on electric vehicle charging as training data to train the electric vehicle load prediction model. By adjusting the hyperparameters corresponding to the model, including the number of neurons in the hidden layer, learning rate, and number of iterations, and minimizing the prediction error, at the same time, input the electric vehicle source-network-load standard data into the trained electric vehicle load prediction model for load prediction calculations, so as to predict and output the electric vehicle charging power distribution corresponding to different time periods and different regions.

[0008] Furthermore, step S25 includes the following steps: Step S251: Obtain historical power generation data, specifically the historical power generation of distributed power sources and the corresponding timestamp information, where the distributed power sources include photovoltaic and wind power; Step S252: Obtain numerical weather forecast data, including solar radiation intensity, wind speed, wind direction, and temperature meteorological parameters; Step S253: Obtain the output time characteristics of the distributed power source corresponding to hours, days, weeks, months, and seasons through historical power generation data; Step S254: Obtain the influence characteristics of solar radiation intensity, wind speed, wind direction, and temperature on the power generation capacity corresponding to the output of the distributed power source through numerical weather forecast data; obtain the potential influence characteristics of the source-network-load state on the output of the distributed power source through the analysis of electric vehicle source-network-load standard data; Step S255: Construct a hybrid prediction model for the distributed power source by selecting the corresponding deep learning algorithm. For photovoltaic power, a hybrid prediction model combining a convolutional neural network and a long short-term memory network is used, where the convolutional neural network is used to extract the spatial characteristics in the meteorological data, and the long short-term memory network is used to capture the time series characteristics corresponding to the power generation power; for wind power, the time series corresponding to the meteorological data is decomposed based on wavelet transform to extract the characteristics of different frequency components, and a hybrid prediction model is constructed in combination with a support vector machine. At the same time, the output time characteristics, power generation capacity influence characteristics, and potential influence characteristics of the source-network-load state on the output of the distributed power source are used as training data to train the hybrid prediction model for the distributed power source. By adjusting the hyperparameters corresponding to the model, including the convolutional kernel size, stride, number of hidden layer neurons, and penalty parameter, and minimizing the prediction error, and inputting the electric vehicle source-network-load standard data into the trained hybrid prediction model for the distributed power source to perform power output prediction calculations, so as to predict and output the distributed power generation power distribution corresponding to different time periods and different regions.

[0009] Further, step S3 includes the following steps: Step S31: Construct a corresponding multi-objective optimization function through the electric vehicle active distribution network, including economic objectives, environmental protection objectives, and grid security objectives; Step S32: Consider the physical constraints corresponding to the operation of the distribution network and the operation characteristics constraints of electric vehicles through the electric vehicle active distribution network to construct corresponding distribution network operation constraint conditions, including grid operation constraints, distributed power source constraints, and electric vehicle constraints; Step S33: Based on the distribution network operation constraint conditions, add the corresponding grid operation constraints, distributed power source constraints, and electric vehicle constraints one by one to construct a source-network-load optimization scheduling model for the multi-objective optimization function, so as to generate a source-network-load collaborative optimization scheduling model.

[0010] Further, the economic objective described in step S31 is specifically to minimize the operation cost of the distribution network, including power purchase cost, power generation cost corresponding to the distributed power source, network loss cost, and charging compensation cost corresponding to the electric vehicle: ; Among them, represents the power purchase cost, represents the total number of distributed power sources, represents the power generation cost corresponding to the th distributed power source, represents the network loss cost, represents the total number of electric vehicles, represents the charging compensation cost corresponding to the th electric vehicle; ; Among them, represents the rd power generation power corresponding to the distributed power source, represents the total number of traditional thermal loads, represents the th traditional thermal load corresponding to the power generation power; The grid security objective described above is specifically to minimize the node voltage deviation and line overload risk: ; Among them, represents the total number of nodes in the electric vehicle active distribution network, represents the th node corresponding to the voltage deviation, represents the total number of lines in the electric vehicle active distribution network, represents the th line corresponding to the current, represents the th line corresponding to the rated current.

[0011] Furthermore, the grid operation constraints described in step S32 include: Node voltage amplitude constraint: , where and are respectively the lower limit value and the upper limit value of the voltage deviation corresponding to the th node; Line current upper limit constraint: , to ensure that the line current does not exceed its rated current; Active power balance constraint: , where represents the distributed power source set, represents the th distributed power source corresponding to the active power, represents the traditional thermal load set, represents the th traditional thermal load corresponding to the active power, Denote the set of electric vehicles, Denote the active power corresponding to the th electric vehicle; Reactive power balance constraint: , where Denote the reactive power corresponding to the th distributed power source; Denote the reactive power corresponding to the th traditional thermal load; Denote the reactive power corresponding to the th electric vehicle; Denote the and reactive power loss of the network; The distributed power source constraints described above include: Generation power upper and lower limit constraints: , where Denote the time parameter, Denote the maximum ramp-up generation rate corresponding to the th distributed power source; The electric vehicle constraints described above include: Active power upper and lower limit constraints: and Denote the active minimum power and active maximum power corresponding to the th electric vehicle; Reactive power upper and lower limit constraints: and Denote the reactive minimum power and reactive maximum power corresponding to the

[0012] Furthermore, step S4 includes the following steps: Step S41: Set the corresponding source-network-load scheduling period, specifically 15 minutes for this source-network-load scheduling period; Step S42: Read the prediction data corresponding to each source-network-load scheduling period from the electric vehicle charging power distribution and distributed power source generation power distribution corresponding to different time periods and different regions based on the source-network-load scheduling period, including the charging power distribution and generation power distribution within the corresponding source-network-load scheduling period; Step S43: Input the corresponding predicted data into the source-grid-load collaborative optimization scheduling model to solve and obtain the corresponding source-grid-load optimized scheduling plan for the source-grid-load scheduling period; Step S44: Apply the source-grid-load optimized scheduling plan to the intelligent charging piles and distributed power source controllers to execute the corresponding scheduling instructions, and respond to the scheduling instructions to adjust the corresponding electric vehicle charging power and distributed power source generation power in real time.

[0013] Furthermore, the present invention also provides a source-grid-load optimized scheduling analysis system for an electric vehicle active distribution network, which is used to execute the source-grid-load optimized scheduling analysis method for the electric vehicle active distribution network as described above. The source-grid-load optimized scheduling analysis system for the electric vehicle active distribution network includes: A source-grid-load data acquisition module, which is used to collect the power source side data, grid side data, and load side data during the operation of the electric vehicle active distribution network in real time, and perform sliding filtering and normalization processing on the power source side data, grid side data, and load side data to obtain the standard source-grid-load data of the electric vehicle; A load and power output prediction module, which is used to obtain historical charging data, traffic flow data, weather data, and user appearance habit data, and combine a gated recurrent unit to construct an electric vehicle load prediction model to perform load prediction calculations on the standard source-grid-load data of the electric vehicle to predict and output the electric vehicle charging power distribution corresponding to different time periods and different regions; obtain historical power generation power data and numerical weather forecast data, and construct a distributed power source hybrid prediction model to perform power output prediction calculations on the standard source-grid-load data of the electric vehicle to predict and output the distributed power source generation power distribution corresponding to different time periods and different regions; A source-grid-load scheduling model generation module, which is used to construct a corresponding multi-objective optimization function and distribution network operation constraint conditions through the electric vehicle active distribution network, and construct a source-grid-load optimized scheduling model for the multi-objective optimization function based on the distribution network operation constraint conditions to generate a source-grid-load collaborative optimization scheduling model; A source-grid-load optimized scheduling execution module, which is used to set a corresponding source-grid-load scheduling period, and read the corresponding predicted data from the electric vehicle charging power distribution and distributed power source generation power distribution corresponding to different time periods and different regions based on the source-grid-load scheduling period; input the corresponding predicted data into the source-grid-load collaborative optimization scheduling model to solve and obtain the corresponding source-grid-load optimized scheduling plan, and adjust the corresponding electric vehicle charging power and distributed power source generation power in real time according to the source-grid-load optimized scheduling plan.

[0014] Advantages of the present invention: 1. Compared with the prior art, the beneficial effect of the proposed method for optimizing the dispatching analysis of the power source, grid, and load of an electric vehicle active distribution network in the present invention is that by collecting data from the power source side, grid side, and load side of the electric vehicle distribution network in real time, the operation status of the power grid and the charging demand of electric vehicles can be comprehensively monitored, providing accurate power data. These data include information such as the charging power, battery status, and charging time of electric vehicles, as well as information such as the voltage, current, frequency of the power grid, and the power generation power of the power source. By performing sliding filtering on these data, the noise in the data can be eliminated, ensuring the smoothness and reliability of the data. Standardization processing further eliminates the dimensional differences between different data sources, enabling various types of data to be directly compared and analyzed, facilitating subsequent prediction and optimization calculations. In addition, the comprehensive data processing of the power source side, grid side, and load side provides a solid foundation for the accurate prediction of electric vehicle charging demand and the optimal dispatching of the power grid load, significantly improving the stability of the distribution network and the efficiency of power supply, providing real and reliable basic data support for the coordination of the power source, grid, and load, and thus enabling a better analysis of the interaction relationship between electric vehicles and the distribution network. Secondly, by combining multiple data sources (such as historical charging data, traffic flow data, weather data, user habit data, etc.) and using a gated recurrent unit (GRU) model for load prediction, accurate prediction of the charging load of electric vehicles can be achieved. When processing time series data, the gated recurrent unit can effectively capture the changing trends and periodic fluctuations of the charging demand, enhancing the prediction accuracy of the charging power distribution in different time periods and different regions. At the same time, the hybrid prediction model of distributed power sources predicts the power generation capacity of distributed power sources based on historical power generation power and weather forecast data, thereby providing the power generation power distribution on the power source side. These two parts of prediction data not only help decision-makers anticipate the power grid load in advance but also can guide the optimization dispatching of electric vehicle charging and distributed power sources, thereby improving the power grid operation efficiency and the utilization rate of charging facilities, reducing unnecessary power waste, and ensuring the stability of the distribution network and the flexibility of electric vehicles in terms of charging time.Then, by introducing a multi-objective optimization function, multiple objectives can be optimized simultaneously, such as the charging power of electric vehicles, the power generation of distributed power sources, and the grid load, etc., so as to achieve the best coordination of the power source, grid, and load. By constructing an appropriate optimization model, the operating constraints of the grid (such as voltage, frequency, etc. limitations) can be considered. On the premise of ensuring the safe and stable operation of the grid, the charging demand of electric vehicles and the power generation capacity of distributed power sources can be rationally allocated. Optimized scheduling under the operating constraints of the distribution network helps to reduce the grid pressure caused by unbalanced load or overcharging, improve the load-carrying capacity of the grid and the charging efficiency of electric vehicles. The source-grid-load collaborative optimization scheduling model can comprehensively consider the electric vehicle load and the power generation of distributed power sources, ensure the optimal allocation of power resources, reduce energy waste, improve the overall operating efficiency of the system, and can achieve the collaborative optimization scheduling between the power source, grid, and load, adapting to the changes of the grid and charging demand. Finally, by setting the source-grid-load scheduling cycle, the charging power of electric vehicles and the power generation of distributed power sources can be dynamically adjusted according to actual needs, so as to achieve real-time optimized scheduling. By inputting the predicted charging power distribution of electric vehicles and the power generation distribution of distributed power sources, the source-grid-load collaborative optimization scheduling model can quickly calculate the best scheduling plan and realize the real-time adjustment of power resources. This flexible scheduling mechanism can not only improve the charging efficiency of electric vehicles, avoid overcharging and grid overload, but also automatically adjust the power output according to the real-time load changes to ensure the stable operation of the grid under various load conditions. This can minimize the impact on the grid, improve the flexibility and reliability of grid operation, and thus achieve the collaborative optimization of the power source, grid, and load, and improve the collaborative scheduling efficiency between the power source, grid, and load.

[0015] 2. The source-grid-load optimization scheduling analysis system for the electric vehicle active distribution network proposed by the present invention is generally composed of a source-grid-load data acquisition module, a load and power output prediction module, a source-grid-load scheduling model generation module, and a source-grid-load optimization scheduling execution module, and can implement the source-grid-load optimization scheduling analysis method for any electric vehicle active distribution network described in the present invention. It is used to realize the source-grid-load optimization scheduling analysis method for the electric vehicle active distribution network through the operation between computer programs running on each module. The internal structure of the system cooperates with each other, which can greatly reduce repetitive work and manpower input, and can quickly and effectively provide a more accurate and efficient source-grid-load optimization scheduling analysis process for the electric vehicle active distribution network, thus simplifying the operation process of the source-grid-load optimization scheduling analysis system for the electric vehicle active distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Other features, objects, and advantages of the present invention will become more apparent by reading the detailed description of the non-limiting embodiments with reference to the following drawings: Figure 1Schematic diagram of the step flow of the source-network-load optimal scheduling analysis method for the electric vehicle active distribution network of the present invention; Figure 2 is Figure 1 detailed step flow diagram of step S1 in Specific implementation manner

[0017] The technical method of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those skilled in the art within the scope of the present invention without creative work based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0018] To achieve the above object, please refer to Figures 1 to 2 , the present invention provides a source-network-load optimal scheduling analysis method for an electric vehicle active distribution network. In the embodiments of the present invention, please refer to Figure 1 shown, which is a schematic diagram of the step flow of the source-network-load optimal scheduling analysis method for the electric vehicle active distribution network of the present invention. In this example, the source-network-load optimal scheduling analysis method for the electric vehicle active distribution network includes the following steps: Step S1: By collecting power source side data, grid side data, and load side data during the operation of the electric vehicle active distribution network in real time, and performing sliding filtering and normalization processing on the power source side data, grid side data, and load side data, standard data of the electric vehicle source-network-load is obtained; In the embodiments of the present invention, in the electric vehicle active distribution network, power source side data, grid side data, and load side data are collected in real time through various sensors and monitoring devices. The power source side data includes the power generation power, remaining capacity, and operating status of distributed power sources (such as photovoltaic power stations and wind farms). For example, the power generation power of a certain photovoltaic power station is measured in real time by a power sensor, and data is recorded every 15 minutes. The remaining capacity is monitored in real time by the battery management system and calculated according to the charge and discharge conditions of the battery. The formula is , where is the current remaining capacity, is the initial remaining capacity, is the current, is the rated capacity of the battery; Grid-side data includes operating parameters such as the voltage of each node in the distribution network, line current, network topology, and transformer capacity limit. The node voltage and line current are measured in real time through voltage sensors and current sensors. The network topology is obtained through the grid geographic information system (GIS), and the transformer capacity limit is determined according to the rated capacity of the transformer; Load-side data includes the real-time power demand of traditional fixed loads, the location of electric vehicles, the remaining battery power, the estimated driving time, and the charging demand, etc. The real-time power demand of traditional fixed loads is measured in real time through smart meters, and the relevant data of electric vehicles is obtained through the vehicle-to-grid communication interface (V2G). The collected power-side data, grid-side data, and load-side data are subjected to sliding filtering processing. Taking the power-side data as an example, assuming that the generated power sequence of a certain distributed power source collected is , set the sliding window size to 4. In the first sliding window, the data included is , calculate the average value of the data in this window , as the filtered value of this window. When the window slides backward by one time point, the second window contains the data , calculate its average value , and so on. Then, perform normalization processing. Adopt the maximum-minimum normalization method, and the formula is , where is the value after normalization,[[]] is the original data,[[]] is the minimum value in the data sequence,[[]] is the maximum value in the data sequence. For example, for the voltage data of a certain node, after normalization processing, its value is mapped to the interval of 0-1, eliminating the dimensional difference between different formats of data, and finally obtaining the standard data of the electric vehicle source-grid-load.[[]]

[0019] Step S2: Obtain historical charging data, traffic flow data, weather data, and user presence habits data, and combine the gated recurrent unit to construct an electric vehicle load prediction model to perform load prediction calculations on the electric vehicle source-grid-load standard data, so as to predict and output the electric vehicle charging power distribution corresponding to different time periods and different regions; Obtain historical generated power data and numerical weather forecast data, and construct a distributed power source hybrid prediction model to perform power output prediction calculations on the electric vehicle source-grid-load standard data, so as to predict and output the distributed power source generated power distribution corresponding to different time periods and different regions; In the embodiments of the present invention, by obtaining historical charging data, which comes from the charging records of electric vehicles in the past 1-2 years, including information such as charging time, charging duration, and charging power. At the same time, traffic flow data is obtained from the traffic management department, including the number of vehicle flows of electric vehicles at different times and regions, weather data is obtained from the meteorological department, including information such as temperature, humidity, and illumination, and user travel habit data is collected through questionnaires or mobile applications, including daily travel time, routes, and destinations, etc. A load prediction model for electric vehicles is constructed by combining a gated recurrent unit (GRU). First, the historical charging data, traffic flow data, weather data, and user travel habit data are preprocessed and converted into a format suitable for model input. Then, the processed data is input into the GRU model for training. During the training process, the number of hidden layer neurons of the GRU model is set to 64, the learning rate is 0.001, and the number of iterations is 100. And by using the mean square error (MSE) as the loss function, the parameters of the model are adjusted through the backpropagation algorithm to minimize the error between the prediction result of the model and the actual charging power distribution. For example, during the training process, the model predicts that the charging power of electric vehicles in a certain area at a certain time period is 500 kW, while the actual charging power is 520 kW, then the mean square error is calculated as (500 - 520) 2= 400. Through continuous iterative training, the prediction accuracy of the model is continuously improved. Finally, the standard data of the electric vehicle source-network-load is input into the trained GRU model for load prediction calculation, and the charging power distribution of electric vehicles corresponding to different time periods and different regions is predicted and output; for distributed power sources, historical power generation data and numerical weather forecast data are obtained. The historical power generation data includes the power generation of distributed power sources and the corresponding timestamp information within the past 1 - 2 years, and the numerical weather forecast data includes meteorological parameters such as solar radiation intensity, wind speed, wind direction, and temperature in the next few days. And a hybrid prediction model for distributed power sources is constructed. For photovoltaic power plants, a hybrid prediction model combining a convolutional neural network (CNN) and a long short-term memory network (LSTM) is used. The CNN is used to extract spatial features in meteorological data, such as the distribution of solar radiation intensity in different regions. The convolutional kernel size of the CNN is set to 3×3, and the stride is 1. Local features of meteorological data are extracted through convolutional operations. The LSTM is used to capture the time series features corresponding to the power generation. The number of neurons in the hidden layer of the LSTM is set to 128. For wind farms, a hybrid prediction model based on wavelet transform combined with support vector machine (SVM) is constructed. First, the wavelet transform is used to decompose meteorological data (such as wind speed time series) to extract features of different frequency components. Then, the extracted features are input into the SVM for training and prediction. By setting the penalty parameter of the SVM to 10 and the kernel function to the radial basis function (RBF), the historical power generation data and numerical weather forecast data are input into the constructed hybrid prediction model for distributed power sources for training. The mean square error (MSE) is used as the loss function. By adjusting the hyperparameters of the model (such as the convolutional kernel size of the CNN, the number of neurons in the hidden layer of the LSTM, the penalty parameter of the SVM, etc.), the error between the prediction result of the model and the actual power generation distribution is minimized. Finally, the standard data of the electric vehicle source-network-load is input into the trained hybrid prediction model for distributed power sources for power output prediction calculation, and the power generation distribution of distributed power sources corresponding to different time periods and different regions is finally predicted and output.

[0020] Step S3: Construct a corresponding multi-objective optimization function and distribution network operation constraints through the electric vehicle active distribution network, and construct a source-network-load optimal scheduling model for the multi-objective optimization function based on the distribution network operation constraints to generate a source-network-load collaborative optimal scheduling model; In the embodiment of the present invention, by considering minimizing the operation cost of the distribution network to form a corresponding economic objective, including power purchase cost, power generation cost corresponding to distributed power sources, network loss cost, and charging compensation cost corresponding to electric vehicles: ; Among them, represents the power purchase cost, represents the total number of distributed power sources, represents the power generation cost corresponding to the th distributed power source, represents the total number of electric vehicles, represents the th electric vehicle's corresponding charging compensation cost; and by considering reducing carbon emissions to maximize the consumption of distributed power sources, the corresponding environmental protection objective is formed: ; where, represents the th distributed power source's corresponding power generation power, represents the total number of traditional thermal loads, represents the th traditional thermal load's corresponding power generation power; meanwhile, by considering the grid voltage stability, frequency stability, and line capacity limitations, etc., to measure, minimizing the node voltage deviation and line overload risk forms the corresponding grid security objective: ; where, represents the total number of nodes in the electric vehicle active distribution network, represents the th node's corresponding voltage deviation, represents the total number of lines in the electric vehicle active distribution network, represents the th line's corresponding current, represents the th line's corresponding rated current, to construct the corresponding multi-objective optimization function according to the above objectives. Meanwhile, by considering the physical constraints corresponding to the operation of the distribution network and the operation characteristics constraints of electric vehicles, the corresponding distribution network operation constraint conditions are constructed, which specifically include 1) Grid operation constraints: Node voltage amplitude constraint: , where and are respectively the lower limit value and upper limit value of the voltage deviation corresponding to the th node; Line current upper limit constraint: , to ensure that the line current does not exceed its rated current; Active power balance constraint: , where represents the distributed power source set, represents the th distributed power source's corresponding active power, represents the traditional thermal load set, represents the th traditional thermal load's corresponding active power, represents the electric vehicle set, represents the The active power corresponding to an electric vehicle represents the active power of network loss; Reactive power balance constraint: , where represents the reactive power corresponding to the th distributed power source, represents the reactive power corresponding to the th traditional thermal load, represents the reactive power corresponding to the th electric vehicle; represents the reactive power of network loss; 2) Distributed power source constraint: Generation power upper and lower limit constraint: , where and respectively represent the minimum generation power and the maximum generation power corresponding to the th distributed power source; Ramp rate constraint: , where represents the time parameter, represents the maximum ramp generation rate corresponding to the th distributed power source; 3) Electric vehicle constraint: Active power upper and lower limit constraint: , where and respectively represent the minimum active power and the maximum active power corresponding to the th electric vehicle; Reactive power upper and lower limit constraint: , where and respectively represent the The minimum reactive power and the maximum reactive power corresponding to an electric vehicle. Then, by adding grid operation constraints, distributed power source constraints, and electric vehicle constraints to the multi-objective optimization function one by one based on the operation constraints of the distribution network, a coordinated optimization scheduling model for the source-grid-load is constructed. First, the line current upper limit constraint, the node voltage amplitude constraint, the active power balance constraint, and the reactive power balance constraint are added to the multi-objective optimization function, and these constraints are satisfied by adjusting the generation power and load distribution. For example, when satisfying the power balance constraint, according to the real-time load condition of the grid and the generation capacity of the distributed power source, the charging power of the electric vehicle is reasonably allocated to ensure the stable operation of the grid. Then, the generation power limit and the ramp rate limit of the distributed power source are added to the multi-objective optimization function. When optimizing the generation power distribution, it is ensured that the generation power of the distributed power source is within its rated power range and the change rate of the generation power does not exceed the ramp rate limit. Finally, the active power and reactive power limits of the electric vehicle are added to the multi-objective optimization function. When arranging the charging plan of the electric vehicle, it is ensured that the charging power is within the allowable range and the charging time meets the travel needs of users. By adding these constraints to the multi-objective optimization function one by one, a coordinated optimization scheduling model for the source-grid-load is constructed. This model can achieve multi-objective optimization scheduling of economy, environmental protection, and grid security on the premise of satisfying the operation constraints of the distribution network and the operation characteristic constraints of the electric vehicle.

[0021] Step S4: By setting the corresponding source-grid-load scheduling period and reading the corresponding prediction data from the charging power distribution of electric vehicles and the generation power distribution of distributed power sources corresponding to different time periods and different regions based on the source-grid-load scheduling period; inputting the corresponding prediction data into the coordinated optimization scheduling model for the source-grid-load to solve for the corresponding source-grid-load optimization scheduling plan, and adjusting the charging power of the corresponding electric vehicle and the generation power of the distributed power source in real time according to the source-grid-load optimization scheduling plan.

[0022] In the embodiment of the present invention, by setting the source-network-load scheduling period to 15 minutes, this period setting is based on the comprehensive consideration of the grid operation stability and the real-time nature of electric vehicle charging demand. A shorter scheduling period can respond more promptly to the state changes of the grid and electric vehicles, but it will also increase the complexity of calculation and control; a longer scheduling period will result in a less sensitive response to real-time changes. After multiple simulations and actual operation tests, a 15-minute scheduling period can, while ensuring the stable operation of the grid, better meet the charging demand of electric vehicles. Based on the set 15-minute source-network-load scheduling period, the prediction data corresponding to each scheduling period is read from the previously predicted electric vehicle charging power distribution and distributed generation power distribution data corresponding to different time periods and different regions. Taking a specific area in a certain city as an example, before the 10:00 - 10:15 scheduling period, the electric vehicle charging power distribution and distributed generation power distribution in this area during this time period have been predicted through the electric vehicle load prediction model and the distributed generation hybrid prediction model. Suppose the prediction results show that there are 50 electric vehicles in this area, the total charging power is expected to be 1000 kW during this scheduling period, and the distributed generation power is 1500 kW. According to the time range of the scheduling period, these prediction data are accurately read to provide data support for the subsequent formulation of the optimized scheduling plan. These data will be used as the input of the source-network-load collaborative optimization scheduling model to ensure that the model can perform accurate optimization calculations according to the actual situation. The prediction data corresponding to each source-network-load scheduling period read, that is, the electric vehicle charging power distribution and distributed generation power distribution data in this area during the 10:00 - 10:15 scheduling period, is input into the source-network-load collaborative optimization scheduling model for solution. The source-network-load collaborative optimization scheduling model is a complex mathematical model that takes into account various factors such as grid operation constraints, distributed generation constraints, and electric vehicle constraints. During the solution process, the model will, according to the multi-objective optimization function, that is, the economic, environmental, and grid security objectives, through a series of algorithms and calculation steps, search for the optimal source-network-load scheduling plan to calculate how to reasonably allocate the charging power of electric vehicles and the distributed generation power during this scheduling period to achieve the optimal comprehensive benefit. Finally, the model solves to obtain the source-network-load optimized scheduling plan corresponding to this source-network-load scheduling period. For example, in this example, the optimized scheduling plan may be to arrange some electric vehicles to charge quickly from 10:00 to 10:05 first, and then charge slowly from 10:05 to 10:15, while adjusting the distributed generation power to ensure the stable operation of the grid and the optimal economic benefit. Intelligent charging piles and distributed generation controllers are the key devices to achieve source-network-load optimized scheduling. Taking intelligent charging piles as an example, they can adjust the charging power of electric vehicles in real time according to the received scheduling instructions.For example, during the period from 10:00 to 10:05, the intelligent charging pile sets the charging power of some electric vehicles to a relatively high value according to the dispatching instruction to meet the demand for fast charging; during the period from 10:05 to 10:15, the charging power is adjusted to a relatively low value to avoid excessive impact on the power grid. Similarly, the distributed power controller adjusts the power generation power of the distributed power source in real time according to the dispatching instruction. For example, the distributed power controller may appropriately increase or decrease the power generation power of a certain photovoltaic power station or wind farm during the period from 10:00 to 10:15 according to the optimized dispatching scheme to maintain the power balance and stable operation of the power grid. Through the response of the intelligent charging pile and the distributed power controller to the dispatching instruction, the real-time adjustment of the charging power of electric vehicles and the power generation power of distributed power sources is realized, thus effectively implementing the source-network-load optimized dispatching scheme and improving the operation efficiency and comprehensive benefits of the electric vehicle active distribution network.

[0023] Further, as an embodiment of the present invention, referring to Figure 2 shown, it is Figure 1 a detailed step flow schematic diagram of step S1 in Step S11: Real-time collect the power source side data during the operation of the electric vehicle active distribution network, including the power generation power, remaining capacity, and real-time operation status data of the distributed power source; In the embodiment of the present invention, in the electric vehicle active distribution network, the power source side data is real-time collected through the distributed power source monitoring device. Taking a regional distribution network as an example, this region contains 10 distributed power sources, such as photovoltaic power stations and wind farms. The monitoring device of each distributed power source collects the power generation power data at a frequency of once per second, and a high-precision power sensor is used with a measurement error not exceeding ±0.5%. For example, the power generation power of photovoltaic power station A at a certain moment is 500 kW, which is real-time transmitted to the data acquisition center through the sensor. For the remaining capacity data, it is obtained through the battery management system and calculated using the ampere-hour integration method. The formula is where is the current remaining capacity, is the initial remaining capacity, is the current, is the rated capacity of the battery. Assume that the initial remaining capacity of the battery corresponding to a certain distributed power source is 80%. After a period of charge and discharge, the current remaining capacity is calculated to be 70% according to the current integration. The real-time operation status data includes the start, stop, fault and other statuses of the power source, which are real-time monitored through the sensor and the control system. For example, a certain wind turbine in wind farm B is in normal operation, and the data acquisition center receives this information in real time. These power source side data are real-time collected and stored in the database to provide basic data for subsequent analysis and dispatching.

[0024] Step S12: Real-time collect the grid-side data during the operation of the electric vehicle active distribution network, including the voltage of each node of the distribution network, line current, network topology, and transformer capacity limit operating parameters; In the embodiment of the present invention, the grid-side data during the operation of the electric vehicle active distribution network is collected in real time by using a grid monitoring system. In this regional distribution network, 50 monitoring nodes are set, distributed in each substation, transmission line, and distribution area. Each node is equipped with a high-precision voltage sensor and current sensor. The voltage measurement error does not exceed ±1%, and the current measurement error does not exceed ±0.8%. Taking a certain moment as an example, the voltage sensor at node C measures a voltage value of 220V, and the current sensor measures a current value of 100A. These data are transmitted to the data acquisition center in real time. The network topology data is obtained through the grid geographic information system (GIS), which records the connection relationship and location information of each component (such as lines, transformers, switches, etc.) in the distribution network. The transformer capacity limit operating parameters are determined according to the rated capacity and actual operating conditions of the transformer. For example, the rated capacity of transformer D is 1000 kVA, and the current load rate is 70%, then its actual operating capacity is 700 kVA. These grid-side data are collected and integrated into the database in real time for analyzing the operating status of the grid and making scheduling decisions.

[0025] Step S13: Real-time collect the load-side data during the operation of the electric vehicle active distribution network, including the real-time power demand of traditional fixed loads, the location of electric vehicles, the remaining battery power, the estimated driving time, and the charging demand data; In the embodiment of the present invention, the load-side data is collected in real time by smart meters and an electric vehicle charging management system. In this area, there are 1000 traditional fixed load users, such as residential houses, commercial stores, and industrial enterprises. The smart meters collect the real-time power demand data of traditional fixed loads at a frequency of once every 15 minutes and upload it to the data acquisition center. For example, the real-time power demand of a certain residential house at a certain moment is 3 kW. For the relevant data of electric vehicles, it is obtained through the vehicle-to-grid communication interface (V2G). Assuming there are 500 electric vehicles in this area, the battery management system of each electric vehicle sends data such as the remaining battery power and estimated driving time to the charging management system in real time. For example, the remaining battery power of electric vehicle E is 50%, and the estimated driving time is 2 hours. The charging demand data is calculated according to the remaining battery power, estimated driving time, and the charging power demand model of the electric vehicle. The formula is , where is the charging power demand, is the current remaining capacity, is the battery rated capacity, For the estimated driving time, assuming that the rated capacity of an electric vehicle battery is 60 kWh, the current remaining capacity is 40%, and the estimated driving time is 1.5 hours, then its charging power demand is = 24 kW. The data on the load side are collected and stored in the database in real time for load forecasting and optimal scheduling.

[0026] Step S14: By setting the sliding window size at 3 - 5 time points, and performing sliding filtering on the power source side data, grid side data, and load side data based on this sliding window size, the sliding filtered data of the electric vehicle's power source, grid, and load are obtained; In the embodiment of the present invention, by setting the sliding window size at 3 - 5 time points, for example, setting the sliding window size to 4 time points, and taking the power source side data as an example for sliding filtering. Assume that the power source side data is the power generation power sequence of a certain distributed power source , within the first sliding window, it contains the data , calculate the average value of the data within this window , as the filtered value of this window. When the window slides backward by one time point, the second window contains the data , calculate its average value , and so on, perform sliding filtering on the entire power source side data sequence. For the grid side data and load side data, the same method is used for sliding filtering. For example, for the voltage data sequence of a certain node , after sliding filtering, the filtered voltage sequence is obtained. Through sliding filtering, the data fluctuations can be smoothed, noise interference can be reduced, and finally the sliding filtered data of the electric vehicle's power source, grid, and load are obtained, providing a more stable data basis for subsequent data processing and analysis.

[0027] Step S15: Perform interpolation filling and standardization processing on the sliding filtered data of the electric vehicle's power source, grid, and load to fill the corresponding missing data using median interpolation and eliminate the dimension between different format data, obtaining the standard data of the electric vehicle's power source, grid, and load.

[0028] In the embodiment of the present invention, through interpolation filling and standardization processing of the sliding filtered data of the electric vehicle's power source, grid, and load, first perform interpolation filling. Assume that the power generation power data at a certain moment in the power source side data is missing. The median interpolation method is used, that is, take the median of the two adjacent data before and after this data point as the filling value. If = 450 kW, = 550 kW, then the filling value is (450 + 550) / 2 = 500 kW. Then perform standardization processing. Taking the power source side data as an example, the maximum - minimum standardization method is used, and the formula is , where is the value after standardization, is the original data, is the minimum value in the data sequence, is the maximum value in the data sequence. Assuming that the minimum value of the power generation power data sequence on the power supply side is 0 kW and the maximum value is 1000 kW, and the original value of a certain data point is 600 kW, then the value after standardization is (600 - 0) / (1000 - 0) = 0.6. The same interpolation filling and standardization processing are performed on the grid-side data and the load-side data. For example, for the voltage data of a certain node, after standardization processing, its value is mapped to the interval of 0 - 1, eliminating the dimensional difference between different formats of data. Finally, the standard data of the electric vehicle source-grid-load are obtained. These standard data can be more conveniently compared and analyzed, providing accurate data support for the optimal scheduling of the source-grid-load of the electric vehicle active distribution network.

[0029] Further, step S2 includes the following steps: Step S21: Obtain historical charging data, specifically the charging records of electric vehicles within the past 1 - 2 years, including charging time, charging duration, and charging power; In the embodiment of the present invention, the historical charging data is obtained from the electric vehicle charging management system, which records the charging records of 5000 electric vehicles in a certain city within the past 1 - 2 years. The data is stored in tabular form, and each row represents a charging event, including information such as charging time (accurate to minutes), charging duration (unit: hours), and charging power (unit: kW). For example, a certain record shows that the charging time is "2022 - 05 - 10 18:30:00", the charging duration is 1.5 hours, and the charging power is 30 kW. By sorting and analyzing these data, the charging behavior patterns of electric vehicles can be understood, providing basic data for subsequent load forecasting.

[0030] Step S22: Obtain traffic flow data, including the number of vehicle flows of electric vehicles at different times and regions; In the embodiment of the present invention, the traffic flow data is obtained from the intelligent transportation system of the urban traffic management department. This system monitors the vehicle flow situation in real time through sensors and cameras installed on the road. The data covers the number of electric vehicle flows at different times (such as the morning and evening rush hours and off-peak hours on weekdays, different times on weekends, etc.) and regions (such as commercial areas, residential areas, industrial areas, etc.). Taking a certain commercial area as an example, during the period from 18:00 to 19:00 on weekdays, an average of 10 electric vehicles pass through this area per minute. By analyzing these data, the travel patterns of electric vehicles at different times and regions can be understood, providing a reference basis for load forecasting.

[0031] Step S23: Obtain weather data and user travel habit data, where the weather data includes temperature, humidity, and light, and the user travel habit data includes daily travel time, route, and destination collected through questionnaires or mobile applications; In the embodiment of the present invention, weather data is obtained from the website of the meteorological department, and the data includes information such as daily temperature (unit: °C), humidity (unit: %), and light (unit: lux). At the same time, user travel habit data is collected through designed questionnaires and mobile applications. The questionnaire content includes information such as daily travel time, route, and destination. The mobile application obtains its travel trajectory data through user authorization. For example, a user fills in their daily travel time as 7:00 - 9:00 and 17:00 - 19:00 on weekdays in the questionnaire, the travel route is mainly from home to the company, and the destination is the company. By sorting and analyzing these data, the impact of weather factors and user travel habits on the charging demand of electric vehicles can be understood, providing more comprehensive data support for load forecasting.

[0032] Step S24: Based on historical charging data, traffic flow data, weather data, and user travel habit data, and combined with a gated recurrent unit, construct an electric vehicle load forecasting model to perform load forecasting calculations on the standard data of the electric vehicle source-network-load, so as to predict and output the electric vehicle charging power distribution corresponding to different time periods and different regions; In the embodiment of the present invention, an electric vehicle load forecasting model is constructed by combining a gated recurrent unit (GRU) based on the obtained historical charging data, traffic flow data, weather data, and user travel habit data. The GRU is a recursive neural network that can effectively process time series data. First, the data is preprocessed, and the charging time, traffic flow, weather data, and user travel habit data are encoded and normalized to have the same scale and format. Then, the processed data is input into the GRU model for training. During the training process, the number of neurons in the hidden layer of the GRU model is set to 64, the learning rate is 0.001, and the number of iterations is 100. By using the mean square error (MSE) as the loss function and adjusting the parameters of the model through the backpropagation algorithm, the error between the prediction result of the model and the actual charging power distribution is minimized. For example, during the training process, the model predicts the electric vehicle charging power in a certain area at a certain time period to be 500 kW, while the actual charging power is 520 kW, then the mean square error is calculated as (500 - 520) 2 = 400. Through continuous iterative training, the prediction accuracy of the model is continuously improved. Finally, the standard data of the electric vehicle source-network-load is input into the trained GRU model for load forecasting calculations, and the electric vehicle charging power distribution corresponding to different time periods and different regions is predicted and output.

[0033] Step S25: Obtain historical power generation data and numerical weather prediction data, and construct a distributed power source hybrid prediction model to perform power output prediction calculations on the standard data of the electric vehicle source-grid-load, so as to predict the distributed power generation power distribution corresponding to different time periods and different regions.

[0034] In the embodiment of the present invention, historical power generation data is obtained from the historical database of the power system. The data includes the power generation (unit: kW) of distributed power sources (such as photovoltaic power stations and wind farms) and the corresponding timestamp information within the past 1-2 years. At the same time, numerical weather prediction data is obtained from the numerical weather prediction system of the meteorological department. The data includes meteorological parameters such as solar radiation intensity (unit: W / m²), wind speed (unit: m / s), wind direction, and temperature (unit: °C) in the next few days. For photovoltaic power stations, a hybrid prediction model combining a convolutional neural network (CNN) and a long short-term memory network (LSTM) is constructed. The CNN is used to extract the spatial features in the meteorological data, such as the distribution of solar radiation intensity in different regions. The convolution kernel size of the CNN is set to 3×3, and the stride is 1. The local features of the meteorological data are extracted through convolution operations. The LSTM is used to capture the time series features corresponding to the power generation. The number of neurons in the hidden layer of the LSTM is set to 128. For wind farms, a hybrid prediction model based on wavelet transform and support vector machine (SVM) is constructed. First, the wavelet transform is used to decompose the meteorological data (such as the wind speed time series) to extract the features of different frequency components. Then, the extracted features are input into the SVM for training and prediction. By setting the penalty parameter of the SVM to 10 and the kernel function to the radial basis function (RBF), the historical power generation data and numerical weather prediction data are input into the constructed distributed power source hybrid prediction model for training. The mean square error (MSE) is used as the loss function. By adjusting the hyperparameters of the model (such as the convolution kernel size of the CNN, the number of neurons in the hidden layer of the LSTM, the penalty parameter of the SVM, etc.), the error between the prediction result of the model and the actual power generation distribution is minimized. Finally, the standard data of the electric vehicle source-grid-load is input into the trained distributed power source hybrid prediction model for power output prediction calculations, and the distributed power generation power distribution corresponding to different time periods and different regions is predicted and output.

[0035] Further, step S24 includes the following steps: Step S241: Obtain the charging time characteristics of electric vehicles corresponding to hours, days, weeks, months, and seasons through historical charging data; In the embodiment of the present invention, historical charging data for the past 1 year is retrieved from the active distribution network data center of electric vehicles. The data covers the charging records of 5,000 electric vehicles in the area and is stored in the "Historical Charging Database" table. Each record contains information such as the charging start time, end time, charging power, and charging duration. Taking the hour dimension as an example, the charging times and the total charging power of electric vehicles within each hour are statistically calculated, and the ratio of the charging times in each hour to the total charging times is calculated to obtain the hourly charging time characteristics. Suppose it is statistically found that the proportion of charging times from 22:00 to 23:00 every day reaches 15%, which is the peak charging period of the day. In the daily dimension, the average charging power of each day within a week is analyzed. For example, it is found that the average charging power on weekdays is 30 kW, and the average charging power on weekends is 25 kW. In the weekly dimension, the change trends of the total charging amounts in different weeks are compared, and the growth rate of the weekly total charging amount is calculated. In the monthly dimension, the total charging duration and charging times of each month are statistically calculated, and seasonal changes are analyzed. For example, due to the increase in travel from June to August in summer, the monthly total charging amount increases by 20% on average compared with other seasons. Through these calculations and analyses, the charging time characteristics of electric vehicles corresponding to hours, days, weeks, months, and seasons are comprehensively obtained, providing regular data in the time dimension for load forecasting.

[0036] Step S242: Obtain the vehicle traffic characteristics of electric vehicles in different regions through traffic flow data, including vehicle traffic density and vehicle congestion index. In the embodiment of the present invention, data is obtained from the real-time traffic flow monitoring system of the traffic management department. The system sets 100 traffic monitoring points in the area and records vehicle passing data at intervals of one minute, which is stored in the "Traffic Flow Database" table and contains information such as the monitoring point location, the number of passing vehicles, and the vehicle speed. The vehicle traffic density is determined by calculating the number of vehicles per unit area. The formula is: traffic density = number of passing vehicles / monitored area. For example, in a commercial area, the monitored area is 1 square kilometer, and the number of passing vehicles at a certain moment is 500, then the vehicle traffic density in this area is 500 vehicles per square kilometer. The vehicle congestion index is calculated based on the ratio of the vehicle speed to the free-flow speed of the road. The formula is: congestion index = free-flow speed of the road / actual vehicle speed. Suppose the free-flow speed of a certain road is 60 km / h and the actual vehicle speed is 20 km / h, then the congestion index of this road is 3, indicating a serious congestion state. The area is divided into different functional areas such as commercial areas, residential areas, and industrial areas, and the vehicle traffic density and congestion index of each area are calculated respectively. Finally, the vehicle traffic characteristics in different regions are obtained, reflecting the impact of traffic conditions on the driving and charging of electric vehicles.

[0037] Step S243: Obtain the weather influence characteristics corresponding to electric vehicles through weather data, including the change rates corresponding to temperature, humidity, and light intensity. In an embodiment of the present invention, weather data is obtained from the data interface of the meteorological department. The data update frequency is once per hour and is stored in the "Weather Database" table, which contains information such as date, time, temperature, humidity, light intensity, etc. The temperature change rate is calculated using the formula: Temperature change rate = (Current temperature - Temperature of the previous hour) / Temperature of the previous hour. For example, if the temperature at a certain moment is 25°C and the temperature of the previous hour is 23°C, then the temperature change rate = (25 - 23) / 23 ≈ 0.087. The calculation method for the humidity change rate is similar, with the formula: Humidity change rate = (Current humidity - Humidity of the previous hour) / Humidity of the previous hour. The formula for the light intensity change rate is: Light intensity change rate = (Current light intensity - Light intensity of the previous hour) / Light intensity of the previous hour. By continuously calculating the change rates of temperature, humidity, and light intensity over a period of time, such as statistically analyzing the change rate data for each hour of each day in the past week, the weather impact characteristics corresponding to the electric vehicle are finally obtained. Through analysis, it is found that for every 1°C increase in temperature, the use of the electric vehicle air conditioner increases, and the average charging demand increases by 3%, providing weather factor-related data for load forecasting.

[0038] Step S244: Obtain the corresponding user travel characteristics through the user travel habit data, including travel distance, travel purpose, and charging frequency; analyze the potential impact characteristics of the source-network-load status on electric vehicle charging through the analysis of electric vehicle source-network-load standard data; In an embodiment of the present invention, user travel habit data is obtained from the electric vehicle user management system, including the travel records of 3000 users in the region, and is stored in the "User Travel Database" table. The recorded content includes information such as the starting place, destination, travel distance, travel time, travel purpose, charging frequency, etc. The average travel distance of users is calculated using the formula: Average travel distance = Total travel distance / Number of trips. For example, if a user travels 100 times in a year and the total travel distance is 5000 kilometers, then the average travel distance is 50 kilometers. The proportion of the number of trips for different travel purposes (such as commuting, shopping, tourism, etc.) is statistically analyzed, and the relationship between the charging frequency and the travel distance and travel purpose is analyzed. For example, it is found that users with a travel distance exceeding 100 kilometers have a 40% higher charging frequency than users with short-distance travel. At the same time, the electric vehicle source-network-load standard data is analyzed, the difference between the power generation power on the power supply side and the charging demand on the load side is calculated, and the potential impact of the source-network-load status on electric vehicle charging is evaluated. For example, when the remaining power generation power on the power supply side is small, the charging of electric vehicles will be restricted. These characteristics are extracted, and finally the user travel characteristics and the potential impact characteristics of the source-network-load status on electric vehicle charging are obtained.

[0039] Step S245: Build an electric vehicle load prediction model by selecting a gated recurrent unit (GRU). Use the charging time characteristics, vehicle traffic characteristics, weather impact characteristics, user travel characteristics, and the potential impact characteristics of the source-network-load status on electric vehicle charging as training data to train the electric vehicle load prediction model. By adjusting the hyperparameters corresponding to the model, including the number of neurons in the hidden layer, learning rate, and number of iterations, and minimizing the prediction error. At the same time, input the standard source-network-load data of electric vehicles into the trained electric vehicle load prediction model for load prediction calculation to predict the charging power distribution corresponding to different time periods and different regions.

[0040] In the embodiment of the present invention, build an electric vehicle load prediction model by using a gated recurrent unit (GRU). The model includes 3 hidden layers, and the number of neurons in the hidden layers is set to 64, 32, and 16 respectively. The initial value of the learning rate is set to 0.001, and the number of iterations is set to 100 times. Integrate the charging time characteristics, vehicle traffic characteristics, weather impact characteristics, user travel characteristics, and the potential impact characteristics of the source-network-load status on electric vehicle charging obtained in the previous steps S241 - S244 to form a training data set. Use the mean square error (MSE) as the prediction error measurement index, and the formula is: , where is the number of samples, is the actual charging power value, is the predicted charging power value. During the training process, adjust the model parameters through the backpropagation algorithm, continuously optimize the connection weights and biases between neurons in the hidden layer, and minimize the mean square error. After the number of iterations reaches 100 times, input the standard source-network-load data of electric vehicles into the trained model for load prediction calculation. For example, input the relevant feature data of a commercial area in the next 24 hours, and the model outputs the predicted values of the charging power distribution per hour in this area. Such as the predicted charging power from 8 - 9 hours is 45kW, and the predicted charging power from 18 - 19 hours is 60kW, providing a basis for the source-network-load optimal scheduling of the active distribution network of electric vehicles for the charging power prediction of different time periods and different regions.

[0041] Further, step S25 includes the following steps: Step S251: Obtain historical power generation data, specifically the historical power generation of distributed power sources and the corresponding timestamp information, where the distributed power sources include photovoltaic and wind power; In an embodiment of the present invention, historical power generation data is obtained from an operation monitoring database of the power system. This database stores relevant information for the past year. Taking distributed power sources in a certain area as an example, which includes 10 photovoltaic power stations and 8 wind farms. For each photovoltaic power station, its power generation at every 15 minutes and the corresponding timestamp information are recorded. For example, the power generation of photovoltaic power station A at 10:00 on January 1, 2020 is 200 kW, and the timestamp is "2020-01-01 10:00:00". For wind farms, the power generation and timestamp at every 15 minutes are also recorded. For example, the power generation of wind farm B at 10:15 on January 1, 2020 is 300 kW, and the timestamp is "2020-01-01 10:15:00". These data are organized into a table form for convenient subsequent analysis and processing, providing basic data for the analysis of the output characteristics of distributed power sources.

[0042] Step S252: Obtain numerical weather forecast data, including meteorological parameters such as solar radiation intensity, wind speed, wind direction, and temperature; In an embodiment of the present invention, data is obtained from the numerical weather forecast database of the meteorological department. This database is updated hourly. Taking a certain area as an example, meteorological parameters such as solar radiation intensity, wind speed, wind direction, and temperature are obtained. For example, at 10:00 on January 1, 2020, the solar radiation intensity in this area is 800 W / m², the wind speed is 5 m / s, the wind direction is northwest, and the temperature is 10 °C. These meteorological parameters have an important impact on predicting the power generation of distributed power sources. For example, the solar radiation intensity directly affects the power generation of photovoltaic power stations, and the wind speed and wind direction affect the power generation of wind farms. The obtained meteorological parameters are associated with the historical power generation data for subsequent analysis of the impact of meteorological factors on the output of distributed power sources.

[0043] Step S253: Obtain the output time characteristics of distributed power sources corresponding to hours, days, weeks, months, and seasons through historical power generation data; In the embodiments of the present invention, by analyzing historical power generation data, the output time characteristics of distributed power sources at different time scales are obtained. Taking a photovoltaic power station as an example, on the hourly scale, the average power generation per hour is statistically calculated. For example, on a certain day in summer, the average power generation from 11:00 to 12:00 is 500 kW. On the daily scale, the total power generation and the power generation duration per day are calculated. For example, the total power generation on a certain day is 8000 kW·h and the power generation duration is 10 hours. On the weekly scale, the changing trend of the power generation per week is analyzed and the weekly average power generation is calculated. For example, the weekly average power generation in a certain week is 6000 kW·h. On the monthly scale, the total power generation and the number of power generation days per month are statistically calculated. For example, the total power generation in a certain month is 180000 kW·h and the number of power generation days is 25 days. On the seasonal scale, the power generation differences in different seasons are compared. For example, the average power generation in summer is significantly higher than that in winter. Through these analyses, the output time characteristics of distributed power sources are comprehensively understood, providing information in the time dimension for the prediction model.

[0044] Step S254: Obtain the influence characteristics of solar radiation intensity, wind speed, wind direction and temperature on the power generation capacity corresponding to the output of distributed power sources through numerical weather forecast data; obtain the potential influence characteristics of the source-grid-load state on the output of distributed power sources through the analysis of electric vehicle source-grid-load standard data; In the embodiments of the present invention, through the correlation analysis of numerical weather forecast data and historical power generation data, the influence characteristics of solar radiation intensity, wind speed, wind direction and temperature on the power generation capacity of distributed power sources are obtained. Taking a photovoltaic power station as an example, a relationship model between solar radiation intensity and power generation is established. Through data analysis, it is found that when the solar radiation intensity increases from 600 W / m² to 800 W / m², the power generation of the photovoltaic power station increases by an average of 20%. For a wind farm, the influence of wind speed and wind direction on power generation is analyzed. When the wind speed is between 5 - 10 m / s, the power generation increases with the increase of wind speed, and the wind direction also has a certain influence on power generation. For example, when the wind direction is consistent with the fan layout direction of the wind farm, the power generation is higher. At the same time, through the analysis of electric vehicle source-grid-load standard data, the potential influence characteristics of the source-grid-load state on the output of distributed power sources are obtained. For example, when the demand on the load side increases, the output of distributed power sources needs to increase accordingly to meet the power balance of the power grid.

[0045] Step S255: Build a distributed power hybrid prediction model by selecting the corresponding deep learning algorithm. For photovoltaic power, a hybrid prediction model combining a convolutional neural network and a long short-term memory network is adopted. The convolutional neural network is used to extract the spatial features in meteorological data, and the long short-term memory network is used to capture the time series features corresponding to the power generation. For wind power, the time series of meteorological data is decomposed based on wavelet transform to extract the features of different frequency components, and a hybrid prediction model is constructed in combination with a support vector machine. At the same time, the output time features, the features affecting the power generation capacity, and the potential impact features of the source-network-load status on the distributed power output are used as training data to train the distributed power hybrid prediction model. By adjusting the hyperparameters corresponding to the model, including the convolutional kernel size, the step size, the number of neurons in the hidden layer, and the penalty parameter, and minimizing the prediction error, and inputting the standard data of the electric vehicle source-network-load into the trained distributed power hybrid prediction model for power output prediction calculation to predict the distributed power generation power distribution corresponding to different time periods and different regions.

[0046] In the embodiment of the present invention, for a photovoltaic power station, a hybrid prediction model is constructed by combining a convolutional neural network (CNN) and a long short-term memory network (LSTM). The CNN is used to extract the spatial features in meteorological data, such as the distribution of solar radiation intensity in different regions. Assuming a 3×3 convolutional kernel and a step size of 1, the meteorological data is convolved to extract local features. The LSTM is used to capture the time series features corresponding to the power generation, and the number of neurons in the hidden layer is set to 128. Taking a certain photovoltaic power station as an example, the historical meteorological data and power generation data are used as training data to train the model. By adjusting the hyperparameters of the model, such as the convolutional kernel size, the step size, the number of neurons in the hidden layer, etc., the prediction error is minimized. The mean square error (MSE) is also used as a measure of the prediction error, and the formula is , where is the number of samples, is the actual power generation, To predict the power generation, for a wind farm, the time series corresponding to the meteorological data is decomposed based on wavelet transform to extract the characteristics of different frequency components. For example, the wind speed time series is decomposed into low-frequency and high-frequency components. The low-frequency component reflects the long-term trend of the wind speed, and the high-frequency component reflects the short-term fluctuations of the wind speed. Then, a hybrid prediction model is constructed by combining the support vector machine (SVM). The penalty parameter is set to 10. The output time characteristics, the characteristics affecting the power generation capacity, and the potential impact characteristics of the source-network-load status on the distributed power generation are used as training data to train the model. Similarly, by adjusting the hyperparameters, such as the penalty parameter, the prediction error is minimized. The standard data of the source-network-load of electric vehicles is input into the trained distributed power hybrid prediction model for power generation prediction calculation. For example, the meteorological data and the source-network-load status data of a certain area in the future 24 hours are input. The model predicts and outputs the distributed power generation power distribution corresponding to different time periods and different regions in this area. For example, it is predicted that within the next 24 hours, the power generation of a certain photovoltaic power station is 450 kW from 11:00 to 12:00, and the power generation of a certain wind farm is 350 kW from 14:00 to 15:00. These prediction results provide an important basis for the source-network-load optimal scheduling of the active distribution network of electric vehicles, which helps to reasonably arrange the power generation plan of distributed power sources and improve the stability and reliability of the power grid.

[0047] Further, step S3 includes the following steps: Step S31: Construct a corresponding multi-objective optimization function through the active distribution network of electric vehicles, including an economic objective, an environmental protection objective, and a power grid security objective; In the embodiment of the present invention, by considering minimizing the operation cost of the distribution network to form the corresponding economic objective, including the power purchase cost, the power generation cost corresponding to the distributed power source, the network loss cost, and the charging compensation cost corresponding to the electric vehicle: ; Among them, represents the power purchase cost, represents the total number of distributed power sources, represents the th power generation cost corresponding to the distributed power source, represents the network loss cost, represents the total number of electric vehicles, represents the th charging compensation cost corresponding to the electric vehicle; and by considering reducing carbon emissions to maximize the consumption of the distributed power source to form the corresponding environmental protection objective: ; Among them, represents the th power generation power corresponding to the distributed power source, represents the total number of traditional thermal loads, represents the The power generation power corresponding to a traditional thermal load; at the same time, measured by considering the voltage stability, frequency stability, and line capacity limit of the power grid, etc., minimizing the node voltage deviation and line overload risk constitutes the corresponding power grid security objective: ; among them, represents the total number of nodes in the electric vehicle active distribution network, represents the th node corresponding voltage deviation, represents the total number of lines in the electric vehicle active distribution network, represents the th line corresponding current, represents the th line corresponding rated current. Finally, construct the corresponding multi-objective optimization function for the above corresponding economic objective, environmental protection objective, and power grid security objective.

[0048] Step S32: Construct the corresponding distribution network operation constraint conditions through the electric vehicle active distribution network considering the physical constraints corresponding to the distribution network operation and the operation characteristics constraints of electric vehicles, including power grid operation constraints, distributed power source constraints, and electric vehicle constraints; In the embodiment of the present invention, by considering the physical constraints corresponding to the distribution network operation and the operation characteristics constraints of electric vehicles, construct the distribution network operation constraint conditions, which specifically include 1) power grid operation constraints: node voltage amplitude constraint: , where and are respectively the lower limit value and upper limit value of the voltage deviation corresponding to the th node; line current upper limit constraint: , to ensure that the line current does not exceed its rated current; active power balance constraint: , where represents the distributed power source set, represents the th distributed power source corresponding active power, represents the traditional thermal load set, represents the th traditional thermal load corresponding active power, represents the electric vehicle set, represents the th electric vehicle corresponding active power, represents the network loss active power; reactive power balance constraint: , where represents the th distributed power source corresponding reactive power, represents the The reactive power corresponding to a traditional thermal load denotes the reactive power corresponding to the th electric vehicle, denotes the reactive power of network loss; 2) Distributed power source constraints: Generation power upper and lower limit constraints: , where and respectively denote the minimum generation power and the maximum generation power corresponding to the th distributed power source; Ramping rate constraint: , where denotes the time parameter, denotes the maximum ramping generation rate corresponding to the th distributed power source; 3) Electric vehicle constraints: Active power upper and lower limit constraints: , where and respectively denote the minimum active power and the maximum active power corresponding to the th electric vehicle; Reactive power upper and lower limit constraints: , where and respectively denote the minimum reactive power and the maximum reactive power corresponding to the th electric vehicle. By combining the power grid operation constraints, distributed power source constraints, and electric vehicle constraints, the corresponding power distribution network operation constraint conditions are constructed.

[0049] Step S33: Based on the power distribution network operation constraint conditions, the corresponding power grid operation constraints, distributed power source constraints, and electric vehicle constraints are added one by one to construct a source-network-load optimization dispatch model for the multi-objective optimization function, so as to generate a source-network-load collaborative optimization dispatch model.

[0050] In the embodiments of the present invention, based on the operating constraints of the distribution network, the grid operating constraints, distributed power source constraints, and electric vehicle constraints are added to the multi-objective optimization function one by one to construct a corresponding source-network-load collaborative optimization scheduling model. First, the line current upper limit constraint, node voltage amplitude constraint, active power balance constraint, and reactive power balance constraint are added to the multi-objective optimization function, and the generation power and load distribution are adjusted to meet these constraints. For example, when satisfying the power balance constraint, according to the real-time load situation of the grid and the generation capacity of the distributed power source, the charging power of the electric vehicle is reasonably allocated to ensure the stable operation of the grid. Then, the generation power limit and ramp rate limit of the distributed power source are added to the multi-objective optimization function. When optimizing the generation power distribution, it is ensured that the generation power of the distributed power source is within its rated power range, and the change rate of the generation power does not exceed the ramp rate limit. For example, when the generation power of the distributed power source is close to its rated power, the generation power of other distributed power sources or the charging power of the electric vehicle is appropriately adjusted to avoid exceeding the generation power limit. Finally, the active power limit and reactive power limit of the electric vehicle are added to the multi-objective optimization function. When arranging the charging plan of the electric vehicle, it is ensured that the charging power is within the allowable range and the charging time meets the travel needs of the user. For example, for electric vehicles with a relatively late expected arrival time and an early expected departure time, they are preferentially arranged to charge after other electric vehicles have completed charging to make full use of the remaining capacity of the grid. By adding these constraints to the multi-objective optimization function one by one, a source-network-load collaborative optimization scheduling model is constructed. This model can achieve multi-objective optimization scheduling of economy, environmental protection, and grid security on the premise of meeting the operating constraints of the distribution network and the operating characteristic constraints of electric vehicles.

[0051] Further, step S4 includes the following steps: Step S41: Set a corresponding source-network-load scheduling period, where the source-network-load scheduling period is specifically 15 minutes; In the embodiment of the present invention, in the management system of the active distribution network of electric vehicles, the source-network-load scheduling period can be set to 15 minutes. In this application, it is assumed to be 15 minutes. This period setting is based on the comprehensive consideration of the grid operation stability and the real-time nature of the charging demand of electric vehicles. A shorter scheduling period can respond more promptly to the state changes of the grid and electric vehicles, but it will also increase the complexity of calculation and control. A longer scheduling period will result in a less sensitive response to real-time changes. After multiple simulations and actual operation tests, a 15-minute scheduling period can, while ensuring the stable operation of the grid, better meet the charging demand of electric vehicles. Within each scheduling period (i.e., 15 minutes), the charging power of electric vehicles and the power generation power of distributed power sources will be optimally scheduled to achieve multi-objective optimization of economy, environmental protection, and grid security. For example, within the scheduling period from 10:00 to 10:15 on a certain day, an appropriate optimal scheduling plan will be formulated according to the grid load situation, the power generation capacity of distributed power sources, and the charging demand of electric vehicles during this period.

[0052] Step S42: Based on the source-network-load scheduling period, read the prediction data corresponding to each source-network-load scheduling period from the charging power distributions of electric vehicles and the power generation power distributions of distributed power sources corresponding to different time periods and different regions, including the charging power distribution and the power generation power distribution within the corresponding source-network-load scheduling period. In the embodiment of the present invention, based on the set 15-minute source-network-load scheduling period, the prediction data corresponding to each scheduling period is read from the charging power distributions of electric vehicles and the power generation power distribution data corresponding to different time periods and different regions predicted previously. Taking a specific area in a certain city as an example, before the scheduling period from 10:00 to 10:15, through the electric vehicle load prediction model and the distributed power source hybrid prediction model, the charging power distribution of electric vehicles and the power generation power distribution of distributed power sources in this area during this period have been predicted. Suppose the prediction results show that there are 50 electric vehicles in this area, the total charging power is expected to be 1000 kW within this scheduling period, and the power generation power of the distributed power source is 1500 kW. And according to the time range of the scheduling period, these prediction data are accurately read to provide data support for the subsequent formulation of the optimal scheduling plan. These data will be used as the input of the source-network-load collaborative optimal scheduling model to ensure that the model can perform accurate optimal calculations according to the actual situation.

[0053] Step S43: Input the corresponding prediction data into the source-network-load collaborative optimal scheduling model to solve and obtain the source-network-load optimal scheduling plan corresponding to this source-network-load scheduling period. In the embodiments of the present invention, by inputting the predicted data corresponding to each source-network-load scheduling cycle read, that is, the electric vehicle charging power distribution and distributed power generation power distribution data in this area during the 10:00-10:15 scheduling cycle, into the source-network-load collaborative optimization scheduling model for solution. The source-network-load collaborative optimization scheduling model is a complex mathematical model that takes into account various factors such as power grid operation constraints, distributed power constraints, and electric vehicle constraints. During the solution process, the model will, according to the multi-objective optimization function, that is, the economic, environmental protection, and power grid security objectives, through a series of algorithms and calculation steps, find the optimal source-network-load scheduling plan. For example, the model will, according to the constraint conditions such as the real-time voltage, frequency, and line capacity of the power grid, as well as factors such as the power generation cost, carbon emissions, and power generation power limit of distributed power sources, combined with the power limit of electric vehicles, calculate how to reasonably allocate the charging power of electric vehicles and the power generation power of distributed power sources during this scheduling cycle to achieve the optimal comprehensive benefit. Finally, the model solves to obtain the source-network-load optimization scheduling plan corresponding to this source-network-load scheduling cycle. For example, in this example, the optimization scheduling plan is to arrange some electric vehicles to perform fast charging first from 10:00 to 10:05, and then perform slow charging from 10:05 to 10:15, while adjusting the power generation power of distributed power sources to ensure the stable operation of the power grid and the optimal economic benefit.

[0054] Step S44: Apply the source-network-load optimization scheduling plan to the intelligent charging piles and distributed power source controllers to execute the corresponding scheduling instructions, and respond to the scheduling instructions to adjust the corresponding electric vehicle charging power and distributed power generation power in real time.

[0055] In the embodiment of the present invention, by applying the source-network-load optimization scheduling scheme obtained by solving the source-network-load collaborative optimization scheduling model to the intelligent charging pile and the distributed power source controller, and executing the corresponding scheduling instructions, the intelligent charging pile and the distributed power source controller are the key devices for realizing the source-network-load optimization scheduling. Taking the intelligent charging pile as an example, it can adjust the charging power of electric vehicles in real time according to the received scheduling instructions. For example, during the period from 10:00 to 10:05, the intelligent charging pile sets the charging power of some electric vehicles to a higher value according to the scheduling instructions to meet the demand for fast charging; during the period from 10:05 to 10:15, the charging power is adjusted to a lower value to avoid excessive impact on the power grid. Similarly, the distributed power source controller adjusts the power generation power of the distributed power source in real time according to the scheduling instructions. For example, the distributed power source controller will appropriately increase or decrease the power generation power of a certain photovoltaic power station or wind farm during the period from 10:00 to 10:15 according to the optimized scheduling scheme to maintain the power balance and stable operation of the power grid. Through the response of the intelligent charging pile and the distributed power source controller to the scheduling instructions, the real-time adjustment of the charging power of electric vehicles and the power generation power of the distributed power source is realized, so as to effectively implement the source-network-load optimization scheduling scheme and improve the operation efficiency and comprehensive benefits of the electric vehicle active distribution network.

[0056] Furthermore, the present invention also provides a source-network-load optimization scheduling analysis system for an electric vehicle active distribution network, which is used to execute the source-network-load optimization scheduling analysis method for the electric vehicle active distribution network as described above. The source-network-load optimization scheduling analysis system for the electric vehicle active distribution network includes: A source-network-load data acquisition module, which is used to collect the power source side data, grid side data, and load side data during the operation of the electric vehicle active distribution network in real time, and perform sliding filtering and standardization processing on the power source side data, grid side data, and load side data to obtain the standard source-network-load data of the electric vehicle. A load and power output prediction module, which is used to obtain historical charging data, traffic flow data, weather data, and user appearance habit data, and construct an electric vehicle load prediction model in combination with a gated recurrent unit to perform load prediction calculations on the standard source-network-load data of the electric vehicle to predict and output the charging power distribution of electric vehicles corresponding to different time periods and different regions; obtain historical power generation power data and numerical weather forecast data, and construct a distributed power source hybrid prediction model to perform power output prediction calculations on the standard source-network-load data of the electric vehicle to predict and output the distributed power source power generation power distribution corresponding to different time periods and different regions. A source-network-load scheduling model generation module, which is used to construct a corresponding multi-objective optimization function and distribution network operation constraint conditions through the electric vehicle active distribution network, and construct a source-network-load optimization scheduling model for the multi-objective optimization function based on the distribution network operation constraint conditions to generate a source-network-load collaborative optimization scheduling model. The source-grid-load optimization scheduling execution module is used to set the corresponding source-grid-load scheduling period, and read the corresponding prediction data from the electric vehicle charging power distribution and distributed power generation power distribution corresponding to different time periods and different regions based on the source-grid-load scheduling period; input the corresponding prediction data into the source-grid-load collaborative optimization scheduling model to solve the corresponding source-grid-load optimization scheduling plan, and adjust the corresponding electric vehicle charging power and distributed power generation power in real time according to the source-grid-load optimization scheduling plan.

[0057] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can 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 these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. An optimization scheduling analysis method for the source-network-load of an electric vehicle active distribution network, characterized in that It includes the following steps: Step S1: By collecting in real time the power source side data, grid side data, and load side data during the operation of the electric vehicle active distribution network, and performing sliding filtering and normalization processing on the power source side data, grid side data, and load side data, the standard data of the electric vehicle power source-grid-load is obtained; Step S2: Obtain historical charging data, traffic flow data, weather data, and user presence habit data, and combine a gated recurrent unit to construct an electric vehicle load prediction model to perform load prediction calculations on the standard data of the electric vehicle power source-grid-load, so as to predict and output the electric vehicle charging power distribution corresponding to different time periods and different regions; Obtain historical power generation power data and numerical weather forecast data, and construct a distributed power hybrid prediction model to perform power output prediction calculations on the standard data of the electric vehicle power source-grid-load, so as to predict and output the distributed power generation power distribution corresponding to different time periods and different regions; Step S3: Construct a corresponding multi-objective optimization function and distribution network operation constraint conditions through the electric vehicle active distribution network, and construct a source-grid-load optimization scheduling model for the multi-objective optimization function based on the distribution network operation constraint conditions to generate a source-grid-load collaborative optimization scheduling model; Step S4: By setting a corresponding source-grid-load scheduling period, and reading the corresponding prediction data from the electric vehicle charging power distribution and distributed power generation power distribution corresponding to different time periods and different regions based on the source-grid-load scheduling period; Input the corresponding prediction data into the source-grid-load collaborative optimization scheduling model to solve for the corresponding source-grid-load optimization scheduling plan, and adjust the corresponding electric vehicle charging power and distributed power generation power in real time according to the source-grid-load optimization scheduling plan.

2. The source-network-load optimal scheduling analysis method for an electric vehicle active distribution network according to claim 1, wherein Step S1 includes the following steps: Step S11: Collect in real time the power source side data during the operation of the electric vehicle active distribution network, including the power generation power, remaining capacity, and real-time operation status data of the distributed power source; Step S12: Collect in real time the grid side data during the operation of the electric vehicle active distribution network, including the voltage of each node of the distribution network, line current, network topology structure, and transformer capacity limit operation parameters; Step S13: Collect in real time the load side data during the operation of the electric vehicle active distribution network, including the real-time power demand of the traditional fixed load, electric vehicle location, remaining battery power, estimated driving time, and charging demand data; Step S14: Set the sliding window size at 3 - 5 time points, and perform sliding filtering processing on the power source side data, grid side data, and load side data based on this sliding window size to obtain the sliding filtered data of the electric vehicle power source-grid-load; Step S15: Perform interpolation filling and normalization processing on the sliding filtered data of the electric vehicle power source-grid-load to fill the corresponding missing data using median interpolation and eliminate the dimension between different format data to obtain the standard data of the electric vehicle power source-grid-load.

3. The source-network-load optimal scheduling analysis method for an electric vehicle active distribution network according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Obtain historical charging data, specifically the electric vehicle charging records within the past 1 - 2 years, including charging time, charging duration, and charging power; Step S22: Obtain traffic flow data, including the number of vehicle flows of electric vehicles at different times and in different regions; Step S23: Obtain weather data and user travel habit data, where the weather data includes temperature, humidity and illumination, and the user travel habit data includes daily travel time, routes and destinations collected through questionnaires or mobile applications; Step S24: Based on historical charging data, traffic flow data, weather data and user travel habit data, and combined with a gated recurrent unit, construct an electric vehicle load prediction model to perform load prediction calculations on the standard data of the electric vehicle source-network-load, so as to predict and output the electric vehicle charging power distribution corresponding to different times and different regions; Step S25: Obtain historical power generation power data and numerical weather forecast data, and construct a distributed power hybrid prediction model to perform power output prediction calculations on the standard data of the electric vehicle source-network-load, so as to predict and output the distributed power generation power distribution corresponding to different times and different regions.

4. The method for optimizing the scheduling analysis of the source-network-load of an electric vehicle integrated distribution network according to claim 3, wherein Step S24 includes the following steps: Step S241: Obtain the charging time characteristics of electric vehicles corresponding to hours, days, weeks, months and seasons through historical charging data; Step S242: Obtain the vehicle traffic characteristics of electric vehicles in different regions through traffic flow data, including vehicle traffic density and vehicle congestion index; Step S243: Obtain the weather influence characteristics corresponding to electric vehicles through weather data, including the change rates corresponding to temperature, humidity and illumination intensity; Step S244: Obtain the corresponding user travel characteristics through user travel habit data, including travel distance, travel purpose and charging frequency; obtain the potential influence characteristics of the source-network-load status on electric vehicle charging through the analysis of the standard data of the electric vehicle source-network-load; Step S245: Construct an electric vehicle load prediction model by selecting a gated recurrent unit GRU, and use the charging time characteristics, vehicle traffic characteristics, weather influence characteristics, user travel characteristics and the potential influence characteristics of the source-network-load status on electric vehicle charging as training data to train the electric vehicle load prediction model. By adjusting the hyperparameters corresponding to the model, including the number of neurons in the hidden layer, the learning rate and the number of iterations, and minimizing the prediction error, at the same time, input the standard data of the electric vehicle source-network-load into the trained electric vehicle load prediction model for load prediction calculations, so as to predict and output the electric vehicle charging power distribution corresponding to different times and different regions.

5. The method for optimizing the scheduling analysis of the source-network-load of an electric vehicle active distribution network according to claim 3, characterized in that Step S25 includes the following steps: Step S251: Obtain historical power generation power data, specifically the historical power generation power corresponding to distributed power sources and the corresponding timestamp information, where the distributed power sources include photovoltaic and wind power; Step S252: Obtain numerical weather forecast data, including solar radiation intensity, wind speed, wind direction, temperature meteorological parameters; Step S253: Obtain the output time characteristics of distributed power sources corresponding to hours, days, weeks, months and seasons through historical power generation power data; Step S254: Obtain the influence characteristics of solar radiation intensity, wind speed, wind direction, and temperature on the power generation capacity corresponding to the distributed power generation output through numerical weather prediction data; obtain the potential influence characteristics of the source-network-load state on the distributed power generation output through the analysis of electric vehicle source-network-load standard data; Step S255: Construct a distributed power generation hybrid prediction model by selecting the corresponding deep learning algorithm. For photovoltaic power generation, a hybrid prediction model combining a convolutional neural network and a long short-term memory network is used. The convolutional neural network is used to extract the spatial features in the meteorological data, and the long short-term memory network is used to capture the time series features corresponding to the power generation power. For wind power generation, the time series of meteorological data is decomposed based on wavelet transform to extract the features of different frequency components, and a hybrid prediction model is constructed in combination with a support vector machine. At the same time, the output time features, power generation capacity influence features, and potential influence features of the source-network-load state on the distributed power generation output are used as training data to train the distributed power generation hybrid prediction model. By adjusting the hyperparameters corresponding to the model, including the convolutional kernel size, stride, number of hidden layer neurons, and penalty parameter, and minimizing the prediction error, and inputting the electric vehicle source-network-load standard data into the trained distributed power generation hybrid prediction model for power generation output prediction calculation to predict the distributed power generation power distribution corresponding to different time periods and different regions.

6. The method for optimizing the scheduling analysis of the source-network-load of an electric vehicle active distribution network according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Construct a corresponding multi-objective optimization function through the electric vehicle active distribution network, including economic objectives, environmental protection objectives, and grid security objectives; Step S32: Construct corresponding distribution network operation constraint conditions through the electric vehicle active distribution network considering the physical constraints of the distribution network operation and the operation characteristics constraints of electric vehicles, including grid operation constraints, distributed power generation constraints, and electric vehicle constraints; Step S33: Based on the distribution network operation constraint conditions, add the corresponding grid operation constraints, distributed power generation constraints, and electric vehicle constraints one by one to construct a source-network-load optimal scheduling model for the multi-objective optimization function, so as to generate a source-network-load coordinated optimal scheduling model.

7. The source-network-load optimal scheduling analysis method for an electric vehicle integrated with an active distribution network according to claim 6, characterized in that The economic objective described in Step S31 is specifically to minimize the distribution network operation cost, including power purchase cost, power generation cost corresponding to distributed power generation, network loss cost, and charging compensation cost corresponding to electric vehicles: ； in, represents the cost of purchasing electricity, represents the total number of distributed generation sources, Indicates The power generation cost corresponding to each distributed power source is: represents the network loss cost, represents the total number of electric vehicles, Indicates Charging compensation cost corresponding to each electric vehicle; The environmental protection objective is specifically to maximize the consumption of distributed power generation; ； in, Indicates The power generation corresponding to each distributed power source is: represents the total amount of traditional fire load, Indicates The power generation corresponding to the traditional thermal load; The grid security objective is specifically to minimize the node voltage deviation and line overload risk; ； Among them, represents the total number of nodes in the active distribution network of electric vehicles, represents the th voltage deviation corresponding to the node, represents the total number of lines in the active distribution network of electric vehicles, represents the th current corresponding to the line, represents the th rated current corresponding to the line.

8. The method for optimizing and scheduling analysis of source-network-load of an electric vehicle integrated with an active distribution network according to claim 7, characterized in that, The grid operation constraints described in Step S32 include: Node voltage magnitude constraint: , where and are respectively the lower limit value and the upper limit value of the voltage deviation corresponding to the th node; Line current upper limit constraint: , to ensure that the line current does not exceed its rated current; Active power balance constraint: , where represents the set of distributed power sources, represents the th active power corresponding to the distributed power source, represents the set of traditional thermal loads, represents the th active power corresponding to the traditional thermal load, represents the set of electric vehicles, represents the th active power corresponding to the electric vehicle, represents the active power loss of the network; Reactive power balance constraint: , where represents the reactive power corresponding to the th distributed power source, represents the reactive power corresponding to the th traditional thermal load, represents the reactive power corresponding to the th electric vehicle, represents the reactive power of network loss; The distributed power generation constraints include: Power generation power upper and lower limit constraints: , where and respectively represent the minimum power generation and the maximum power generation corresponding to the th distributed power source; Ramp rate constraint: , where represents the time parameter, represents the maximum ramp-up power generation rate corresponding to the The electric vehicle constraints include: Active power upper and lower limit constraints: , where and respectively represent the minimum active power and the maximum active power corresponding to the th electric vehicle; Reactive power upper and lower limit constraints: , where and respectively represent the minimum reactive power and the maximum reactive power corresponding to the th electric vehicle.

9. The source-network-load optimal scheduling analysis method for an electric vehicle active distribution network according to claim 1, characterized in that Step S4 includes the following steps: Step S41: Set a corresponding source-network-load scheduling period, and the source-network-load scheduling period is specifically 15 minutes; Step S42: Based on the source-network-load scheduling period, read the prediction data corresponding to each source-network-load scheduling period from the electric vehicle charging power distribution and distributed power generation power distribution corresponding to different time periods and different regions, including the charging power distribution and power generation power distribution within the corresponding source-network-load scheduling period; Step S43: Input the corresponding prediction data into the source-grid-load collaborative optimization scheduling model to solve and obtain the corresponding source-grid-load optimized scheduling plan for the source-grid-load scheduling period; Step S44: Apply the source-grid-load optimized scheduling plan to the intelligent charging piles and distributed power source controllers to execute the corresponding scheduling instructions, and respond to the scheduling instructions to adjust the charging power of the corresponding electric vehicles and the power generation power of the distributed power sources in real time.

10. An active distribution network source-network-load optimal scheduling analysis system for electric vehicles, characterized in that, For implementing the source-grid-load optimized scheduling analysis method of the electric vehicle active distribution network as described in claim 1, the source-grid-load optimized scheduling analysis system of the electric vehicle active distribution network includes: A source-grid-load data acquisition module, configured to collect power source side data, grid side data, and load side data during the operation of the electric vehicle active distribution network in real time, and perform sliding filtering and normalization processing on the power source side data, grid side data, and load side data, so as to obtain the standard source-grid-load data of the electric vehicle; A load and power output prediction module, configured to obtain historical charging data, traffic flow data, weather data, and user appearance habit data, and combine a gated recurrent unit to construct an electric vehicle load prediction model to perform load prediction calculations on the standard source-grid-load data of the electric vehicle, so as to predict and output the charging power distribution of electric vehicles corresponding to different time periods and different regions; obtain historical power generation power data and numerical weather forecast data, and construct a distributed power source hybrid prediction model to perform power output prediction calculations on the standard source-grid-load data of the electric vehicle, so as to predict and output the distributed power source power generation power distribution corresponding to different time periods and different regions; A source-grid-load scheduling model generation module, configured to construct a corresponding multi-objective optimization function and distribution network operation constraint conditions through the electric vehicle active distribution network, and construct a source-grid-load optimized scheduling model for the multi-objective optimization function based on the distribution network operation constraint conditions, so as to generate a source-grid-load collaborative optimization scheduling model; A source-grid-load optimized scheduling execution module, configured to set a corresponding source-grid-load scheduling period, and read the corresponding prediction data from the charging power distribution of electric vehicles and the distributed power source power generation power distribution corresponding to different time periods and different regions based on the source-grid-load scheduling period; input the corresponding prediction data into the source-grid-load collaborative optimization scheduling model to solve and obtain the corresponding source-grid-load optimized scheduling plan, and adjust the charging power of the corresponding electric vehicles and the distributed power source power generation power in real time according to the source-grid-load optimized scheduling plan.

Images