Electric vehicle charging load prediction method and system

Abstract

The present disclosure discloses a method and system for predicting electric vehicle charging load, including: obtaining the active power output by each power supply in the system; inputting the active power output by each power supply into a constructed electric vehicle charging load prediction model to obtain the connectable electric vehicle charging load. The electric vehicle charging load prediction model is constructed based on the constraints of minimizing the combined initial investment, maintenance, and operating costs of each power supply, system power balance, upper and lower limits on the active power output by each power supply, and the upper limit of the charging power of all electric vehicles, with the goal of minimizing equivalent load fluctuation after connecting an electric vehicle. This ensures that the equivalent load fluctuation is minimized when an electric vehicle is connected.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid planning, and in particular to a method and system for predicting charging load of an electric vehicle. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] With the popularization of electric vehicles in the future, a large number of electric vehicles will be connected to the power grid for charging and discharging. As the number of electric vehicles increases significantly, due to the randomness of their own charging behavior, large-scale access of electric vehicles will have a significant impact on the operation and control of the power system. How the power grid can accept the large-scale connection of electric vehicles without exceeding the capacity of the power grid has become a major issue that must be solved in the future development and popularization of electric vehicles.

[0004] The inventors found that the existing planning for electric vehicle grid connection does not take into account the access of distributed power sources and the fluctuation of equivalent load in the power grid. As a result, after the electric vehicle is connected to the grid, the equivalent load may fluctuate significantly, thereby affecting the operation of conventional load equipment. Summary of the Invention

[0005] In order to solve the above problems, the present disclosure proposes a method and system for predicting electric vehicle charging load. When predicting the accessible electric vehicle charging load, the equivalent load fluctuation is fully considered, so that after the accessible electric vehicle charging load is added to the power grid, the fluctuation of the equivalent load is minimized, thereby ensuring the normal operation of conventional load equipment.

[0006] To achieve the above objectives, the present disclosure adopts the following technical solutions:

[0007] First, a method for predicting electric vehicle charging load is proposed, including:

[0008] Obtain the active power output by each power supply in the system;

[0009] The active power output by each power source is input into the constructed electric vehicle charging load prediction model to obtain the accessible electric vehicle charging load;

[0010] Among them, the electric vehicle charging load prediction model is based on the constraints of minimizing the comprehensive cost of initial investment, maintenance and operation of each power source, system power balance, the upper and lower limits of the active power output of each power source, and the upper limit of the charging power of all electric vehicles. It is constructed with the goal of minimizing the equivalent load fluctuation after connecting electric vehicles.

[0011] Secondly, an electric vehicle charging load forecasting system is proposed, including:

[0012] A data acquisition module is used to obtain the active power output by each power supply in the system;

[0013] The charging load prediction module is used to input the active power output by each power supply into the constructed electric vehicle charging load prediction model to obtain the accessible electric vehicle charging load;

[0014] Among them, the electric vehicle charging load prediction model is based on the constraints of minimizing the comprehensive cost of initial investment, maintenance and operation of each power source, system power balance, the upper and lower limits of the active power output of each power source, and the upper limit of the charging power of all electric vehicles. It is constructed with the goal of minimizing the equivalent load fluctuation after connecting electric vehicles.

[0015] In a third aspect, an electronic device is proposed, comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps described in a method for predicting charging load of an electric vehicle are completed.

[0016] In a fourth aspect, a computer-readable storage medium is proposed for storing computer instructions. When the computer instructions are executed by a processor, the steps of a method for predicting charging load of an electric vehicle are completed.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] 1. This disclosure uses the minimum fluctuation of equivalent load after connecting electric vehicles as a constraint and the minimum comprehensive cost of initial investment, maintenance and operation of the system as a goal to predict the charging load of electric vehicles that can be connected. This ensures that after the charging load of electric vehicles is connected to the power grid, the equivalent load fluctuation in the power grid is minimized, thereby ensuring the normal use of conventional load equipment in the power grid.

[0019] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings in the specification, which constitute a part of this application, are used to provide further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute improper limitations on this application.

[0021] Figure 1 This is a flow chart of the method disclosed in Example 1 of the present disclosure;

[0022] Figure 2 This is a structural diagram of the deep neural network disclosed in Example 1 of the present disclosure;

[0023] Figure 3 This is the equivalent load prediction flow chart disclosed in Example 1 of the present disclosure. DETAILED DESCRIPTION

[0024] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0025] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present application belongs.

[0026] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0027] Example 1

[0028] In this embodiment, a method for predicting charging load of an electric vehicle is disclosed, comprising:

[0029] Obtain the active power output of each power supply in the system;

[0030] The active power output by each power source is input into the constructed electric vehicle charging load prediction model to obtain the accessible electric vehicle charging load;

[0031] Among them, the electric vehicle charging load prediction model is based on the constraints of minimizing the comprehensive cost of initial investment, maintenance and operation of each power source, system power balance, the upper and lower limits of the active power output of each power source, and the upper limit of the charging power of all electric vehicles. It is constructed with the goal of minimizing the equivalent load fluctuation after connecting electric vehicles.

[0032] Furthermore, the power supply in the system includes conventional units and distributed power supply.

[0033] Furthermore, the maximum charging power of electric vehicles is determined based on the maximum charging power of the charging and swapping stations that the busbar in the area can access and the charging power of the charging piles in the area when charging the maximum number of electric vehicles at the same time.

[0034] Furthermore, the number of charging piles is determined based on the area where the charging piles are located, the construction indicators of the building type, and the building area.

[0035] Furthermore, the specific process of obtaining the accessible electric vehicle charging load is as follows:

[0036] The active power output by each power source is input into the constructed electric vehicle charging load prediction model to obtain the allowable equivalent load;

[0037] The allowed equivalent load is subtracted from the equivalent load forecast value during the electric vehicle access time period to obtain the accessible electric vehicle charging load.

[0038] Furthermore, the average temperature, average humidity, number of holidays, GDP growth rate, industrial structure, electricity consumption structure and saturated load density of the electric vehicle access period are obtained;

[0039] The average temperature, average humidity, number of holidays, GDP growth rate, industrial structure, electricity consumption structure and saturated load density of the electric vehicle access period are input into the trained equivalent load prediction model to obtain the equivalent load prediction value when the electric vehicle is charging.

[0040] Furthermore, the equivalent load forecasting model adopts deep neural network.

[0041] A method for predicting charging load of an electric vehicle disclosed in this embodiment is described in detail.

[0042] like Figure 1 As shown, a method for predicting charging load of an electric vehicle includes:

[0043] S1: Get the active power output by each power supply in the system.

[0044] In specific implementation, the power sources in the power grid system include conventional generator sets and distributed power sources, and the distributed power sources also include photovoltaic power generation, wind power generation and gas turbine power generation.

[0045] Among them, the power generation power of the conventional generator set is G 机组 Indicates that the power generated by distributed power is expressed in G 分布式电源 express.

[0046] Among them, the photovoltaic power generation power is obtained by obtaining the sunshine duration and sunlight radiation intensity of the area to be planned;

[0047] By obtaining the wind speed in the area to be planned, the power of wind power generation is obtained.

[0048] S2: Input the active power output by each power source into the constructed electric vehicle charging load prediction model to obtain the accessible electric vehicle charging load;

[0049] Among them, the electric vehicle charging load prediction model is based on the constraints of minimizing the comprehensive cost of initial investment, maintenance and operation of each power source, system power balance, the upper and lower limits of the active power output of each power source, and the upper limit of the charging power of all electric vehicles. It is constructed with the goal of minimizing the equivalent load fluctuation after connecting electric vehicles.

[0050] In the specific implementation, the constructed electric vehicle charging load prediction model is:

[0051]

[0052] The constraints are:

[0053]

[0054] G 分布式电源 +G 机组 =P eq

[0055] G 机组 =P 常规负荷

[0056]

[0057]

[0058]

[0059]

[0060] Where, f(p) is the system load fluctuation, is the active power of the conventional generator set. is the power generation cost of each conventional generator set power source i, Active power generated by distributed power sources, including photovoltaic, wind, gas turbines, etc. The active load of the power grid includes conventional load, electric vehicle charging, energy storage and other demand-side response loads. and They are the maximum and minimum active power of conventional unit power supply, P eq (i) is the equivalent load at time i (MW); P av is the average value of equivalent load (MW), and They are the maximum active power of regional photovoltaic power and the maximum charging power of regional electric vehicles.

[0061] Among them, the power balance equation of the system can also be written as:

[0062] G 分布式电源 +G 机组 =L 常规负荷 +L 充电负荷 +L 其他柔性负荷

[0063] When the unit transformer capacity is determined, without considering the load reduction effect of distributed power sources, the size of the electric vehicle charging load that can be connected is determined by dispatching the electric vehicle load. The maximum charging power of the electric vehicle is determined by the maximum charging power Pmax of the charging and swapping station that the busbar in the area can access and the charging power Pmax of the maximum number of charging piles in the area when electric vehicles are charging at the same time. EV Sure.

[0064] Among them, the number of charging piles is determined based on the area where the charging piles are located, the construction indicators of the building type and the building area.

[0065] Assume that the connected electric vehicle charging load P EV If all of them participate in the grid regulation as interruptible loads, then all electric vehicles are charged during the off-peak period, and the peak load P eq =P 常规负荷 The size of electric vehicle access load is determined by the planned number of charging piles in the region, transformer and line capacity, load rate, and power factor. Specifically:

[0066]

[0067] P EV =n*P 充 *Simultaneous rate

[0068] P max =βS N cosθ / k s -P H

[0069] P max is the maximum charging power (kW) of the charging and swapping station that can be connected to the 10kV bus; β is the load rate of the transformer; S N is the transformer capacity (kVA); P H k is the active power of the normal load carried by the transformer (kW); s is the simultaneous coefficient of the user's power load; cosθ is the power factor.

[0070] P EV =n*P 充 *In the simultaneous rate, n is the number of charging piles in the area, which is limited by the regional construction area, and P charging is the charging power of a single electric vehicle.

[0071] The number of charging piles required in a region is related to the construction area. The number of parking spaces N for a construction project is calculated based on the project's location, building type, and construction area. According to the "Interim Provisions on the Planning and Construction of Charging Facilities in Workplaces, Residential Areas, and Parking Lots," the proportion of electric vehicle charging facilities in parking lots must be at least 10%, or a reserved proportion. Therefore, the planned number of charging piles in a region is n = N * 10%.

[0072] Referring to the special plans for electric vehicle charging infrastructure in other provinces, the simultaneous rate of electric vehicle charging piles can be taken as 0.7.

[0073] Without increasing the transformer capacity, the maximum connected electric vehicle charging load can be determined The maximum charging power of the connected electric vehicles.

[0074] The electric vehicle charging load prediction model is constructed by inputting the active power output of each power source. When predicting the electric vehicle charging load, it is also necessary to predict the equivalent load of the power grid when the electric vehicle is connected to obtain the equivalent load prediction value.

[0075] The process of obtaining the accessible electric vehicle charging load is as follows:

[0076] The active power output by each power source is input into the constructed electric vehicle charging load prediction model to obtain the allowable equivalent load;

[0077] The allowable equivalent load P eq (i) Subtract the equivalent load forecast value from the electric vehicle access time period to obtain the charging load of the electric vehicle that can be connected.

[0078] The specific process of obtaining the grid equivalent load forecast value during the electric vehicle access period is as follows:

[0079] Obtain the average temperature, average humidity, number of holidays, GDP growth rate, industrial structure, electricity consumption structure and saturated load density of the electric vehicle access period;

[0080] The average temperature, average humidity, number of holidays, GDP growth rate, industrial structure, electricity consumption structure and saturated load density of the electric vehicle access period are input into the trained equivalent load prediction model to obtain the equivalent load prediction value when the electric vehicle is charging.

[0081] The equivalent load prediction model of this embodiment adopts a deep neural network, the structure of which is as follows: Figure 2 shown.

[0082] Deep Neural Networks (DNNs) are multilayer perceptrons with multiple hidden layers. A typical DNN model consists of an input layer, multiple intermediate layers, and an output layer. DNN models employ a hybrid training approach. During layer-by-layer training, a temporary output layer is stacked on top of the hidden layer to be trained. This layer is then trained using a hybrid training method that combines unsupervised training (to reconstruct the input) with supervised training (to reduce prediction error). The updated hidden layer weights from unsupervised training are summed with the updated values ​​from supervised training. Unsupervised training uses the radial basis function, while supervised training uses the sigmoid function.

[0083] The learning process of deep neural networks utilizes forward and backward propagation of signals to achieve training. During forward propagation, input information passes through the input layer, is trained in the hidden layer, and then propagated to the output layer. Generally, the predicted result is not achieved, so the error is reversely propagated through the output layer, re-entering the hidden layer for learning. Through repeated forward and backward iterations, when the predicted result reaches the expected value, it is output, completing the training.

[0084] For a DNN network, assuming its input is X, the activation value calculation of its hidden layer and output layer can be expressed by the following formula:

[0085]

[0086] v k (x)=∑w kj u j (x)+w k0

[0087]

[0088] Where u j (x) represents the output of the jth node in the hidden layer, m j represents the center of the jth hidden layer node. ||·|| represents the Euclidean norm, σ j Represents the Gauss distribution width of node j.

[0089] where w kj and w k0 Represent the weight and bias vector of the kth layer of the network, y k (x) represents the output of the network. The forward propagation algorithm of DNN uses several weight coefficient matrices W and bias vectors w k0To perform a series of nonlinear operations and activation operations on the input value vector x, starting from the input layer, and calculating backward layer by layer until the operation reaches the output layer, the output result is value.

[0090] In the neural network input, due to the different units of each input, the order of magnitude varies greatly. If direct input is used, the neuron training will be saturated. Therefore, before input training, the data must be normalized to make it at the same order of magnitude to accelerate the convergence of the neural network. Finally, the real value is obtained through denormalization. The normalization method used in this embodiment normalizes the data to [0, 1], and the formula is as follows: Denormalization: x i =(x max -x min )y i +x min Where x max 、x min is the maximum and minimum value of the training sample input, x i 、y i are the values ​​of the input samples before and after normalization.

[0091] Many factors influence power load, generally categorized as economic, temporal, climatic, and random. Weather and holidays are generally the most influential factors on short-term load forecasting, while medium- and long-term load forecasting primarily considers socioeconomic development, population changes, and geographic and climatic variations.

[0092] The monthly load of the power grid exhibits distinct seasonal characteristics. November, December, and January are winter months, with generally low loads. In late December and early January, due to the Spring Festival holiday, industrial load drops significantly, bringing the total load to its lowest point. In July and August, summer months, air conditioning loads increase significantly due to the hot weather, accounting for a significant portion of the load, leading to a significant increase in the total power grid load, reaching its annual peak. In March, April, May, October, and November, the grid load is less affected by meteorological factors and remains relatively stable. Therefore, the main factors influencing monthly load are weather, climate, historical load, economic factors, industrial structure, and electricity consumption structure. The corresponding factor indicators are: average temperature, average humidity, number of holidays, GDP growth rate, industrial structure, electricity consumption structure, and saturated load density.

[0093] The training process of deep neural networks, such as Figure 3 As shown,

[0094] Obtain historical sample data for the time period to be predicted;

[0095] The acquired historical sample data is input into the constructed deep neural network for partial supervised training to obtain a trained deep neural network.

[0096] The historical sample data obtained include: average temperature, average humidity, number of holidays, regional GDP growth rate, industrial structure, electricity consumption structure, saturated load density and corresponding equivalent load data for the forecast period.

[0097] This embodiment discloses a method for predicting the charging load of an electric vehicle. When predicting the charging load of an electric vehicle, the method fully considers the equivalent load fluctuation of the power grid. With the goal of minimizing the equivalent load fluctuation after the electric vehicle is connected, the method obtains the accessible charging load of the electric vehicle. When the accessible charging load of the electric vehicle is connected to the power grid, the equivalent load fluctuation is minimized, thereby not affecting the normal use of conventional load equipment in the power grid.

[0098] In addition, when predicting the equivalent load when electric vehicles are connected to the power grid, the present invention fully considers the average temperature, average humidity, regional GDP growth rate, industrial structure, electricity consumption structure and saturated load density during the electric vehicle connection period, so that the equivalent load of the power grid can be more reasonably predicted according to the different electricity demand in different seasons, ensuring the accuracy of the equivalent prediction results. On this basis, the charging load of electric vehicles can be accurately predicted.

[0099] Example 2

[0100] In this embodiment, a system for predicting charging load of an electric vehicle is disclosed, comprising:

[0101] A data acquisition module is used to obtain the active power output by each power supply in the system;

[0102] The charging load prediction module is used to input the active power output by each power supply into the constructed electric vehicle charging load prediction model to obtain the accessible electric vehicle charging load;

[0103] Among them, the electric vehicle charging load prediction model is based on the constraints of minimizing the comprehensive cost of initial investment, maintenance and operation of each power source, system power balance, the upper and lower limits of the active power output of each power source, and the upper limit of the charging power of all electric vehicles. It is constructed with the goal of minimizing the equivalent load fluctuation after connecting electric vehicles.

[0104] Example 3

[0105] In this embodiment, an electronic device is disclosed, including a memory and a processor, and computer instructions stored in the memory and running on the processor. When the computer instructions are executed by the processor, the steps described in the electric vehicle charging load prediction method disclosed in Example 1 are completed.

[0106] Example 4

[0107] In this embodiment, a computer-readable storage medium is disclosed for storing computer instructions. When the computer instructions are executed by a processor, the steps of the electric vehicle charging load prediction method disclosed in Example 1 are completed.

[0108] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0109] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0110] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0111] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A method for predicting electric vehicle charging load, characterized in that: include: Obtain the active power output of each power supply in the system; The active power output by each power source is input into the constructed electric vehicle charging load prediction model to obtain the accessible electric vehicle charging load; The electric vehicle charging load forecasting model is built with the constraints of minimizing the combined costs of initial investment, maintenance, and operation of each power source, system power balance, the upper and lower limits of each power source's output active power, and the upper limit of all electric vehicle charging power, with the goal of minimizing equivalent load fluctuations after connecting electric vehicles. The constructed electric vehicle charging load prediction model is: The constraints are: G 分布式电源 +G 机组 =P eq G 机组 =P 常规负荷 Where, f(p) is the system load fluctuation, is the active power of the conventional generator set. is the power generation cost of each conventional generator set power source i, and They are the maximum and minimum active power of conventional unit power supply, P eq (i) is the equivalent load at time i; P av is the average value of the equivalent load, and They are the maximum active power of regional photovoltaic power and the maximum charging power of regional electric vehicles; The power balance equation of the system is G 分布式电源 +G 机组 =L 常规负荷 +L 充电负荷 +L 其他柔性负荷 ; When the unit transformer capacity is determined, without considering the load reduction effect of distributed power sources, the size of the electric vehicle charging load that can be connected is determined by dispatching the electric vehicle load. The maximum charging power of the electric vehicle is determined by the maximum charging power Pmax of the charging and swapping station that the busbar in the area can access and the charging power Pmax of the maximum number of charging piles in the area when electric vehicles are charging at the same time. EV Sure; The number of charging piles is determined based on the area where the charging piles are located, the construction indicators of the building type, and the building area; When the connected electric vehicle charging load P EV If all of them participate in the grid regulation as interruptible loads, then all electric vehicles are charged during the off-peak period, and the peak load P eq =P 常规负荷 The size of the electric vehicle access load is determined by the planned number of charging piles in the region, the capacity of the transformer and line, the load rate, and the power factor. Specifically: P EV =n*P 充 *Simultaneous rate P max =βS N cosθ / k s -P H P max is the maximum charging power of the charging and swapping station that can be connected to the 10kV bus; β is the load rate of the transformer; S N is the transformer capacity; P H is the active power of the normal load carried by the transformer; k s is the simultaneous coefficient of the user's electricity load; cosθ is the power factor; n is the number of charging piles in the area, P 充 is the charging power of a single electric vehicle; The maximum charging power of the connected electric vehicle; The obtained active power output of each power source is input into the electric vehicle charging load prediction model. When predicting the electric vehicle charging load, it is also necessary to predict the equivalent load of the power grid when the electric vehicle is connected to obtain the equivalent load prediction value; The specific process of obtaining the accessible electric vehicle charging load is as follows: The active power output by each power source is input into the constructed electric vehicle charging load prediction model to obtain the allowable equivalent load; The allowable equivalent load P eq (i) Subtract the equivalent load forecast value from the electric vehicle access time period to obtain the accessible electric vehicle charging load; The specific process of obtaining the equivalent load forecast value during the electric vehicle access time period is as follows: Obtain the average temperature, average humidity, number of holidays, GDP growth rate, industrial structure, electricity consumption structure and saturated load density of the electric vehicle access period; The average temperature, average humidity, number of holidays, GDP growth rate, industrial structure, electricity consumption structure and saturated load density of the electric vehicle access period are input into the trained equivalent load prediction model to obtain the equivalent load prediction value when the electric vehicle is charging; The equivalent load forecasting model uses a deep neural network, which contains a multi-layer perceptron with multiple hidden layers. The deep neural network model uses a hybrid training method. During the layer-by-layer training process, a temporary output layer is stacked on top of the hidden layer to be trained. This layer is then trained using a hybrid training method that combines unsupervised training with supervised training. The updated values ​​of the hidden layer weights from the unsupervised training process are added to the updated values ​​from the supervised training process as the updated weight values. Unsupervised training uses the radial basis function, while supervised training uses the sigmoid function. The learning process of deep neural networks uses forward and backward propagation of signals to achieve learning and training. In the forward propagation learning process, the input information passes through the input layer, is trained in the hidden layer, and then propagated to the output layer. In the output layer, the error change value is reversely propagated and re-enters the hidden layer for learning. Through repeated forward and backward iterations, when the prediction result reaches the expected value, its result is output, and finally the learning and training is completed. Deep neural network, input is X, the activation value calculation formula of the hidden layer and output layer is: v k (x)=∑w kj u j (x)+w k0 Where u j (x) represents the output of the jth node in the hidden layer, m j represents the center of the jth hidden layer node, ||·|| represents the Euclidean norm, σ j represents the Gauss distribution width of node j; w kj and w k0 Represent the weight and bias vector of the kth layer of the network, y k (x) represents the output of the network; the forward propagation algorithm of the deep neural network uses several weight coefficient matrices W and bias vectors w k0 To perform nonlinear operations and activation operations on the input value vector x, starting from the input layer, and calculating backward layer by layer until the operation reaches the output layer and the output result is obtained; Before inputting the training data, normalize them to the same level to speed up the convergence of the neural network. Finally, denormalize them to get the real value. Specifically, the normalization method normalizes the data to [0,1]. The formula is: Denormalization is: x i =(x max -x min )y i +x min Among them, x max 、x min is the maximum and minimum value of the training sample input, x i 、y i is the value of the input sample before and after normalization; The training process of a deep neural network is: Obtain historical sample data for the time period to be predicted; Input the acquired historical sample data into the constructed deep neural network, perform partial supervised training, and obtain a trained deep neural network; The historical sample data obtained include: average temperature, average humidity, number of holidays, regional GDP growth rate, industrial structure, electricity consumption structure, saturated load density and corresponding equivalent load data for the forecast period.

2. The method for predicting charging load of an electric vehicle according to claim 1, wherein: The power sources in the system include conventional units and distributed power sources.

3. An electric vehicle charging load prediction system, characterized in that: include: A data acquisition module is used to obtain the active power output by each power supply in the system; The charging load prediction module is used to input the active power output by each power supply into the constructed electric vehicle charging load prediction model to obtain the accessible electric vehicle charging load; The electric vehicle charging load forecasting model is built with the constraints of minimizing the combined costs of initial investment, maintenance, and operation of each power source, system power balance, the upper and lower limits of each power source's output active power, and the upper limit of all electric vehicle charging power, with the goal of minimizing equivalent load fluctuations after connecting electric vehicles. The constructed electric vehicle charging load prediction model is: The constraints are: G 分布式电源 +G 机组 =P eq G 机组 =P 常规负荷 Where, f(p) is the system load fluctuation, is the active power of the conventional generator set. is the power generation cost of each conventional generator set power source i, and They are the maximum and minimum active power of conventional unit power supply, P eq (i) is the equivalent load at time i; P av is the average value of the equivalent load, and They are the maximum active power of regional photovoltaic power and the maximum charging power of regional electric vehicles; The power balance equation of the system is G 分布式电源 +G 机组 =L 常规负荷 +L 充电负荷 +L 其他柔性负荷 ; When the unit transformer capacity is determined, without considering the load reduction effect of distributed power sources, the size of the electric vehicle charging load that can be connected is determined by dispatching the electric vehicle load. The maximum charging power of the electric vehicle is determined by the maximum charging power Pmax of the charging and swapping station that the busbar in the area can access and the charging power Pmax of the maximum number of charging piles in the area when electric vehicles are charging at the same time. EV Sure; The number of charging piles is determined based on the area where the charging piles are located, the construction indicators of the building type, and the building area; When the connected electric vehicle charging load P EV If all of them participate in the grid regulation as interruptible loads, then all electric vehicles are charged during the off-peak period, and the peak load P eq =P 常规负荷 The size of the electric vehicle access load is determined by the planned number of charging piles in the region, the capacity of the transformer and line, the load rate, and the power factor. Specifically: P EV =n*P 充 *Simultaneous rate P max =βS N cosθ / k s -P H P max is the maximum charging power of the charging and swapping station that can be connected to the 10kV bus; β is the load rate of the transformer; S N is the transformer capacity; P H is the active power of the normal load carried by the transformer; k s is the simultaneous coefficient of the user's electricity load; cosθ is the power factor; n is the number of charging piles in the area, P 充 is the charging power of a single electric vehicle; The maximum charging power of the connected electric vehicle; The obtained active power output of each power source is input into the electric vehicle charging load prediction model. When predicting the electric vehicle charging load, it is also necessary to predict the equivalent load of the power grid when the electric vehicle is connected to obtain the equivalent load prediction value; The specific process of obtaining the accessible electric vehicle charging load is as follows: The active power output by each power source is input into the constructed electric vehicle charging load prediction model to obtain the allowable equivalent load; The allowed equivalent load is subtracted from the equivalent load forecast value during the electric vehicle access time period to obtain the accessible electric vehicle charging load; The specific process of obtaining the equivalent load forecast value during the electric vehicle access time period is as follows: Obtain the average temperature, average humidity, number of holidays, GDP growth rate, industrial structure, electricity consumption structure and saturated load density of the electric vehicle access period; The average temperature, average humidity, number of holidays, GDP growth rate, industrial structure, electricity consumption structure and saturated load density of the electric vehicle access period are input into the trained equivalent load prediction model to obtain the equivalent load prediction value when the electric vehicle is charging; The equivalent load forecasting model uses a deep neural network, which contains a multi-layer perceptron with multiple hidden layers. The deep neural network model uses a hybrid training method. During the layer-by-layer training process, a temporary output layer is stacked on top of the hidden layer to be trained. This layer is then trained using a hybrid training method that combines unsupervised training with supervised training. The updated values ​​of the hidden layer weights from the unsupervised training process are added to the updated values ​​from the supervised training process as the updated weight values. Unsupervised training uses the radial basis function, while supervised training uses the sigmoid function. The learning process of deep neural networks uses forward and backward propagation of signals to achieve learning and training. In the forward propagation learning process, the input information passes through the input layer, is trained in the hidden layer, and then propagated to the output layer. In the output layer, the error change value is reversely propagated and re-enters the hidden layer for learning. Through repeated forward and backward iterations, when the prediction result reaches the expected value, its result is output, and finally the learning and training is completed. Deep neural network, input is X, the activation value calculation formula of the hidden layer and output layer is: v k (x)=∑w kj u j (x)+w k0 Where u j (x) represents the output of the jth node in the hidden layer, m j represents the center of the jth hidden layer node, ||·|| represents the Euclidean norm, σ j represents the Gauss distribution width of node j; w kj and w k0 Represent the weight and bias vector of the kth layer of the network, y k (x) represents the output of the network; the forward propagation algorithm of the deep neural network uses several weight coefficient matrices W and bias vectors w k0 To perform nonlinear operations and activation operations on the input value vector x, starting from the input layer, and calculating backward layer by layer until the operation reaches the output layer and the output result is obtained; Before inputting the training data, normalize them to the same level to speed up the convergence of the neural network. Finally, denormalize them to get the real value. Specifically, the normalization method normalizes the data to [0,1]. The formula is: Denormalization is: x i =(x max -x min )y i +x min Among them, x max 、x min is the maximum and minimum value of the training sample input, x i 、y i is the value of the input sample before and after normalization; The training process of a deep neural network is: Obtain historical sample data for the time period to be predicted; Input the acquired historical sample data into the constructed deep neural network, perform partial supervised training, and obtain a trained deep neural network; The historical sample data obtained include: average temperature, average humidity, number of holidays, regional GDP growth rate, industrial structure, electricity consumption structure, saturated load density and corresponding equivalent load data for the forecast period.

4. An electronic device, characterized in that: The method comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps of the electric vehicle charging load prediction method according to any one of claims 1 to 2 are completed.

5. A computer-readable storage medium, characterized in that Used to store computer instructions, when the computer instructions are executed by a processor, the steps of the electric vehicle charging load prediction method according to any one of claims 1-2 are completed.

Images