Electric energy peak regulation method based on V2G technology and machine learning

Abstract

The invention provides an electric energy peak regulation method based on a V2G technology and machine learning, and the method comprises the steps: carrying out the clustering through combining vehicle charging and discharging related operation data with digital map information, and carrying out the calculation to obtain charging demands and reverse power supply capacities corresponding to different sub-regions of a power distribution network; the power grid load fluctuation caused by the charging and discharging behaviors of the individual vehicles with relatively strong randomness can be relatively easily processed; then a deep neural network model is established and trained by using the demand and the power supply capacity of each sub-region based on the thought of time sequence prediction, and the load of each sub-region in a subsequent short-term continuous period can be predicted, so that a peak regulation decision can be made timely and efficiently according to a prediction result; the source-load interaction level is improved from the vehicle network level, and the positive effect of the V2G technology is fully played.

Description

Technical Field

[0001] The present invention belongs to the field of V2G technology, and in particular relates to an electric energy allocation method based on V2G technology and machine learning. Background Art

[0002] At present, with the rapid growth of the number of new energy electric vehicles and charging demand, the power grid has been under unprecedented load pressure during the peak period of electricity consumption, which has seriously affected the normal operation and dispatch of the power grid. V2G technology can use the idle power stored in the on-board power battery to reversely transmit it to the energy storage device or the power grid, which can relieve the burden of the power grid. However, due to the high randomness of the charging behavior of the vehicle and the reverse charging behavior of the energy supply vehicle, it is difficult to achieve a balance between the actual power supply and demand. It is very common that some nodes in the power grid are overloaded while others have excess power. The traditional "source follows load" peak-shaving method cannot effectively solve this problem even if V2G technology is superimposed. Especially for some areas with excess power, connecting new energy power generation from vehicles and other sources to the grid will not only fail to save energy, but will also lead to unnecessary power conversion costs and reverse loads. Although the load-bearing capacity of the power grid and the storage capacity of excess power can be improved by simply increasing the number of V2G facilities, it is obviously impossible to solve this contradiction between power supply and demand in the short term due to its high construction cost. Therefore, how to make the existing relatively small number of V2G facilities better play their role in peak load shaving and valley filling in the power grid is a technical problem that urgently needs to be solved in this field. Summary of the invention

[0003] In view of this, in order to solve the technical problems existing in the art, the present invention provides a method for peak load regulation of electric energy based on V2G technology and machine learning, which specifically includes the following steps: Step 1. Extract historical operation data of new energy electric vehicles within a certain period, and perform data cleaning and segmentation on the historical operation data to obtain charging segments and reverse charging segments when the vehicle is used as an energy source; Step 2. Extract the charge and discharge status, battery current and voltage, vehicle speed, segment start and end SOC, parking charging or reverse charging location, vehicle identification and other information from the charging segment and the reverse charging segment respectively; Step 3. Create a grid digital map for the area where the vehicle travels on a daily basis, and mark the V2G facilities in the digital map; match the parking charging location and the reverse charging location with the grid cells in the digital map; take each grid cell as the central cell, and splice it with its adjacent grids to obtain a merged grid cell, and then add corresponding merged grid cell labels to the charging segment and the reverse charging segment respectively; Step 4. Cluster the charging segments and reverse charging segments to obtain the corresponding charging cluster centers and reverse charging cluster centers; match and select the positions of each charging cluster center with the corresponding central unit positions, and ignore the corresponding charging clusters whose charging cluster centers are not located in the central unit; based on the regional distribution network topology, node settings, and load distribution, add corresponding sub-region labels to each segment in the selected charging clusters and reverse charging clusters; Step 5. Divide a single day into several time periods at equal intervals. For the charging segments and reverse charging segments in a specific time period, calculate the charging capacity and reverse charging capacity respectively by using the start and end SOC of the segments combined with the time series; count the charging demand capacity of each sub-area in the time period on a single day, and the power supply capacity obtained from vehicles in the sub-area containing V2G facilities; respectively use the sub-area demand capacity and power supply capacity and their corresponding time information to establish a training sample set; Step 6. Establish two time series prediction models for demand capacity and power supply capacity based on deep neural networks respectively, and use the training sample set to train them respectively until they converge stably or reach the maximum number of training times; Step 7. Use the two trained models to predict the demand capacity and power supply capacity of each sub-region in the subsequent short-term period, and determine the equivalent demand capacity of each sub-region. On this basis, make peak-shaving decisions for each sub-region, including adjusting load distribution and adjusting power generation output.

[0004] Furthermore, in step 3, a digital map of multi-level hexagonal grid units is specifically used; before clustering is specifically performed in step 4, the number of charging and reverse charging occurring in each merged grid unit is firstly counted, and the merged grid units whose number does not reach the predetermined number are eliminated, so that private or household charging facilities that provide scattered charging services in the same period are excluded as much as possible to reduce the amount of calculation during clustering; during clustering, the DBSCAN algorithm is specifically used, and after clustering the fragment positions, the selected cluster centers are assigned corresponding clustering cluster numbers, i.e., sub-region labels, according to their different regional levels in the distribution network and the degree of load fluctuation in different time periods.

[0005] Furthermore, the prediction model establishment and training process in step 6 includes: ① Define the input and output sequence using the labels of each sub-region, the corresponding total demand capacity or total power supply capacity, and the time information including date and time period, and establish the following prediction model to predict the output sequence of the subsequent short-term period based on the previous k-th period input sequence:

[0006] in, f (·) is the mapping of the subsequent time period sequence to the previous time period sequence; Qis the sequence element of a sub-region in a certain period of time; the subscript indicates the corresponding period of time, m and p are the number of neurons in the input layer and output layer of the deep neural network respectively; ② Initialize the neural network parameters of the model and establish a training sample pool for each sub-region. Each training sample pool consists of two parts of data: a training update set and a validation set. ③ Use each set of training sample pools to train the model. First, use the data in the training update set to train and update the neural network parameters. Then, input the validation set data into the model with updated parameters, and use the output results to calculate the model loss and its partial derivatives with respect to the neural network parameters. ④ Repeat step ③ in a gradient descent manner until the model converges stably or reaches the maximum number of training times; The demand capacity and power supply capacity time series prediction models are established and trained using the same process as above.

[0007] Furthermore, in step 7, for the sub-areas with V2G facilities, the equivalent demand capacity in the subsequent short-term period is determined based on the difference between the predicted demand capacity and the power supply capacity; for the sub-areas without V2G facilities, the predicted demand capacity is directly used as the equivalent demand capacity.

[0008] Furthermore, in step 7, a global peak load regulation decision is made based on the equivalent demand capacity of each sub-region and the load fluctuation changes in different time periods, including reducing the corresponding available capacity of the sub-region whose equivalent demand capacity is less than a predetermined value.

[0009] Furthermore, the equivalent demand capacity of each sub-region is also used to guide the subsequent construction planning of charging facilities and V2G facilities.

[0010] The electric energy peak-shaving method based on V2G technology and machine learning provided by the present invention utilizes vehicle charging and discharging related operation data combined with digital map information for clustering, and calculates the charging demand and reverse power supply capacity corresponding to different sub-areas of the distribution network, so that the grid load fluctuations caused by the charging and discharging behavior of individual vehicles with strong randomness can be handled more easily; then, based on the idea of ​​time series prediction, a deep neural network model is established and trained using the demand and power supply capacity of each sub-area, which can predict the load of each sub-area in subsequent short-term continuous time periods, so that peak-shaving decisions can be made in a timely and efficient manner based on the prediction results, thereby improving the level of "source-load interaction" from the vehicle-grid level and giving full play to the positive role of V2G technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 A flow chart of the method provided by the present invention; Figure 2This is a structural diagram of a deep neural network. DETAILED DESCRIPTION

[0012] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0013] The present invention provides a method for peak load regulation of electric energy based on V2G technology and machine learning, such as Figure 1 As shown, the specific steps include: Step 1. Extract the historical operation data of new energy electric vehicles in a certain period, clean and segment the historical operation data, and obtain the charging segment and the reverse charging segment when the vehicle is used as an energy source; since the energy consumption of the vehicle group is highly correlated with the season, ambient temperature, climate and other conditions and shows regular changes, the historical operation data can be extracted separately for winter and summer, and the subsequent clustering and modeling training process can be performed to ensure the accuracy of the final peak load regulation decision; Step 2. Extract the charge and discharge status, battery current and voltage, vehicle speed, segment start and end SOC, parking charging or reverse charging location, vehicle identification and other information from the charging segment and the reverse charging segment respectively; Step 3. Create a grid digital map for the area where the vehicle travels on a daily basis, and mark the V2G facilities in the digital map; match the parking charging location and the reverse charging location with the grid cells in the digital map; take each grid cell as the central cell, and splice it with its adjacent grids to obtain a merged grid cell, and then add corresponding merged grid cell labels to the charging segment and the reverse charging segment respectively; Step 4. Cluster the charging segments and reverse charging segments to obtain the corresponding charging cluster centers and reverse charging cluster centers; match and select the positions of each charging cluster center with the corresponding central unit positions, and ignore the corresponding charging clusters whose charging cluster centers are not located in the central unit; based on the regional distribution network topology, node settings, and load distribution, add corresponding sub-region labels to each segment in the selected charging clusters and reverse charging clusters; Step 5. Divide a single day into several time periods at equal intervals, such as dividing the whole day into 144 time periods at intervals of 10 minutes. For the charging segment and reverse charging segment in the evening power consumption peak period of 17:00-20:00, use the segment start and end SOC combined with the time series to calculate the charging capacity and reverse charging capacity respectively; count the charging demand capacity of each sub-area in this period on a single day, and the power supply capacity obtained from the vehicle in the sub-area containing the V2G facilities; respectively use the sub-area demand capacity and power supply capacity and their corresponding time information to establish a training sample set; Step 6. Establish two time series prediction models for demand capacity and power supply capacity based on deep neural networks respectively, and use the training sample set to train them respectively until they converge stably or reach the maximum number of training times; Step 7. Use the two trained models to predict the demand capacity and power supply capacity of each sub-region in the subsequent short-term period, and determine the equivalent demand capacity of each sub-region. On this basis, make peak-shaving decisions for each sub-region, including adjusting load distribution and adjusting power generation output.

[0014] In a preferred embodiment of the present invention, a digital map of multi-level hexagonal grid units is specifically used in step 3; before clustering is specifically performed in step 4, the number of charging and reverse charging occurring in each merged grid unit is firstly counted, and the merged grid units whose number does not reach the predetermined number are eliminated, so that private or household charging facilities that provide scattered charging services in the same period are excluded as much as possible to reduce the amount of calculation during clustering; the DBSCAN algorithm is specifically used during clustering, and after clustering the fragment positions, the corresponding clustering cluster number, i.e., the sub-region label, is assigned to each selected cluster center according to its different regional levels in the distribution network and the degree of load fluctuation in different time periods.

[0015] In a preferred embodiment of the present invention, the prediction model establishment and training process in step 6 includes: ① Define the input and output sequence using the labels of each sub-region, the corresponding total demand capacity or total power supply capacity, and the time information including date and time period, and establish the following prediction model to predict the output sequence of the subsequent short-term period based on the previous k-th period input sequence:

[0016] in, f (·) is the mapping of the subsequent time period sequence to the previous time period sequence; Q is the sequence element of a sub-region in a certain period of time; the subscript indicates the corresponding period of time, m and p are the number of neurons in the input layer and output layer of the deep neural network respectively. The network structure is as follows Figure 2 As shown; ② Initialize the neural network parameters of the model and establish a training sample pool for each sub-region. Each training sample pool consists of two parts of data: a training update set and a validation set. ③ Use each set of training sample pools to train the model. First, use the data in the training update set to train and update the neural network parameters. Then, input the validation set data into the model with updated parameters, and use the output results to calculate the model loss and its partial derivatives with respect to the neural network parameters. ④ Repeat step ③ in a gradient descent manner until the model converges stably or reaches the maximum number of training times; The demand capacity and power supply capacity time series prediction models are both established and trained using the same process as above. Compared with the power load time series prediction method using LSTM and RNN algorithms, the neural network structure of the present invention is simpler, easier to adjust parameters and has lower computational overhead.

[0017] In a preferred embodiment of the present invention, in step 7, for the sub-area with V2G facilities, the equivalent demand capacity in the subsequent short-term period is determined based on the difference between the predicted demand capacity and the power supply capacity; for the sub-area without V2G facilities, the predicted demand capacity is directly used as the equivalent demand capacity.

[0018] In a preferred embodiment of the present invention, in step 7, a global peak-shaving decision is made based on the equivalent demand capacity of each sub-region and the load fluctuation changes in different time periods, including indicating that there is excess and waste of electricity in the sub-region where the equivalent demand capacity is less than a predetermined value, so the corresponding available capacity should be reduced, and the excess power supply capacity should be restricted from being connected to the grid to avoid unnecessary reverse load and conversion costs.

[0019] In a preferred embodiment of the present invention, the equivalent demand capacity of each sub-region is also used to guide the subsequent construction planning of charging facilities and V2G facilities, such as adding V2G facilities to some sub-regions with long-term tight charging demand, and reducing or dismantling some charging facilities that are often idle.

[0020] It should be understood that the order of execution of each step in the embodiment of the present invention does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0021] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for peak-shaving of electric energy based on V2G technology and machine learning, characterized in that: The specific steps include: Step 1. Extract historical operation data of new energy electric vehicles within a certain period of time, and divide them into charging segments and reverse charging segments when the vehicle is used as an energy source; Step 2. Extracting information including vehicle status and location from the charging segment and the reverse charging segment respectively; Step 3: Create a grid digital map for the area where vehicles travel on a daily basis, and mark the location of V2G facilities on the digital map; Step 4. Cluster the charging segments and reverse charging segments to obtain corresponding charging clusters, reverse charging cluster centers and clusters; based on the regional distribution network topology, node settings and load distribution, add corresponding sub-region labels to the segments in each cluster; Step 5. Divide a single day into several time periods at equal intervals. For the charging segments and reverse charging segments in a specific time period, count the charging demand capacity and the power supply capacity from vehicles in each sub-area in different time periods of the single day and establish a training sample set. Step 6. Establish two time series prediction models for demand capacity and power supply capacity based on deep neural network respectively, and train them respectively using training sample sets; Step 7. Use the two trained models to predict the demand capacity and power supply capacity of each sub-region in the subsequent short-term period, and determine the equivalent demand capacity of each sub-region. On this basis, make peak-shaving decisions for each sub-region, including adjusting load distribution and adjusting power generation output.

2. The method according to claim 1, characterized in that: In step 3, a digital map of multi-level hexagonal grid cells is specifically used to match the parking charging position and the reverse charging position with the grid cells in the digital map; each grid cell is used as a central cell, and is spliced ​​with its adjacent grids to obtain a merged grid cell, and then corresponding merged grid cell labels are added to the charging segment and the reverse charging segment respectively; Before clustering is specifically performed in step 4, the number of charging and reverse charging occurring in each merged grid unit is first counted, and the merged grid units whose number does not reach the predetermined number are eliminated to reduce the amount of calculation during clustering; the DBSCAN algorithm is specifically used during clustering to match and screen the position of each charging cluster center with the corresponding central unit position, and ignore the corresponding charging cluster clusters whose charging cluster center positions are not in the central unit; after clustering the fragment positions, the selected cluster centers are assigned corresponding cluster cluster numbers, i.e., sub-region labels, according to their different regional levels in the distribution network, grid topology, node settings, and the degree of load fluctuation in different time periods.

3. The method according to claim 1, characterized in that: The process of establishing the sequence sample set in step 5 includes: using the start and end SOC of the segment combined with the time series to calculate the charging capacity and the reverse charging capacity respectively; counting the charging demand capacity of each sub-area in the time period on a single day, and the power supply capacity obtained from the vehicle in the sub-area containing the V2G facilities; and establishing the training sample set respectively using the sub-area demand capacity and power supply capacity and their corresponding time information; The prediction model building and training process in step 6 includes: ① Define the input and output sequence using the labels of each sub-region, the corresponding total demand capacity or total power supply capacity, and the time information including date and time period, and establish the following prediction model to predict the output sequence of the subsequent short-term period based on the previous k-th period input sequence: ， in, f (·) is the mapping of the subsequent time period sequence to the previous time period sequence; Q is the sequence element of a sub-region in a certain period of time, and the subscript indicates the corresponding period of time. m and p are the number of neurons in the input layer and output layer of the deep neural network respectively; ② Initialize the neural network parameters of the model and establish a training sample pool for each sub-region. Each training sample pool consists of two parts of data: a training update set and a validation set. ③ Use each set of training sample pools to train the model. First, use the data in the training update set to train and update the neural network parameters. Then, input the validation set data into the model with updated parameters, and use the output results to calculate the model loss and its partial derivatives with respect to the neural network parameters. ④ Repeat step ③ in a gradient descent manner until the model converges stably or reaches the maximum number of training times.

4. The method according to claim 1, characterized in that: In step 7, for sub-areas with V2G facilities, the equivalent demand capacity in the subsequent short-term period is determined based on the difference between the predicted demand capacity and the power supply capacity; for sub-areas without V2G facilities, the predicted demand capacity is directly used as the equivalent demand capacity.

5. The method according to claim 1, characterized in that: In step 7, a global peak load regulation decision is made based on the equivalent demand capacity of each sub-region and the load fluctuation changes in different time periods, including reducing the corresponding available capacity of the sub-region whose equivalent demand capacity is less than a predetermined value.

6. The method according to claim 1, characterized in that: The equivalent demand capacity of each sub-region is also used to guide the subsequent construction planning of charging facilities and V2G facilities.

Images