A power grid load prediction method considering charging modes of multiple types of electric vehicles under extremely high temperature weather

Abstract

The application discloses a power grid load prediction method considering charging modes of multiple types of electric vehicles under extremely high-temperature weather, first, electric vehicle charging characteristics and travel information of the owner under extremely high-temperature weather are collected, and the electric vehicles are classified according to the collected information; the travel law of the electric vehicles under extremely high-temperature weather is simulated by using a Monte Carlo sampling method; the charging behavior of each type of electric vehicle under high-temperature weather is simulated by considering two factors, i.e., a market time-of-use electricity price policy and the growth of air conditioner load of the electric vehicles under extremely high-temperature weather; finally, a similar day method combined with a neural network is used to predict the electric vehicle load under extremely high-temperature weather, and the total load of different types of electric vehicles connected to the power grid is calculated. The method can accurately predict the charging load of various electric vehicles, is beneficial to the economic operation and energy management of the electric vehicle charging station, and is also beneficial to helping the power grid realize stable operation by peak shaving and frequency modulation.

Description

Technical Field

[0001] This invention relates to the field of power system load forecasting and planning technology, specifically a power grid load forecasting method that takes into account various electric vehicle charging methods under extreme high-temperature weather conditions. Background Technology

[0002] With increasing environmental pollution and energy shortages, electric vehicles (EVs), with their unparalleled advantages in energy conservation and emission reduction compared to gasoline-powered vehicles, are poised to replace gasoline-powered cars in the future. To address resource and environmental issues, the Chinese government attaches great importance to the development of EVs and has introduced numerous policies to support and promote the industry. As EVs become more widespread, large-scale integration into the power grid will have a significant impact on the operation and planning of the power system.

[0003] The large-scale integration of electric vehicles into the power grid will impact the operation of the existing distribution network. Therefore, understanding the charging load characteristics of various types of electric vehicles, especially their charging load characteristics under extreme high-temperature weather, can enable them to play a positive and crucial role in peak shaving, valley filling, and renewable energy absorption within the distribution network system.

[0004] Most current methods for predicting electric vehicle load do not differentiate between the charging load characteristics of different types of electric vehicles, and do not fully consider the impact of different types of electric vehicle charging characteristics on the charging load under extreme high-temperature weather. This makes it impossible to accurately predict the charging load of electric vehicles under extreme high temperatures, which is not conducive to the economic operation and energy management of electric vehicle charging stations, the development of interaction strategies between electric vehicles and the power grid, and the achievement of stable operation of power grid peak regulation and frequency regulation. Summary of the Invention

[0005] Based on this, in order to accurately predict the load of a large number of electric vehicles connected to the grid, this invention provides a grid load prediction method that considers the charging methods of various types of electric vehicles under extreme high temperature weather. The method divides electric vehicles into four categories, considers the market time-of-use pricing policy and the increase in air conditioning load of electric vehicles under extreme high temperature weather, and uses a combination of similar day method and neural network method to predict the charging load of each type of electric vehicle under extreme weather.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solution:

[0007] This invention is a power grid load forecasting method that takes into account various electric vehicle charging methods under extreme high-temperature weather conditions, and includes the following steps:

[0008] Step 1: Collect information on the charging characteristics of electric vehicles and the travel information of vehicle owners under extreme high temperature weather, and classify them according to the charging characteristics and travel information of electric vehicles;

[0009] Step 2: Use Monte Carlo sampling to simulate the travel patterns of electric vehicles under extreme weather conditions;

[0010] Step 3: Based on the travel patterns of electric vehicles obtained in Step 2, and considering the market time-of-use electricity pricing policy and the increase in air conditioning load of electric vehicles under extreme high temperature weather, simulate the charging behavior of each type of electric vehicle under extreme high temperature weather.

[0011] Step 4: Based on the electric vehicle charging behavior simulation results from Step 3, a method combining similar day method and neural network is used to predict the charging load of various types of electric vehicles under extreme high temperature weather, and the total load of each type of electric vehicle connected to the power grid is obtained by superimposing the results.

[0012] A further improvement of the present invention is that, in step 1, electric vehicles are divided into four categories: electric private cars, electric buses, electric taxis, and electric official vehicles; and travel information of owners of each type of vehicle is collected, specifically including: the departure time and departure location of the vehicle, the driving route of the vehicle, and the arrival time and arrival location of the vehicle.

[0013] A further improvement of the present invention is that the Monte Carlo sampling method is performed as follows:

[0014] Based on travel time, travel location, and travel route, construct an n*n location matrix L for electric private cars, electric buses, electric taxis, and electric official vehicles:

[0015]

[0016] This represents the position in the φ-th row of the position matrix. The location of the column. When At that time, the elements in the position matrix L Indicates the starting point of the electric vehicle. When At that time, the elements in the position matrix L This indicates the destination of the electric vehicle.

[0017] The probability matrix P of an electric vehicle appearing at a certain location is:

[0018]

[0019] Elements in probability matrix P Indicates the location where the electric vehicle is located. The probability of finding the destination. Let distance matrix D represent the distance between the starting point and the destination. Distance matrix D is:

[0020]

[0021] d elements in distance matrix D δγ This represents the distance between the departure point and the destination point. δ represents the departure point, and γ represents the destination point.

[0022] Based on the required accuracy, the number of simulations, N, is determined. New random numbers are generated using the probability matrix P, and these generated random numbers are then substituted into the established probability matrix P. Each type of electric vehicle is simulated N times, generating N sample values ​​for each. Statistical analysis is performed on the sample values ​​to obtain the travel patterns of each type of electric vehicle.

[0023] A further improvement of the present invention is that, in step 3, the formula for calculating the charging time of each type of electric vehicle under extreme high-temperature weather is as follows:

[0024]

[0025] In the formula, T ev D represents the daily mileage of various types of electric vehicles. ev Let α represent the electricity consumption per kilometer for each type of electric vehicle, α represent the charging efficiency of each charging station, and P represent the total electricity consumption per kilometer. ev Charging efficiency for various types of electric vehicles;

[0026] Among them, the daily driving mileage T of various types of electric vehicles ev The calculation formula is:

[0027]

[0028] In the formula, x represents the daily mileage of various types of electric vehicles based on historical data, and δ ev and β ev These are the expected value and variance of the logarithm lnx of the daily mileage x, respectively, and the δ of different types of electric vehicles. ev and β ev The values ​​are different.

[0029] A further improvement of this invention lies in considering that under the market-based time-of-use electricity pricing policy, the charging behavior of owners of various types of electric vehicles will change. Taxi and bus companies, for economic reasons, will adopt a battery swapping model for charging electric vehicles. The swapped-out batteries will be recharged uniformly during periods of low electricity prices. The charging start time of electric vehicles using the battery swapping model follows a uniform distribution, with the following probability density function:

[0030]

[0031] In the formula, b is the time when the off-peak electricity price ends, and a is the time when the off-peak electricity price begins.

[0032] The start-up charging time of electric vehicles charging in non-battery-swapping mode all follow a normal distribution, and their probability density function is:

[0033]

[0034] In the formula, μ represents the mean of the initial charging time following a normal distribution, and the standard deviation σ determines the amplitude of the distribution.

[0035] A further improvement of the present invention is that the specific operation of step 4 is as follows:

[0036] Step 4.1: Acquire a large amount of historical charging station data, including the start time of electric vehicle charging, the remaining battery power of the electric vehicle at the start of charging, and the duration of a single charging session. Quantify the similarity using a similarity evaluation function, and select historical charging station data for each type of electric vehicle that meets the similarity standard (i.e., has the smallest Euclidean distance to the predicted date) as the data for similar days.

[0037] Step 4.2: Determine the number of hidden layer nodes in the neural network, and use the selected historical data of similar days as samples to input the network for training;

[0038] Step 4.3: Obtain the charging load of each type of electric vehicle under extreme high temperature on the predicted day. Add up the charging load of each type of electric vehicle to obtain the total load of each type of electric vehicle connected to the power grid.

[0039] P alli =P allsi +P allbi +P allti +P alloi i = 1, 2, 3...T

[0040] In the formula, P alli P represents the total load of all electric vehicles at time i; allsi P represents the total load of the electric private car at time i; allbi The total load of the electric bus at time i; P allti P represents the total load of the electric taxis at time i; alloi Let be the total load of the electric official vehicle at time i; divide the day into T time periods.

[0041]

[0042] In the formula, P si Let N be the load of the electric private car at time i. s P represents the number of electric private cars. bi Let N be the load of the electric bus at time i. b P represents the number of electric buses. ti Let N be the load of the electric taxi at time i. t P represents the number of electric taxis. oiLet N be the load of the electric official vehicle at time i. o This refers to the number of electric private cars.

[0043] A further improvement of the present invention is that the specific steps for calculating similar days are as follows: Let the meteorological feature vector of each day be...

[0044]

[0045] In the formula, T gmax , T gmin The maximum, average, and minimum temperatures for day g;

[0046] Y g Y is the meteorological feature vector for the predicted day. j For historical days with the same day type and season type, their meteorological feature vector is:

[0047]

[0048] Using Euclidean distance d gj To describe the overall difference in meteorological factors over the past two days, the Euclidean distance d gj The expression is:

[0049]

[0050] In the formula, k is the index of the feature vector, and m is the number of feature vectors. The historical day with the highest similarity (smallest Euclidean distance) to the predicted day is calculated using the above formula. The historical day with the highest similarity is used as the similar day to the predicted day.

[0051] The beneficial effects of this invention are as follows: This invention provides a power grid load forecasting method considering various types of electric vehicle charging methods under extreme high-temperature weather. Electric vehicles are subdivided into four categories for load forecasting. Monte Carlo sampling is used to obtain the travel patterns of different types of electric vehicles under extreme high-temperature weather. Based on this, the charging behavior of each type of electric vehicle under extreme high-temperature weather is simulated by comprehensively considering four factors: charging efficiency, charging time, market electricity price, and remaining power state (SOC). A combination of the similarity day method and neural network method is used to predict the load of electric vehicles under extreme high-temperature weather. This method predicts the charging load of electric vehicles more accurately than traditional methods, which is beneficial not only to the economic operation and energy management of electric vehicle charging stations and the formulation of interaction strategies between electric vehicles and the power grid, but also to helping the power grid achieve stable operation through peak shaving and frequency regulation. Attached Figure Description

[0052] Figure 1 A flowchart for load forecasting of multiple types of electric vehicles;

[0053] Figure 2 Flowchart for Monte Carlo simulation;

[0054] Figure 3 This is a fully connected neural network topology. Detailed Implementation

[0055] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0056] Figure 1 This is a flowchart of the method of the present invention; a grid load forecasting method of the present invention considering multiple types of electric vehicle charging methods under extreme high temperature weather includes the following steps:

[0057] Step 1: Collect the charging characteristics of electric vehicles (EVs) and the travel information of vehicle owners under extreme high temperature weather, and classify them according to the charging characteristics and travel information of electric vehicles (EVs);

[0058] Step 2: Use Monte Carlo sampling to simulate the travel patterns of electric vehicles under extreme weather conditions;

[0059] Step 3: Based on the travel patterns of electric vehicles obtained in Step 2, and considering the market electricity price time-of-use pricing policy and the increase in air conditioning load of electric vehicles under extreme high temperature weather, simulate the charging behavior of each type of electric vehicle under extreme high temperature weather.

[0060] Step 4: Based on the electric vehicle charging behavior simulation results from Step 3, a method combining the similarity day method and neural networks is used to predict the charging load of various types of electric vehicles under extreme high-temperature weather. The total load of each type of electric vehicle connected to the power grid is obtained by superimposing the results. In Step 1, electric vehicle types are divided into four categories: electric private cars, electric buses, electric taxis, and electric official vehicles; the travel information of vehicle owners is collected, specifically including: vehicle departure time and departure location, vehicle route, and vehicle arrival time and arrival location.

[0061] The travel patterns in step 2 determine the charging behavior of car owners. After obtaining the charging behavior of electric vehicles, the charging load can be obtained by combining the charging behavior with the charging power of electric vehicles. Using a combination of similar day method and neural network can make the load prediction results more accurate. Figure 2 The flowchart for Monte Carlo simulation is as follows. Step 2 uses Monte Carlo sampling to simulate the travel patterns of electric vehicles under extreme weather conditions. The process is as follows:

[0062] Establish a probability and statistics model, specifically as follows:

[0063] Based on travel time, travel location, and travel route, construct an n*n location matrix L for electric private cars, electric buses, electric taxis, and electric official vehicles:

[0064]

[0065] This represents the position in the φ-th row of the position matrix. The location of the column. When At that time, the elements in the position matrix L Indicates the starting point of the electric vehicle. When At that time, the elements in the position matrix L This indicates the destination of the electric vehicle.

[0066] The probability matrix P of an electric vehicle appearing at a certain location is:

[0067]

[0068] Elements in probability matrix P Indicates the location where the electric vehicle is located. The probability of finding the destination. Let distance matrix D represent the distance between the starting point and the destination. Distance matrix D is:

[0069]

[0070] d elements in distance matrix D dγ This represents the distance between the departure point and the destination point. δ represents the departure point, and γ represents the destination point.

[0071] Based on the required accuracy, the number of simulations, N, is determined. New random numbers are generated using the probability matrix P, and these generated random numbers are then substituted into the established probability matrix P. Each type of electric vehicle is simulated N times, generating N sample values ​​for each. Statistical analysis is performed on the sample values ​​to obtain the travel patterns of each type of electric vehicle.

[0072] In step 3, the charging behavior of each type of electric vehicle under extreme high-temperature weather includes charging time, t. ev The calculation formula is:

[0073]

[0074] In the formula, T ev This represents the daily mileage of various types of electric vehicles. (D) ev α represents the electricity consumption per kilometer for each type of electric vehicle. α represents the charging efficiency of each charging station. P ev Charging efficiency for various types of electric vehicles.

[0075] Among them, the daily mileage T of various types of electric vehiclesev The calculation formula is:

[0076]

[0077] In the formula, x represents the daily mileage of various types of electric vehicles based on historical data. δ ev and β ev Let be the expected value and variance of the logarithm lnx of the daily mileage x, respectively. δ for different types of electric vehicles. ev and β ev The values ​​are different.

[0078] Considering the changes in charging behavior among various types of electric vehicle owners under the time-of-use (TOU) pricing policy, the primary influence is the charging time. Charging data is collected and statistically analyzed for different types of electric vehicles under the TOU pricing policy. Taxis and company vehicles, for economic reasons, use battery swapping to charge their electric vehicles, with the swapped-out batteries being recharged during off-peak electricity periods. Among these, the charging start time of electric vehicles using the battery swapping model follows a uniform distribution, and its probability density...

[0079] The function is:

[0080]

[0081] In the formula, b is the time when the off-peak electricity price ends, and a is the time when the off-peak electricity price begins.

[0082] Except for car owners using battery swapping, the charging start times for all other car owners follow a normal distribution. The probability density function is: 4

[0083]

[0084] In the formula, μ represents the mean of the initial charging time following a normal distribution, and the standard deviation σ determines the amplitude of the distribution.

[0085] Figure 3 For a fully connected neural network topology, the specific operation of step 4 is as follows:

[0086] Step 4.1: Acquire a large amount of historical data on charging stations, including the start time of electric vehicle charging, the remaining battery power of the electric vehicle at the start of charging, and the duration of a single charging session. Quantify the similarity using a similarity evaluation function, and select historical data of electric vehicle charging stations during extreme high-temperature weather on historical days that meet the similarity criteria as historical data for similar days.

[0087] Step 4.2: Determine the number of hidden layer nodes in the neural network, and use the selected historical data of similar days as samples to input the network for training;

[0088] Step 4.3: Obtain the charging load of each type of electric vehicle under extreme high temperature on the predicted day. Add up the charging load of each type of electric vehicle to obtain the total load of each type of electric vehicle connected to the power grid.

[0089] P alli =P allsi +P allbi +P allti +P alloi ,i=1,2,3...T (8)

[0090] In the formula, P alli P represents the total load of all electric vehicles at time i; allsi P represents the total load of the electric private car at time i; allbi The total load of the electric bus at time i; P allti P represents the total load of the electric taxis at time i; alloi Let be the total load of the electric official vehicle at time i; divide the day into T time periods.

[0091]

[0092] In the formula P si Let N be the load of the electric private car at time i. s P represents the number of electric private cars. bi Let N be the load of the electric bus at time i. b P represents the number of electric buses. ti Let N be the load of the electric taxi at time i. t P represents the number of electric taxis. oi Let N be the load of the electric official vehicle at time i. o The number of electric private cars;

[0093] The specific steps for calculating similar days are as follows: Let the meteorological feature vector of each day be:

[0094]

[0095] In the formula, T gmax , T gmin The maximum, average, and minimum temperatures for day g;

[0096] remember Y g is the meteorological feature vector for the predicted day, Y j For historical days with the same day type and season type, their meteorological feature vector is: Using Euclidean distance d gj To describe the overall difference in meteorological factors over the past two days, the Euclidean distance d gj The expression is:

[0097]

[0098] In the formula, k is the index of the feature vector, and m is the number of feature vectors. The historical day with the highest similarity (smallest Euclidean distance) is calculated using formula (11). The historical day with the highest similarity is used as the similar day for the predicted day.

[0099] The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of the claims.

Claims

1. A grid load forecasting method considering multiple types of electric vehicle charging methods under extreme high-temperature weather, characterized in that: Includes the following steps: Step 1: Collect electric vehicle charging characteristics and owner travel information under extreme high temperature weather, and classify electric vehicles according to their charging characteristics and travel information; Step 2: Use Monte Carlo sampling to simulate the travel patterns of electric vehicles under extreme weather conditions; Step 3: Based on the travel patterns of electric vehicles obtained in Step 2, and considering the market time-of-use electricity pricing policy and the increase in air conditioning load of electric vehicles under extreme high temperature weather, simulate the charging behavior of each type of electric vehicle under extreme high temperature weather. Step 4: Based on the electric vehicle charging behavior simulation results from Step 3, a method combining similar day method and neural network is used to predict the charging load of various types of electric vehicles under extreme high temperature weather, and the total load of each type of electric vehicle connected to the power grid is obtained by superimposing the results. In step 3, the formula for calculating the charging time for each type of electric vehicle under extreme high-temperature weather is as follows: In the formula, Indicates electric vehicle, The daily driving mileage for various types of electric vehicles, Electricity consumption per kilometer for various types of electric vehicles. The charging efficiency of each charging station, Charging efficiency for various types of electric vehicles; Among them, the daily driving mileage of various types of electric vehicles The expression is: In the formula, The daily mileage of various types of electric vehicles was collected based on historical data. and These are the daily mileages. logarithm Expectation and variance of different types of electric vehicles and Different values; Considering the changing charging behaviors of various types of electric vehicle owners under the time-of-use pricing policy, and for economic reasons, electric taxis and electric buses adopt a battery swapping model for charging. The swapped-out batteries are then recharged during off-peak electricity periods. For electric vehicles using the battery swapping model, the charging start time follows a uniform distribution, with the following probability density function: In the formula, b is the time when the off-peak electricity price ends, a is the time when the off-peak electricity price begins, and t is the charging start time; The start-up charging time of electric vehicles charging in non-battery-swapping mode all follow a normal distribution, and their probability density function is: In the formula, This represents the mean and standard deviation of the initial charging time, which follow a normal distribution. This determines the magnitude of its distribution; The specific operation of step 4 is as follows: Step 4.1: Obtain a large amount of historical charging pile data, including the start time of electric vehicle charging, the remaining power of the electric vehicle at the start of charging, and the duration of a single charging session. Combine this with a similarity evaluation function to quantify the similarity and select historical charging pile data of various types of electric vehicles that meet the similarity criteria as data for similar days of the prediction day. Step 4.2: Determine the number of hidden layer nodes in the neural network, and use the selected historical data of similar days as samples to input the network for training; Step 4.3: Obtain the charging load of each type of electric vehicle under extreme high temperature conditions on the predicted day. Add up the charging loads of each type of electric vehicle to obtain the total load of each type of electric vehicle connected to the power grid, expressed as: In the formula, For the first The total load of all electric vehicles at any given moment; For the first The total load of electric private cars at any given time; No. The total load of the electric bus at any given time; For the first Total load of electric taxis at all times; For the first The total load of electric official vehicles at all times; dividing the day into T time periods. ， ， ， In the formula, For the first The load on electric private cars at all times. The number of electric private cars; For the first The load on electric buses at all times The number of electric buses; For the first The load of electric taxis at all times The number of electric taxis; For the first The load of electric official vehicles at all times The number of electric private cars; In step 4.1, the specific steps for filtering similar days are as follows: Let the meteorological feature vector of each day be: In the formula, , , For the first Daily maximum temperature, average temperature, and minimum temperature; remember The meteorological feature vector for the predicted day is given by the historical day with the same day type and season type. Using Euclidean distance To describe the overall difference in meteorological factors over the past two days, the Euclidean distance is used. The expression is: In the formula, The feature vector index, Given the number of eigenvectors, calculate the historical day with the smallest Euclidean distance to the predicted day, and use this historical day as the similar day to the predicted day.

2. The grid load forecasting method considering multiple types of electric vehicle charging methods under extreme high-temperature weather as described in claim 1, characterized in that: In step 1, electric vehicles are divided into four categories: electric private cars, electric buses, electric taxis, and electric official vehicles; travel information of owners of each type of vehicle is collected, specifically including: vehicle departure time and departure location, vehicle route, and vehicle arrival time and arrival location.

3. The grid load forecasting method considering multiple types of electric vehicle charging methods under extreme high-temperature weather as described in claim 1, characterized in that: In step 2, the Monte Carlo sampling method proceeds as follows: Based on travel time, travel location, and travel route, construct an n*n location matrix for electric private cars, electric buses, electric taxis, and electric official vehicles. for: Represents the position matrix of the first position. Line number The location of the column, when At that time, position matrix medium elements Indicates the starting point of the electric vehicle, when At that time, position matrix elements in Indicates the destination of the electric vehicle; Probability matrix of an electric vehicle appearing at a certain location for: probability matrix medium elements Indicates the location where the electric vehicle is located. The probability at a location is expressed using the distance matrix. Distance matrix represents the distance between the departure point and the destination point. for: Distance matrix medium elements Indicates the distance between the departure point and the destination point. Indicates the departure point. Indicates the destination; Based on the accuracy requirements, determine the number of simulations N, and then use the probability matrix... Generate new random numbers; input the generated new random numbers into the probability matrix. In the process, each type of electric vehicle was simulated N times, generating N sample values ​​for each type. Statistical analysis was performed on the sample values ​​to obtain the travel patterns of each type of electric vehicle.

Images