A method for distributing and scheduling charging power of electric vehicles based on water filling algorithm

Abstract

The present invention relates to a method for distributing and scheduling charging power of electric vehicles based on a water filling algorithm, and belongs to the field of electric vehicles. The method comprises: collecting historical load data of the distribution network of a target residential area to construct a first data set; constructing a prediction model, and training the prediction model based on the first data set to obtain a trained prediction model; constructing a second data set, and inputting the second data set into the trained prediction model to obtain the daily basic load curve predicted on the current date of the distribution network of the residential area and setting the valley period; taking the minimum variance of the total load of the distribution network as the objective function, and taking the capacity of the distribution network transformer and the charge state of the electric vehicle as constraints, and using the water filling algorithm to generate the optimal charging power distribution scheduling plan for the electric vehicle on the current date. The method uses a water filling algorithm to distribute the charging power of electric vehicles, does not require complex mathematical models and parameter settings, has a fast execution speed, and solves the problems of large calculation amount and long time consumption of the current electric vehicle charging distribution algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of electric vehicles, and in particular to a method for distributing and scheduling charging power of electric vehicles based on a water filling algorithm. Background Art

[0002] As the global warming problem becomes increasingly serious, reducing carbon emissions has become a consensus of the international community. Electric vehicles are developing rapidly as one of the effective means to reduce environmental pollution. However, the construction of electric vehicle charging facilities and power grid supporting facilities lags behind. In particular, old communities cannot support a large number of electric vehicles for centralized charging due to the power limitation of distribution transformers, which easily causes overload of the distribution network.

[0003] At present, the existing electric vehicle charging allocation methods usually find the optimal solution based on algorithms such as particle swarm algorithms and genetic algorithms. The parameter setting and solution process are relatively complex, resulting in large amount of calculation and long time consumption. In addition, the existing electric vehicle charging allocation methods are usually solved based on fixed valley time periods. With the access of various new energy sources such as wind power and solar power generation to the power grid, the fixed valley time period will cause electric vehicles to miss the best charging time or fail to meet the charging needs of users, resulting in unstable power supply of the distribution network.

[0004] Therefore, it is an urgent problem to use a short-time algorithm to allocate charging for electric vehicles and predict the valley period based on actual conditions, so as to ensure the stability of power supply in the distribution network and meet the charging needs of electric vehicle users. Summary of the invention

[0005] In view of the above analysis, the present invention aims to provide an electric vehicle charging power allocation and scheduling method based on a water filling algorithm, so as to solve the problems that the current electric vehicle charging allocation algorithm has large computational complexity, long time consumption and is solved based on a fixed valley period.

[0006] The present invention provides a method for distributing and scheduling charging power of electric vehicles based on a water filling algorithm, the method comprising the following steps:

[0007] Collect historical load data of the distribution network of the target residential area, and select corresponding data from the historical load data of the distribution network according to the current date to construct a first data set;

[0008] Building a prediction model based on Gaussian process regression, and training the prediction model based on the first data set to obtain a trained prediction model;

[0009] Dividing the current date into a time point sequence based on a preset time interval to construct a second data set, and inputting the second data set into a trained prediction model to obtain a daily basic load curve predicted on the current date of the residential community distribution network;

[0010] The valley period is set according to the predicted daily basic load curve, the minimum total load variance of the distribution network is taken as the objective function, the distribution network transformer capacity and the charge state of the electric vehicle are taken as constraints, and the water filling algorithm is used to generate the optimal charging power distribution scheduling plan for the electric vehicle on the current date.

[0011] Furthermore, setting the valley period according to the predicted daily basic load curve includes:

[0012] Setting a valley period threshold, and setting the start time and end time of each valley period according to the predicted daily basic load curve and the valley period threshold, so as to obtain each valley period corresponding to the current date;

[0013] For each valley period, a plurality of sub-intervals are obtained according to preset time intervals.

[0014] Furthermore, the objective function is obtained by the following formula:

[0015] ,

[0016] in, is the charging power of the nth electric vehicle in the kth sub-interval, is the predicted base load for the kth sub-interval on the current date, N is the total number of electric vehicles in the charging waiting sequence, is the sub-interval contained in the valley period corresponding to the nth electric car, is the arrival time of the nth electric car, represents the subinterval corresponding to the arrival time of the nth electric car, is the departure time of the nth electric car, It represents the subinterval corresponding to the departure time of the nth electric car.

[0017] Furthermore, the constraint condition of the objective function is obtained by the following formula:

[0018] ,

[0019] in, is the desired state of charge of the nth electric vehicle, is the initial state of charge of the nth electric vehicle, is the charging efficiency of the nth electric vehicle, is the total charge demand of the nth electric vehicle.

[0020] Furthermore, the method of using the water filling algorithm to generate the optimal charging power distribution scheduling plan for electric vehicles on the current date includes:

[0021] The water filling algorithm is used to solve the objective function of each electric vehicle in the charging waiting sequence to obtain the charging power in each sub-interval corresponding to each electric vehicle; the charging power in each sub-interval corresponding to all electric vehicles in the charging waiting sequence is the optimal charging power allocation scheduling plan for electric vehicles on the current date, and orderly charging is carried out according to the plan.

[0022] Furthermore, selecting corresponding data from the distribution network historical load data according to the current date to construct the first data set includes:

[0023] Divide the historical daily base load curve of the distribution network into “weeks”;

[0024] According to the current date, the corresponding "weekday" is obtained, and the daily basic load curves of the same "weekday" in the previous three weeks constitute the set corresponding to the current date;

[0025] For the daily basic load curve of each "weekday" in the set corresponding to the current date: divide it according to the preset time interval to obtain a time point sequence and its corresponding basic load sequence, thereby forming a first data set corresponding to the current date.

[0026] Furthermore, the prediction model is expressed as:

[0027] ,

[0028] in, is the basic load vector output by the prediction model for the training set, is Gaussian noise, is the basic load vector predicted by the prediction model for the training set, is the input training set time point sequence vector, is the time point sequence vector corresponding to the output predicted basic load vector, is the mean function, represents the mean value, is the covariance function, represents the Gaussian process regression model, is the input vector The variance of is the iteration step size of the covariance function.

[0029] Furthermore, the prior distribution of the prediction model is:

[0030] ,

[0031] in, is an n-dimensional symmetric positive definite matrix, is the covariance matrix of the noise, is the n-dimensional identity matrix, where n is the dimension of the input vector, represents a normal distribution, for and The similarity between the corresponding loads, is the i-th time point of the input training set, is the jth time point of the input training set.

[0032] Furthermore, the hyperparameters are solved by the following formula: , , :

[0033] ,

[0034] in, for The target value of , that is, the corresponding value when minimizing NLML value; is the probability density function, is a hyperparameter, , For function calls, is the dimension of the input vector, represents minimizing the negative log-likelihood function;

[0035] Take the partial derivative of the above formula and use the conjugate gradient method to get the extreme point, so as to solve the value of the hyperparameter:

[0036] ,

[0037] in, for The transposed matrix of .

[0038] Furthermore, the following steps are performed for each electric vehicle in the charging waiting sequence:

[0039] Step S401, initializing the error value, and randomly initializing the charging power in each sub-interval corresponding to the electric vehicle within the upper and lower limits of the charging power of the electric vehicle;

[0040] Step S402: obtaining a minimum value and a maximum value of a water filling variable based on an initial charging power in each sub-interval;

[0041] Step S403: if the difference between the maximum value and the minimum value of the water filling variable is less than or equal to the error value, then output the charging power in the subinterval corresponding to the electric vehicle; otherwise, execute step S404;

[0042] Step S404: Get the current water filling variable value through formula 16, and update the charging power in each sub-interval corresponding to the electric vehicle through formula 17. If the sum of the charging powers in each sub-interval corresponding to the electric vehicle after the update is greater than its total charge demand Then set , otherwise set ; Calculate the difference between the maximum value and the minimum value of the water filling variable. If the difference between the maximum value and the minimum value of the water filling variable is less than or equal to the error value, output the charging power in the subinterval corresponding to the electric vehicle; otherwise, repeat step S404 until the difference between the maximum value and the minimum value of the water filling variable is less than or equal to the error value;

[0043] The current water filling variable value is obtained by the following formula:

[0044] , (16)

[0045] in, Fill the variable value with the current water, is the maximum value of the water filling variable, Minimum value for water fill variable;

[0046] The charging power in each sub-interval corresponding to the electric vehicle is obtained by the following formula:

[0047] , (17)

[0048] in, is the charging power of the nth electric vehicle in the kth sub-interval, is the total load excluding the charging power of the nth electric vehicle in the kth subinterval.

[0049] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0050] 1. The present invention adopts a water filling algorithm to distribute the charging power of electric vehicles. It does not require complex mathematical models and parameter settings, has a fast execution speed, and can support near real-time performance, solving the problem of large calculation amount and long time consumption of current electric vehicle charging distribution algorithms.

[0051] 2. The present invention uses the “same-day” prediction method to predict the basic load of the distribution network based on the historical data of the previous three weeks, so that the prediction model makes predictions based on the latest information, thereby improving the accuracy and efficiency of the prediction while using less historical data, and providing a basis for achieving the optimal charging scheduling of electric vehicles.

[0052] 3. The present invention flexibly sets the valley period based on the prediction of the distribution network load, and solves the electric vehicle charging power allocation algorithm based on the changing valley period, thereby improving the stability of the distribution network power supply while meeting the user's charging needs.

[0053] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be achieved and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings are only used for the purpose of illustrating specific embodiments and are not to be considered as limiting the present invention. In the entire drawings, the same reference symbols represent the same components;

[0055] Figure 1 This is a flow chart of a method for distributing and scheduling charging power for electric vehicles based on a water filling algorithm according to an embodiment of the present invention;

[0056] Figure 2 A schematic diagram of the predicted basic load of a residential area according to an embodiment of the present invention;

[0057] Figure 3 This is a schematic diagram comparing the orderly charging scheduling of electric vehicles in a residential area and the basic load according to an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0059] A specific embodiment of the present invention discloses a method for distributing and scheduling charging power of electric vehicles based on a water filling algorithm. Figure 1 As shown, the method comprises the following steps:

[0060] Step S1, collecting historical load data of the distribution network of the target residential area, and selecting corresponding data from the historical load data of the distribution network according to the current date to construct a first data set;

[0061] Step S2: constructing a prediction model based on Gaussian process regression, and training the prediction model based on the first data set to obtain a trained prediction model;

[0062] Step S3, dividing the current date into a time point sequence based on a preset time interval to construct a second data set, and inputting the second data set into a trained prediction model to obtain a daily basic load curve predicted on the current date of the residential community distribution network;

[0063] Step S4, setting the valley period according to the predicted daily basic load curve, taking the minimum variance of the total load of the distribution network as the objective function, taking the capacity of the distribution network transformer and the charge state of the electric vehicle as the constraints, and using the water filling algorithm to generate the optimal charging power distribution scheduling plan for the electric vehicle on the current date.

[0064] Specifically, in step S1, the historical load data of the target residential area distribution network includes a historical daily load curve, and the daily load curve includes a daily total load curve and a daily electric vehicle charging load curve.

[0065] It should be noted that the daily load curve represents the curve of the distribution network load changing with time during the day. The daily maximum load, daily minimum load, daily average load, and daily peak-to-valley difference can be obtained from the daily load curve. The daily total load curve represents the curve of the total load of the distribution network changing with time during the day. The total load refers to the sum of the total active power, transformer ratio, and electric vehicle (EV) charging power of the residential community distribution network. The electric vehicle charging load is obtained from the parking time period and charging power of the electric vehicle. The daily total active power, transformer ratio, transformer capacity, electric vehicle parking time period, and electric vehicle charging power of the residential community distribution network are obtained through intelligent terminal measurement equipment, so as to obtain the daily total load curve and daily electric vehicle charging load curve of the distribution network history.

[0066] It can be understood that the electric vehicle has a corresponding charging load during the parking period, and the charging load outside the parking period is 0, thereby obtaining a corresponding daily electric vehicle charging load curve for the electric vehicle.

[0067] Furthermore, the corresponding daily basic load curve is obtained by subtracting the corresponding daily electric vehicle charging load curve from the historical daily total load curve of the distribution network.

[0068] It should be noted that the basic load refers to the sum of the total active power and transformer ratio of the residential community distribution network. The above operation is used to distinguish the basic power load of the residential community distribution network from the electric vehicle charging load.

[0069] Furthermore, selecting corresponding data from the distribution network historical load data according to the current date to construct the first data set includes:

[0070] Divide the historical daily base load curve of the distribution network into “weeks”;

[0071] According to the current date, the corresponding "weekday" is obtained, and the daily basic load curves of the same "weekday" in the previous three weeks constitute the set corresponding to the current date;

[0072] For the daily basic load curve of each "weekday" in the set corresponding to the current date: divide it according to the preset time interval to obtain a time point sequence and its corresponding basic load sequence, thereby forming a first data set corresponding to the current date.

[0073] Exemplarily, the daily basic load curves of the distribution network in the previous three weeks are divided into "Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday" and "Sunday"; according to the current date, the corresponding "Sunday" is "Monday", and the daily basic load curves of "Monday" last week, "Monday the week before" and "Monday" three weeks ago constitute the set corresponding to the current date; for the daily basic load curve of "Monday" last week: divide it according to time intervals of 15 minutes to obtain a time point sequence containing 96 time points and its corresponding basic load sequence; similarly, divide the daily basic load curves of "Monday" the week before and three weeks ago to obtain a time point sequence containing 96 time points and its corresponding basic load sequence respectively; the time point sequence obtained from the daily basic load curves of "Monday" last week, "Monday the week before" and "Monday" three weeks ago and its corresponding basic load sequence constitute the first data set corresponding to the current date.

[0074] Furthermore, the first data set is divided into a training set and a validation set, wherein the training set is composed of a time point sequence and its corresponding basic load sequence obtained from the daily basic load curve of the "Sunday" last week and the week before last, and a sample includes a time point sequence and its corresponding basic load sequence obtained from the daily basic load curve of a "Sunday"; the validation set is composed of a time point sequence and its corresponding basic load sequence obtained from the daily basic load curve of the "Sunday" three weeks ago.

[0075] It can be understood that the distribution network load is greatly affected by weather factors such as temperature, humidity, sunshine duration and holidays. The present invention uses the historical load data of the "same day" in the previous three weeks as the training set and verification set, which is similar to the weather factors and holiday conditions of the current date, thereby improving the accuracy and efficiency of the prediction while using less historical data.

[0076] Specifically, in step S2, the expression of the prediction model is:

[0077] , (1)

[0078] in, is the basic load vector output by the prediction model for the training set, is Gaussian noise, is the basic load vector predicted by the prediction model for the training set, is the input training set time point sequence vector, is the time point sequence vector corresponding to the output predicted basic load vector, is the mean function, represents the mean value, is the covariance function, represents the Gaussian process regression model, is the input vector The variance of is the iteration step size of the covariance function.

[0079] It should be noted that the Gaussian process regression model is defined by the mean function and the covariance function. In the above formula, is Gaussian noise that conforms to the normal distribution. Before training the prediction model, the data in the first data set needs to be preprocessed so that its mean function is equal to 0, that is, , to simplify calculations and improve the accuracy of model predictions.

[0080] Furthermore, the prior distribution of the prediction model is:

[0081] , (2)

[0082] in, is an n-dimensional symmetric positive definite matrix, is the covariance matrix of the noise, is the n-dimensional identity matrix, where n is the dimension of the input vector, represents a normal distribution, for and The similarity between the corresponding loads, is the i-th time point of the input training set, is the jth time point of the input training set.

[0083] It should be noted that and The higher the similarity between the corresponding loads, The larger the value of .

[0084] Furthermore, the hyperparameters in Formula 2 are solved by the following formula: , , :

[0085] , (3)

[0086] in, for The target value is to minimize The corresponding value; is the probability density function, is a hyperparameter, , For function calls, is the dimension of the input vector, stands for minimizing the negative log-likelihood function.

[0087] Specifically, the partial derivative of the above formula 3 is obtained, and the conjugate gradient method is used to obtain the extreme point, so as to solve the value of the hyperparameter:

[0088] , (4)

[0089] in, for The transposed matrix of .

[0090] Furthermore, the posterior distribution of the prediction model is:

[0091] , (5)

[0092] in, is the basic load vector output by the prediction model for the validation set, is the input validation set time point sequence vector, for The mean of for The variance of for The transposed matrix of Represents the inverse matrix.

[0093] It should be noted that since any finite-dimensional subset of random variables in the Gaussian process follows a joint normal distribution, the joint distribution of the training set, the validation set, and their corresponding predicted values ​​conforms to the joint normal distribution.

[0094] Further, the training the prediction model based on the first data set to obtain a trained prediction model includes:

[0095] The time point sequence in the training set of the first data set is input into the prediction model, and the predicted basic load corresponding to each time point is obtained by formula 2; based on the predicted basic load, the optimal hyperparameters are solved by minimizing the negative log-likelihood function by formulas 3 and 4 to determine the kernel function of the prediction model, thereby obtaining the trained prediction model; the time point sequence in the validation set of the first data set is input into the trained prediction model, and the predicted basic load corresponding to each time point is obtained by formula 2, and the posterior distribution is calculated by formula 5. If the confidence threshold is reached, the trained prediction model is obtained.

[0096] It should be noted that the statistical experimental results show that the confidence level of the base load prediction using the above-mentioned trained prediction model can reach more than 95%.

[0097] It can be understood that the present invention predicts the basic load of the distribution network based on the historical data of the previous three weeks by adopting a Gaussian regression prediction model and a "same-day" prediction method, so that the prediction model makes predictions based on the latest information, thereby improving the accuracy and efficiency of the prediction while using less historical data, thereby providing a basis for achieving optimal charging scheduling for electric vehicles.

[0098] Specifically, in step S3, the corresponding "day of the week" is obtained according to the current date, and the current date is divided into a time point sequence according to preset time intervals, thereby forming a second data set corresponding to the current date.

[0099] It should be noted that the time interval preset in step S3 is the same as that in step S2, and therefore, the time point sequence in the second data set is the same as the time point sequence of a sample in the first data set.

[0100] Furthermore, the second data set is input into the trained prediction model to obtain the predicted basic load sequence corresponding to the time point sequence in the second data set, and the predicted basic load sequence is fitted into a curve to obtain the daily basic load curve predicted on the current date of the residential community distribution network.

[0101] It should be noted that the predicted basic load sequence can be linearly interpolated to fit the daily basic load curve through existing linear interpolation methods, such as Lagrange interpolation method, polynomial interpolation method, and tri-spline interpolation method.

[0102] For example, the above method is used to predict the basic load of a residential area distribution network on a certain day. The prediction result is as follows: Figure 2 As shown, the maximum relative error between the true value and the predicted value is less than 8%. The predicted basic load sequence is fitted into a curve by the three-spline interpolation method to obtain the daily basic load curve predicted for the residential community distribution network on that day.

[0103] Specifically, in step S4, setting the valley period according to the predicted daily basic load curve includes:

[0104] Setting a valley period threshold, and setting the start time and end time of each valley period according to the predicted daily basic load curve and the valley period threshold, so as to obtain each valley period corresponding to the current date;

[0105] For each valley period, a plurality of sub-intervals are obtained according to preset time intervals.

[0106] Specifically, the valley period is expressed by the following formula:

[0107] , (6)

[0108] in, Represents the preset time interval, represents the kth subinterval.

[0109] It should be noted that the time interval preset in step S4 is the same as that in step S2. The valley period is subdivided into K sub-intervals according to the time interval, where 0-(K-1) is the index number of the sub-interval. If the length of the sub-interval is less than the time interval during the division, the length of the sub-interval is supplemented by extending the valley period to ensure that the lengths of the sub-intervals are equal.

[0110] For example, the valley period threshold is set to 3200kW. Figure 3 The predicted daily basic load curve and valley period threshold shown in the figure set the valley period start time as 00:00 and end time as 06:50, thus obtaining the valley period corresponding to the day. For the valley period, 28 sub-intervals are obtained at 15-minute time intervals, of which the length of the last sub-interval is less than 15 minutes, so the end time of the extended valley period is 07:00, and the length of the last sub-interval is supplemented to 15 minutes.

[0111] It can be understood that the present invention flexibly sets the valley period based on the prediction of the distribution network load, and solves the electric vehicle charging power allocation algorithm based on the changing valley period, thereby improving the stability of the distribution network power supply while meeting the user's charging needs.

[0112] Furthermore, when the electric vehicle is connected to the residential community distribution network through the charging pile, the initial state of charge (SOC), battery charging efficiency, battery capacity, and arrival time of the electric vehicle are obtained through the charging pile; at the same time, the user is prompted to enter the following information: the selected charging mode, the expected state of charge, the expected departure time, and whether he is willing to participate in the orderly charging scheduling.

[0113] Specifically, electric vehicles that are willing to participate in orderly charging scheduling are placed in a charging waiting sequence, and for each electric vehicle in the charging waiting sequence: the shortest charging time is obtained according to the initial state of charge of the electric vehicle, the expected state of charge, the battery charging efficiency, and the maximum charging power; if the shortest charging time is greater than the length of the corresponding valley period, the electric vehicle is charged at the maximum charging power during the corresponding valley period and outside the valley period according to the shortest charging time; otherwise, the water filling algorithm is used to solve the objective function to obtain the optimal charging power allocation scheduling plan.

[0114] It should be noted that the parking time of the electric vehicle is obtained according to the arrival time and departure time of the electric vehicle, and the portion of the parking time of the electric vehicle that overlaps with the valley period corresponding to the current date is the valley period corresponding to the electric vehicle. If the shortest charging time of the electric vehicle is greater than its corresponding valley period, it means that its corresponding valley period cannot meet its charging demand, and it is necessary to charge at the maximum charging power during the entire valley period corresponding to it, and it is also necessary to charge at the maximum charging power outside the valley period according to its shortest charging time. When a new electric vehicle is connected, it is determined whether to put it into the charging waiting sequence according to its willingness to participate; when the electric vehicle in the charging waiting sequence leaves, it is deleted from the charging waiting sequence, so the electric vehicles in the charging waiting sequence are dynamically changing, and the minimum time interval of the change is the preset time interval; when the electric vehicles in the charging waiting sequence change, it is necessary to regenerate the allocation scheduling plan for the charging power of all electric vehicles therein according to the method provided by the present invention. It can be seen that the present invention requires an algorithm with fast execution speed and small calculation amount to solve the charging power distribution.

[0115] Specifically, the objective function is obtained by the following formula:

[0116] , (7)

[0117] in, is the charging power of the nth electric vehicle in the kth sub-interval, is the predicted base load for the kth sub-interval on the current date, N is the total number of electric vehicles in the charging waiting sequence, is the sub-interval contained in the valley period corresponding to the nth electric car, is the arrival time of the nth electric car, represents the subinterval corresponding to the arrival time of the nth electric car, is the departure time of the nth electric car, It represents the subinterval corresponding to the departure time of the nth electric car.

[0118] It should be noted that the sub-interval corresponding to the electric vehicle is obtained according to the valley period corresponding to the electric vehicle and the sub-interval divided by the valley period; if the arrival time of the electric vehicle is within a sub-interval, the sub-interval corresponding to the arrival time of the electric vehicle is the next sub-interval; if the departure time of the electric vehicle is within a sub-interval, the sub-interval corresponding to the departure time of the electric vehicle is the previous sub-interval. The electric vehicle is set to start charging at the beginning time of the sub-interval corresponding to its arrival time. In the same sub-interval, its charging power remains constant; in different sub-intervals, its charging power can be different, ranging from 0 to varies between is the maximum charging power of the nth electric vehicle.

[0119] Exemplarily, assuming that the valley period corresponding to the current date is 0-8 o'clock, the valley period is divided into 32 sub-intervals at intervals of 15 minutes, and the sub-intervals have index numbers 0-31; assuming that the arrival time of an electric car is 1:25 and the departure time is 6:38, the arrival time of the electric car is within the 5th sub-interval [1:16, 1:30], then the sub-interval corresponding to the arrival time of the electric car is the 6th sub-interval; the departure time of the electric car is within the 26th sub-interval [1:16, 1:30], then the sub-interval corresponding to the departure time of the electric car is the 25th sub-interval.

[0120] Understandably, due to differences in battery capacity, charging mode and other factors, the maximum charging power supported by each electric vehicle is different.

[0121] Specifically, the constraint condition of the objective function is obtained by the following formula:

[0122] , (8)

[0123] in, is the desired state of charge of the nth electric vehicle, is the initial state of charge of the nth electric vehicle, is the charging efficiency of the nth electric vehicle, is the total charge demand of the nth electric vehicle.

[0124] It can be understood that the present invention reasonably allocates the charging power in each sub-interval corresponding to the electric vehicle in real time based on the actual arrival time and initial charge state of the electric vehicle, charges according to the allocated plan, and improves the stability of power supply of the distribution network while meeting the charging needs of the electric vehicle itself.

[0125] Furthermore, since the computational complexity increases with the increase of n and k, when solving formula (7), by separating the charging power of the electric vehicle in formula (7) and solving the charge distribution of each electric vehicle separately, we can obtain:

[0126] . (9)

[0127] Further, let is the total load in the kth subinterval, and its expression is:

[0128] . (10)

[0129] Further, let is the total load excluding the charging power of the nth electric vehicle in the kth subinterval, and its expression is:

[0130] . (11)

[0131] Furthermore, for each electric vehicle, its objective function is:

[0132] . (12)

[0133] Specifically, for each electric vehicle, the constraints of its objective function are:

[0134] . (13)

[0135] Furthermore, the method of using the water filling algorithm to generate the optimal charging power distribution scheduling plan for electric vehicles on the current date includes:

[0136] The water filling algorithm is used to solve the objective function of each electric vehicle in the charging waiting sequence to obtain the charging power in each sub-interval corresponding to each electric vehicle; the charging power in each sub-interval corresponding to all electric vehicles in the charging waiting sequence is the optimal charging power allocation scheduling plan for electric vehicles on the current date, and orderly charging is carried out according to the plan.

[0137] Specifically, the following steps are performed for each electric vehicle in the charging waiting sequence:

[0138] Step S401, initializing the error value, and randomly initializing the charging power in each sub-interval corresponding to the electric vehicle within the upper and lower limits of the charging power of the electric vehicle;

[0139] Step S402: obtaining a minimum value and a maximum value of a water filling variable based on an initial charging power in each sub-interval;

[0140] Step S403: if the difference between the maximum value and the minimum value of the water filling variable is less than or equal to the error value, then output the charging power in the subinterval corresponding to the electric vehicle; otherwise, execute step S404;

[0141] Step S404: Get the current water filling variable value through formula 16, and update the charging power in each sub-interval corresponding to the electric vehicle through formula 17. If the sum of the charging powers in each sub-interval corresponding to the electric vehicle after the update is greater than its total charge demand Then set , otherwise set ; Calculate the difference between the maximum and minimum values ​​of the water filling variable. If the difference between the maximum and minimum values ​​of the water filling variable is less than or equal to the error value, output the charging power in the subinterval corresponding to the electric vehicle; otherwise, repeat step S404 until the difference between the maximum and minimum values ​​of the water filling variable is less than or equal to the error value.

[0142] It should be noted that the electric vehicles are charged in an orderly manner according to the charging power within the sub-intervals corresponding to the output.

[0143] Specifically, the minimum value of the water filling variable is obtained by the following formula:

[0144] ; (14)

[0145] The maximum value of the water filling variable is obtained by the following formula:

[0146] , (15)

[0147] in, This is the transformer power constraint for the residential area's distribution network.

[0148] Specifically, the current water filling variable value is obtained by the following formula:

[0149] . (16)

[0150] Specifically, the charging power in each sub-interval corresponding to the electric vehicle is obtained by the following formula:

[0151] . (17)

[0152] It should be noted that the initial charging power in each sub-interval corresponding to each electric vehicle can be obtained based on the predicted daily basic load curve. .because It changes over time, so Need to be based on the current The value is adjusted; at each iteration, by updating and , get the right water level , and then according to the Value and current Worth it .

[0153] Furthermore, when updating the charging power in each subinterval corresponding to the electric vehicle, if the charging power calculated in the subinterval is less than zero, it is set to zero; if the charging power calculated in the subinterval is greater than , then set it to .

[0154] For example, the error value is set , there are 100 electric vehicles in the charging waiting sequence, based on Figure 2 The predicted basic load shown in the figure is used to obtain the total load curve after orderly charging scheduling of electric vehicles, as shown in Figure 3 As shown, the valley period is from 00:00 to 07:00. It can be seen that the method provided by the present invention can well fill the load gap in the valley period and realize the peak shaving and valley filling of the load curve.

[0155] It can be understood that the present invention flexibly sets the valley period based on the prediction of the distribution network load, and solves the electric vehicle charging power allocation algorithm based on the changing valley period, thereby improving the stability of the distribution network power supply while meeting the user's charging needs.

[0156] Compared with the prior art, the electric vehicle charging power allocation and scheduling method based on the water filling algorithm provided by the present invention has the following beneficial effects:

[0157] 1. The present invention adopts a water filling algorithm to distribute the charging power of electric vehicles. It does not require complex mathematical models and parameter settings, has a fast execution speed, and can support near real-time performance, solving the problem of large calculation amount and long time consumption of current electric vehicle charging distribution algorithms.

[0158] 2. The present invention uses the “same-day” prediction method to predict the basic load of the distribution network based on the historical data of the previous three weeks, so that the prediction model makes predictions based on the latest information, thereby improving the accuracy and efficiency of the prediction while using less historical data, and providing a basis for achieving the optimal charging scheduling of electric vehicles.

[0159] 3. The present invention flexibly sets the valley period based on the prediction of the distribution network load, and solves the electric vehicle charging power allocation algorithm based on the changing valley period, thereby improving the stability of the distribution network power supply while meeting the user's charging needs.

[0160] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0161] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for distributing and scheduling charging power of electric vehicles based on a water filling algorithm, characterized in that: The method comprises the following steps: Collect historical load data of the distribution network of the target residential area, and select corresponding data from the historical load data of the distribution network according to the current date to construct a first data set; Building a prediction model based on Gaussian process regression, and training the prediction model based on the first data set to obtain a trained prediction model; Dividing the current date into a time point sequence based on a preset time interval to construct a second data set, and inputting the second data set into a trained prediction model to obtain a daily basic load curve predicted on the current date of the residential community distribution network; The valley period is set according to the predicted daily basic load curve, the total load variance of the distribution network is minimized as the objective function, the capacity of the distribution network transformer and the state of charge of the electric vehicle are used as constraints, and the water filling algorithm is used to generate the optimal charging power distribution scheduling plan for the electric vehicle on the current date; The expression of the prediction model is: ， in, is the basic load vector output by the prediction model for the training set, is Gaussian noise, is the basic load vector predicted by the prediction model for the training set, is the input training set time point sequence vector, is the time point sequence vector corresponding to the output predicted basic load vector, is the mean function, represents the mean value, is the covariance function, represents the Gaussian process regression model, is the input vector The variance of is the iteration step size of the covariance function.

2. The electric vehicle charging power allocation and scheduling method based on the water filling algorithm according to claim 1 is characterized in that: The setting of the valley period according to the predicted daily basic load curve comprises: Setting a valley period threshold, and setting the start time and end time of each valley period according to the predicted daily basic load curve and the valley period threshold, so as to obtain each valley period corresponding to the current date; For each valley period, a plurality of sub-intervals are obtained according to preset time intervals.

3. The electric vehicle charging power allocation and scheduling method based on the water filling algorithm according to claim 1 is characterized in that: The objective function is obtained by the following formula: ， in, is the charging power of the nth electric vehicle in the kth subinterval, is the predicted base load for the kth sub-interval on the current date, N is the total number of electric vehicles in the charging waiting sequence, is the sub-interval contained in the valley period corresponding to the nth electric car, is the arrival time of the nth electric car, represents the subinterval corresponding to the arrival time of the nth electric car, is the departure time of the nth electric car, It represents the subinterval corresponding to the departure time of the nth electric car.

4. The electric vehicle charging power allocation and scheduling method based on the water filling algorithm according to claim 3 is characterized in that: The constraint conditions of the objective function are obtained by the following formula: ， in, is the desired state of charge of the nth electric vehicle, is the initial state of charge of the nth electric vehicle, is the charging efficiency of the nth electric vehicle, is the total charge demand of the nth electric vehicle.

5. The electric vehicle charging power allocation and scheduling method based on the water filling algorithm according to claim 2 is characterized in that: The method of using the water filling algorithm to generate the optimal charging power distribution scheduling plan for electric vehicles on the current date includes: The water filling algorithm is used to solve the objective function of each electric vehicle in the charging waiting sequence to obtain the charging power in each sub-interval corresponding to each electric vehicle; the charging power in each sub-interval corresponding to all electric vehicles in the charging waiting sequence is the optimal charging power allocation scheduling plan for electric vehicles on the current date, and orderly charging is carried out according to the plan.

6. The electric vehicle charging power allocation and scheduling method based on the water filling algorithm according to claim 1 is characterized in that: The selecting corresponding data from the distribution network historical load data according to the current date to construct the first data set comprises: Divide the historical daily basic load curve of the distribution network into "weeks"; According to the current date, the corresponding "weekday" is obtained, and the daily basic load curves of the same "weekday" in the previous three weeks constitute the set corresponding to the current date; For the daily basic load curve of each "weekday" in the set corresponding to the current date: divide it according to the preset time interval to obtain a time point sequence and its corresponding basic load sequence, thereby forming a first data set corresponding to the current date.

7. The electric vehicle charging power allocation and scheduling method based on the water filling algorithm according to claim 1 is characterized in that: The prior distribution of the prediction model is: ， in, is an n-dimensional symmetric positive definite matrix, is the covariance matrix of the noise, is the n-dimensional identity matrix, where n is the dimension of the input vector, represents a normal distribution, for and The similarity between the corresponding loads, is the i-th time point of the input training set, is the jth time point of the input training set.

8. The electric vehicle charging power allocation and scheduling method based on the water filling algorithm according to claim 7 is characterized in that: The hyperparameters are solved by the following formula , , : ， in, for The target value of , that is, the corresponding value when minimizing NLML value; is the probability density function, is a hyperparameter, , For function calls, is the dimension of the input vector, represents minimizing the negative log-likelihood function; Take the partial derivative of the above formula and use the conjugate gradient method to get the extreme point, so as to solve the value of the hyperparameter: ， in, for The transposed matrix of .

9. The electric vehicle charging power allocation and scheduling method based on the water filling algorithm according to claim 5 is characterized in that: Perform the following steps for each electric vehicle in the charging waiting sequence: Step S401, initializing the error value, and randomly initializing the charging power in each sub-interval corresponding to the electric vehicle within the upper and lower limits of the charging power of the electric vehicle; Step S402: obtaining a minimum value and a maximum value of a water filling variable based on the initial charging power in each sub-interval; Step S403: if the difference between the maximum value and the minimum value of the water filling variable is less than or equal to the error value, then output the charging power in the subinterval corresponding to the electric vehicle; otherwise, execute step S404; Step S404: Get the current water filling variable value through formula 16, and update the charging power in each sub-interval corresponding to the electric vehicle through formula 17. If the sum of the charging powers in each sub-interval corresponding to the electric vehicle after the update is greater than its total charge demand Then set , otherwise set ; Calculate the difference between the maximum value and the minimum value of the water filling variable. If the difference between the maximum value and the minimum value of the water filling variable is less than or equal to the error value, output the charging power in the subinterval corresponding to the electric vehicle; otherwise, repeat step S404 until the difference between the maximum value and the minimum value of the water filling variable is less than or equal to the error value; The current water filling variable value is obtained by the following formula: ， （16） in, Fill the variable value with the current water, is the maximum value of the water filling variable, Minimum value for water fill variable; The charging power in each sub-interval corresponding to the electric vehicle is obtained by the following formula: ， （17） in, is the charging power of the nth electric vehicle in the kth subinterval, is the total load excluding the charging power of the nth electric vehicle in the kth subinterval.

Images