A multi-energy virtual power plant optimization scheduling method and system

Abstract

The application relates to a multi-energy virtual power plant optimization scheduling method and system, belongs to the technical field of power optimization scheduling, and solves the problems of incapability of a single prediction model to meet the demand of multi-energy virtual power plant optimization scheduling, a large data set used in prediction calculation, and high operation cost; the application adopts a random forest algorithm to screen influencing factor indexes of wind and light electric set output, uses an artificial neural network algorithm to predict the wind and light electric set output, further makes up for the deficiency of a single algorithm through a hybrid prediction algorithm, and improves the prediction accuracy; a Monte Carlo method is used for simulation analysis for load prediction of electric vehicles; an optimization scheduling model of the multi-energy virtual power plant is established by taking the minimum virtual power plant operation cost as an objective function, and the problem that only a single distributed energy can be predicted and optimized in the prior art is solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power optimization scheduling, and relates to a multi-energy virtual power plant optimization scheduling method and system. BACKGROUND

[0002] The core idea of a virtual power plant is to gather various types of distributed adjustable power sources and loads, form a virtual "power plant" through digital means for unified management and scheduling, and participate in the electricity market as a main body. With the development of renewable energy, more and more wind farms and photovoltaic power stations are integrated into virtual power plants, and mobile electric vehicles are also connected as a kind of energy storage carrier. In a multi-energy virtual power plant, in order to realize efficient use and flexible configuration of energy, various distributed energy sources such as distributed power generation equipment and energy storage equipment need to be coordinated and optimized. However, due to the strong uncertainty of wind power generation and photovoltaic power generation, how to accurately predict the power generation output and load of the virtual power plant has become one of the current optimization scheduling problems.

[0003] Due to the intermittency and volatility of multi-energy, power quality and economic efficiency of power generation have become difficult points, so in the process of controlling various distributed energy generation equipment, energy storage systems and controllable loads in the virtual power plant, their coordinated control is the most critical part. For example, the invention patent with the application publication number CN117196234A discloses a two-stage scheduling method for a virtual power plant integrated with renewable energy and electric vehicles, which realizes the best bidding of the virtual power plant in the day-ahead market and the optimal real-time scheduling of cold and heat load, and facilitates the reasonable distribution of cold and heat energy in the virtual power plant.

[0004] At present, many mature algorithms have been formed in the computer field for numerical prediction, and have a high prediction accuracy. Based on the meteorological, regional and other related data collected by the virtual power plant, numerical prediction algorithms can be used to predict the load and power generation in the system. The commonly used prediction algorithms at present include the ARIMA model based on statistical model, the support vector regression (SVR) model based on machine learning, and the artificial neural network (ANN) model.

[0005] The existing technology for optimizing the scheduling of a virtual power plant through machine learning mainly predicts a certain type of distributed energy source. However, with more and more types of distributed energy sources being connected to the virtual power plant, a single prediction model cannot meet the needs of multi-energy virtual power plant optimization scheduling, and a more comprehensive analysis system is needed to predict power generation and load. At the same time, a single prediction model lacks feature selection, resulting in a large data set used in prediction calculation and high computational cost. Therefore, there is an urgent need for a method that can combine multiple algorithms to optimize the scheduling of multiple types of distributed energy sources in a virtual power plant, improve the efficiency and reliability of the energy system. SUMMARY

[0006] The technical scheme of the present application is used to solve the problem that the existing single prediction model cannot meet the demand of multi-energy virtual power plant optimal scheduling, and the data set used in prediction calculation is large and the operation cost is high.

[0007] The present application solves the above technical problems by the following technical scheme:

[0008] A multi-energy virtual power plant optimal scheduling method, comprising the following steps:

[0009] S1, collect wind turbine output influencing factor data, photovoltaic turbine output influencing factor data and electric vehicle related data, and perform data cleaning, abnormal value processing and missing value processing on the collected data and normalize;

[0010] S2, the random forest algorithm is used to screen the wind and photovoltaic turbine output influencing factors respectively, and the main factor indexes required for prediction are obtained respectively;

[0011] S3, according to the main factor indexes screened in step S2, a wind and photovoltaic turbine output prediction model based on artificial neural network is established respectively;

[0012] S4, the Monte Carlo method is used to simulate the unordered charging load of electric vehicles;

[0013] S5, taking the maximization of virtual power plant economic benefits as the objective function, constructing a constraint model of each module, and establishing a multi-energy virtual power plant optimal scheduling model;

[0014] S6, according to the data calculated by the model, an optimal scheme of virtual power plant energy scheduling is formulated.

[0015] Further, the wind turbine output influencing factor data includes wind direction, wind speed, humidity, temperature, atmospheric pressure and air density data; the photovoltaic turbine output influencing factor data includes wind speed, humidity, temperature, solar direct radiation intensity, scattered radiation intensity, total radiation intensity, weather type and temperature; and the electric vehicle related data includes departure time, home charging time, daily driving distance, charging and discharging power and time length.

[0016] Further, in step S2, the random forest algorithm is used to screen the wind and photovoltaic turbine output influencing factors respectively, and the main factor indexes required for prediction are obtained respectively, which specifically includes:

[0017] S21, the historical data set of wind and photovoltaic turbine output influencing factors is taken as a training set respectively, Bootstraping method is used for random sampling with replacement, m samples are selected, n samplings are performed, n training sets are generated, and n decision tree models are trained respectively for the n training sets;

[0018] S22, calculate the out-of-bag data error errOOB1 corresponding to each decision tree; add noise interference to the features X of all samples of the out-of-bag data OOB at random, and calculate the out-of-bag data error errOOB2 again to calculate the feature importance of the features X;

[0019] S23, remove the corresponding proportion of features according to the feature importance to obtain a new feature set, and repeat the above process until p features are left;

[0020] S24, calculate the out-of-bag error rate corresponding to each feature set, and select the feature set with the lowest out-of-bag error rate as the main factor index; the out-of-bag error rate is the proportion of the out-of-bag samples that are misclassified.

[0021] Further, the importance of the features X in step S22 is expressed as:

[0022] Feature importance = ∑(errOOB2-errOOB1) / N

[0023] In the formula, N is the number of decision trees in the random forest.

[0024] Further, the wind and light electric machine group output prediction model based on artificial neural network established according to the main factor index selected in step S2 in step S3 is specifically:

[0025] The feature set selected in step S2 is selected as the input variable of the artificial neural network model, and the output variable is the power generation of the corresponding unit. 80% of the data set is selected as the training set to train the neural network. The artificial neural network parameters are continuously adjusted through the back propagation algorithm. The gradient of the neural network parameters is calculated through the cost function to adjust the weights and biases of the neural network. 20% of the data set is selected as the validation set, and the trained model is evaluated by the validation set or cross-validation method. The model is optimized according to the evaluation result.

[0026] Further, the Monte Carlo method is used to simulate the disordered charging load of the electric vehicle in step S4, which is specifically:

[0027] S41, fit the electric vehicle travel characteristics, the electric vehicle travel characteristics including the distribution of electric vehicle arrival and departure time, electric vehicle charging and discharging power and time;

[0028] S42, based on the fitted electric vehicle travel characteristics and data, randomly extract the initial parameters of the electric vehicle by using the Monte Carlo method, and set the scale of the electric vehicle;

[0029] S43, according to the electric vehicle return home time density function simulation return home and grid connected with travel off-grid time, using the daily mileage density function simulation mileage to calculate the electric vehicle power consumption, and then calculate the charging time of a single electric vehicle;

[0030] S44, the steps S42 to S43 are cycled until the number of simulated electric vehicles reaches the set scale value, and then the final total charging load is accumulated.

[0031] Further, in step S5, a virtual power plant economic benefit maximization is taken as a target function, a constraint model of each module is constructed, and a multi-energy virtual power plant optimization scheduling model is established.

[0032] S51, taking the minimum virtual power plant operation cost as the target, a target function is constructed.

[0033] S52, a constraint condition model is established, and the constraint condition model includes wind, light, and unit output constraint, gas unit constraint, energy storage constraint, electric vehicle constraint, grid interaction constraint, and balance constraint.

[0034] Further, in step S51, the target function expression is:

[0035] min Cost=C abondon +C gas +C w +C pv +C storage +C ev +C g

[0036] In the formula, min Cost is the target function, C abondon is the penalty cost of abandoned light and wind, C gas is the gas unit power generation cost, C w is the wind turbine power generation cost, C pv is the light power generation cost, C ev is the electric vehicle scheduling cost, C storage is the energy storage charging and discharging cost, C g is the grid interaction cost.

[0037] Further, in step S52, the balance constraint expression is:

[0038]

[0039] In the formula, is the virtual power plant day-ahead planning output at t, P w,t is the wind turbine power generation output power at t, P pv,t is the wind turbine power generation output power at t, P gas,tP is the power of the gas turbine generator at time t dis,t P is the power of the energy storage discharge at time t, P is the power bought and sold by the virtual power plant to the grid at time t, P is the power of the electric vehicle discharge at time t.

[0040] The application also provides a multi-energy virtual power plant optimization scheduling system, comprising a data acquisition module, an influence factor screening module, a prediction model construction module, a load simulation module, a scheduling model construction module and a scheduling scheme formulation module.

[0041] The data acquisition module is used to collect wind turbine output influence factor data, photovoltaic turbine output influence factor data and electric vehicle related data, and to perform data cleaning, abnormal value and missing value processing and normalization on the collected data.

[0042] The influence factor screening module is used to screen the wind and photovoltaic turbine output influence factors using a random forest algorithm, respectively, to obtain the main factor indicators required for prediction.

[0043] The prediction model construction module is used to establish wind and photovoltaic turbine output prediction models based on artificial neural networks according to the screened main factor indicators.

[0044] The load simulation module is used to simulate the disordered charging load of electric vehicles using the Monte Carlo method.

[0045] The scheduling model construction module is used to construct constraint models of each module and establish a multi-energy virtual power plant optimization scheduling model with the maximum economic benefit of the virtual power plant as the objective function.

[0046] The scheduling scheme formulation module is used to formulate an optimal energy scheduling scheme for the virtual power plant according to the data calculated by the model.

[0047] The application has the following advantages:

[0048] (1) The application screens the influence factor indicators of wind and photovoltaic turbine output using a random forest algorithm, eliminates indicators with low classification effect, and uses an artificial neural network algorithm to predict the wind and photovoltaic turbine output, and trains the wind and photovoltaic turbine output based on the random forest feature selection artificial neural network prediction model according to the data of the wind and photovoltaic turbines. The hybrid prediction algorithm further makes up for the shortcomings of a single algorithm to improve the prediction accuracy. Feature selection reduces the data set used in calculation, screens and simplifies the influence factor indicators of wind and photovoltaic turbine output, reduces the operation cost, simplifies the redundant analysis indicators of the prediction model in the prior art, and improves the model applicability.

[0049] (2) This invention uses the Monte Carlo method to simulate and analyze the load forecast of electric vehicles. It establishes an optimal scheduling model for a multi-energy virtual power plant with the objective function of minimizing the operating cost of the virtual power plant, which includes wind, solar, electric vehicles, gas turbines and energy storage equipment. This solves the problem that existing technologies can only predict and optimize for a single distributed energy source, improves the applicability of the model, reduces the cost of virtual power plant data collection and processing, and finally realizes the optimal scheduling of multi-energy virtual power plants based on the model. Attached Figure Description

[0050] Figure 1 This is a flowchart of the multi-energy virtual power plant optimization scheduling method according to an embodiment of the present invention;

[0051] Figure 2 This is a flowchart illustrating the process of using the random forest algorithm to filter feature sets according to an embodiment of the present invention;

[0052] Figure 3 This is a schematic diagram of the process of simulating electric vehicle charging load using the Monte Carlo method according to an embodiment of the present invention;

[0053] Figure 4 This is a schematic diagram of the multi-energy virtual power plant optimization scheduling model constructed according to an embodiment of the present invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0056] Example 1

[0057] like Figure 1 As shown, specifically, a multi-energy virtual power plant optimization scheduling method is disclosed, including the following steps:

[0058] S1. Collect data on factors affecting the output of wind turbines, solar turbines, and electric vehicles. Clean the collected data, remove outliers and missing values, and normalize them.

[0059] The data on factors affecting the output of the wind turbine include: wind direction, wind speed, humidity, temperature, atmospheric pressure, and air density.

[0060] The data on factors affecting the output of the photovoltaic generator include: wind speed, humidity, temperature, direct solar radiation intensity, diffuse radiation intensity, total radiation intensity, weather type, and temperature;

[0061] The electric vehicle-related data includes: departure time, return time for charging, daily mileage, charging / discharging power and duration.

[0062] S2. The random forest algorithm (RF) is used to screen the factors affecting the output of wind and solar generators, and the main factor indicators required for prediction are obtained respectively.

[0063] S21. Using the historical datasets of factors affecting the output of wind and solar generators as training sets, the Bootstraping method is used to perform random sampling with replacement, select m samples, perform n samplings in total, generate n training sets, and train n decision tree models for the n training sets respectively.

[0064] like Figure 2 As shown, in the above sampling method, a data point is selected from the training set each time and then put back into the original training set. This ensures that the selection opportunities of all data points in the original training dataset remain unchanged after each sampling. This means that the same data point may be selected multiple times in a bootstrap sample, which can improve the stability of the model when the sample size is limited or the data fluctuates greatly.

[0065] S22. Calculate the out-of-bag error errOOB1 for each decision tree corresponding to the out-of-bag data (OOB); randomly add noise interference to the feature X of all samples of the out-of-bag data (OOB), calculate the out-of-bag error errOOB2 again, and calculate the feature importance of feature X.

[0066] Out-of-bag data refers to the data obtained through repeated sampling when building a decision tree. Approximately one-third of the data remains unused and not used in the tree's construction. This unused data can be used to evaluate the decision tree's performance and calculate the model's prediction error rate, known as out-of-bag error.

[0067] Specifically, in step S22, for each decision tree, the out-of-bag (OOB) error is calculated and denoted as errOOB1. Noise interference is randomly added to the feature X of all samples in the OOB data, i.e., the value of all samples at feature X is randomly changed, and the OOB error is calculated again and denoted as errOOB2. Assuming there are N decision trees in the random forest, the importance expression for feature X is:

[0068] Feature importance = ∑(errOOB2 - errOOB1) / N

[0069] S23, according to the feature importance, a corresponding proportion of features is removed to obtain a new feature set, and the above process is repeated until p features are left, p being a preset value;

[0070] Since the greater the value of the feature importance, the greater the contribution of the feature to the prediction ability of the model, that is, the more important the feature. Therefore, by setting a threshold of feature importance, the features with feature importance higher than the threshold are retained, and the features with feature importance lower than the threshold are removed.

[0071] S24, calculating the out-of-bag error rate corresponding to each feature set, and screening the feature set with the lowest out-of-bag error rate as the main factor index; the out-of-bag error rate is the proportion of the out-of-bag samples that are misclassified;

[0072] According to the feature sets of the wind and photovoltaic generator output respectively obtained in the above process, and the out-of-bag error rate corresponding to the feature set, the feature set with the lowest out-of-bag error rate is selected;

[0073] For each sample, the classification probability of the sample can be estimated by the proportion of the sample that is correctly classified by multiple decision trees. For a classification problem, the out-of-bag error rate is the proportion of the out-of-bag samples that are misclassified.

[0074] S3, according to the main factor index screened in step S2, respectively establishing a wind and photovoltaic generator output prediction model based on artificial neural network (ANN);

[0075] Specifically, the feature set screened in step S2 is selected as the model input variable of the artificial neural network (ANN), and the output variable is the power generation of the corresponding generator set. 80% of the data set is selected as the training set to train the neural network. The artificial neural network parameters are continuously adjusted through the backpropagation algorithm (backpropagation). The gradient of the neural network parameters is calculated through the cost function to adjust the weights and biases of the neural network, thereby optimizing the performance of the network so that the model can better fit the data.

[0076] 20% of the data set is selected as the validation set, and the trained model is evaluated using the validation set or cross-validation method. According to the evaluation result, the model is optimized to improve the generalization ability of the model.

[0077] Through the above steps, the wind and photovoltaic generator output prediction model based on random forest feature selection of artificial neural network is trained according to the data of wind and photovoltaic generator. Through feature selection, the data set used for calculation is reduced, the influence factor index of wind and photovoltaic generator output is screened and simplified, the operation cost is reduced, the redundant analysis index of the existing technology for the prediction model is simplified, and the applicability of the model is improved.

[0078] S4, using Monte Carlo method to simulate the unordered charging load of electric vehicles, for reflecting the charging load of electric vehicles in the region, facilitating optimal scheduling; as shown in Figure 3 the following steps are included:

[0079] S41, fitting electric vehicle travel characteristics, the electric vehicle travel characteristics including electric car home and away time distribution, electric vehicle charging and discharging power and time;

[0080] Based on a private car travel behavior report to determine the electric vehicle (EV) travel characteristics, according to the above-mentioned optimization instructions.

[0081] The last time the electric car user returns home, that is, the probability density function f b t is:

[0082]

[0083] In the formula, t is the time constant, σ b is the sample standard deviation of the electric car home time, μ b is the sample mean of the electric car home time.

[0084] The electric car user travel time, the electric car away from home time probability density function is:

[0085]

[0086] In the formula, σ l is the sample standard deviation of the electric car away time, μ l is the sample mean of the electric car away time.

[0087] The probability density function of the daily travel mileage of electric vehicles is:

[0088]

[0089] In the formula, S x is the daily mileage, σ d is the sample standard deviation of the daily travel mileage of electric vehicles, μ x is the sample mean of the daily travel mileage of electric vehicles.

[0090] The charging and discharging power expression of electric vehicles is:

[0091]

[0092] In the formula, The charging and discharging power of the i-th electric vehicle is respectively, and the charging and discharging power of the single electric vehicle is accumulated. respectively, taking 1 indicates the i-th electric vehicle charging and discharging, otherwise taking 0; n is the number of electric vehicles participating in charging and discharging at this time.

[0093] Discharge duration of the electric vehicle that can accept the discharging scheduling electric quantity The calculation formula is:

[0094]

[0095] In the formula, is the discharging duration of the electric vehicle, is the dispatchable discharging state of charge of the electric vehicle, is the rated capacity of the battery of the electric vehicle, is the discharging efficiency of the electric vehicle, is the discharging power of the electric vehicle.

[0096] Dispatchable discharging state of charge of the electric vehicle The calculation formula is:

[0097]

[0098]

[0099] In the formula, respectively, the state of charge of the electric vehicle connected to the grid, the initial state of charge, S is the daily mileage of the electric vehicle, and a is the EV power consumption per 100 kilometers, is the minimum state of charge of the electric vehicle accepting discharging scheduling.

[0100] Further, the charging duration of the electric vehicle is divided into two cases:

[0101] For the electric vehicle not participating in discharging, the charging duration is calculated as follows:

[0102]

[0103] In the formula, represents the expected state of charge after the electric vehicle finishes charging; represents the charging efficiency of the electric vehicle.

[0104] For the electric vehicle participating in discharging, the charging duration is calculated as follows:

[0105]

[0106] S42, based on the fitted electric vehicle travel characteristics and data, the initial parameters of the electric vehicle are randomly extracted by using the Monte Carlo method, and the scale number of the electric vehicle is set.

[0107] S43, according to the electric vehicle return home time density function simulation return home and grid connected with travel off-grid time, using the daily mileage density function simulation mileage to calculate the electric vehicle power consumption, and then calculate the charging time of a single electric vehicle.

[0108] S44, the steps S42 to S43 are cycled until the number of simulated electric vehicles reaches the set scale value, and then the final total charging load is accumulated.

[0109] S5, taking the maximum economic benefit of the virtual power plant as the objective function, constructing a constraint model of each module, and establishing a multi-energy virtual power plant optimization scheduling model;

[0110] S51, taking the minimum virtual power plant operation cost as the target, constructing the target function; the target function min Cost expression is as follows:

[0111] min Cost=C abondon +C gas +C w +C pv +C storage +C ev +C g

[0112] In the formula, C abondon is the penalty cost of abandoned light and wind, C gas is the gas turbine generator cost, C w is the wind turbine generator cost, C pv is the photovoltaic generator cost, C ev is the electric vehicle scheduling cost, C storage is the storage charge and discharge cost, and C g is the grid interaction cost.

[0113] Specifically, the calculation formula of the abandoned light and wind penalty cost C abondon is:

[0114]

[0115] In the formula, P abw,t and P abpv,t are the abandoned wind and light power at t, p ab is the abandoned wind and light penalty unit price, and Δt is the unit time.

[0116] The calculation formula of the gas turbine generator cost C gas is:

[0117]

[0118] In the formula, P gas,t is the gas turbine generator output power at t, and pab Marginal cost of wind and solar power generation at time t.

[0119] The calculation formula of the wind and solar power generation cost is:

[0120]

[0121] In the formula, P w,t is the wind power generation output power at time t, P pv,t is the wind power generation output power at time t; p w,t and p pv,t are the marginal costs of wind and solar power generation at time t, respectively.

[0122] The electric vehicle scheduling cost is composed of the charging income of the electric vehicle participating in demand response and the discharging cost of the electric vehicle, so the electric vehicle scheduling cost C ev The calculation formula is:

[0123]

[0124] In the formula, C is the discharging compensation cost of the electric vehicle at time t, is the income of the electric vehicle participating in demand response at time t, is the discharging power of the i-th electric vehicle at time t, is the virtual power plant discharging compensation price of the electric vehicle at time t, N dis is the number of electric vehicles discharging at time t, are the charging loads before and after the implementation of demand response at time t, respectively, are the charging prices before and after the implementation of demand response at time t, respectively.

[0125] According to the demand theory of economics, when the price of a commodity increases at a certain time, the demand for the commodity in the market will decrease; conversely, when the price decreases, the demand for the commodity will increase; the same is true for the charging load. Since the charging demand of the electric vehicle is related to the electricity price at the time, an elasticity coefficient is introduced to represent the relationship between the charging demand of the electric vehicle and the electricity price and the relationship between the actual charging load demand and the real-time charging price:

[0126]

[0127] In the formula, ε st is the electricity demand price elasticity, s and t are two scheduling periods, s=t is the self-elasticity coefficient when s=t, and otherwise is the cross-elasticity coefficient.

[0128] The calculation formula of the energy storage charging and discharging cost C storage is:

[0129]

[0130] In the formula, P ch,t P dis,t Let p be the energy storage charging and discharging power at time t. ch,t p dis,t Let t be the cost of charging and discharging the stored energy.

[0131] The grid interaction cost is the cost incurred when a virtual power plant needs to purchase or sell electricity from the grid after a discrepancy exists between its actual and planned output. This grid interaction cost C... g The formula for calculation is:

[0132]

[0133] In the formula, Let t be the power that the virtual power plant buys from and sells to the grid. Let t be the price at which the virtual power plant buys and sells electricity to the grid.

[0134] S52, such as Figure 4 As shown, a constraint model is established based on the resources and configuration of each module. The constraint model includes constraints on wind and solar turbine output, gas turbine constraints, energy storage constraints, electric vehicle constraints, grid interaction constraints, and balance constraints.

[0135] The output constraints of the wind and solar turbine generators are as follows:

[0136]

[0137] In the formula, For the predicted output of the hybrid wind and solar turbine model at time t, μ α The coefficients for wind and solar power curtailment.

[0138] The constraints of the gas turbine unit are:

[0139]

[0140] In the formula, Let θ be the upper and lower limits of the gas turbine unit's output power at time t. h θ l This represents the ramp rate of the gas turbine unit per unit time.

[0141] The energy storage constraint is:

[0142]

[0143] The State of Charge (SOC) is calculated based on the charging and discharging state of an energy storage device over a specific period. It is related to the energy storage capacity, the device's charging and discharging power, and its efficiency. The formula for calculating energy storage capacity is:

[0144]

[0145] E = E + P * (1 - η) - P * η t S (t) = S (t - 1) + P * (1 - η) - P * η ch,t P (t) = P (t - 1) + P * (1 - η) - P * η dis,t η = η (t - 1) + η ch η = η (t - 1) - η dis P = P (t - 1) + P * (1 - η) - P * η

[0146] When the energy storage device is charging and discharging, its state of charge at time t is:

[0147]

[0148] Since the state of charge of the electric vehicle cannot exceed the maximum state of charge after charging, and cannot be less than the minimum state of charge after discharging, and the total discharge capacity of the electric vehicle should be less than the total dispatchable discharge capacity, and the total charge capacity cannot exceed the total chargeable capacity, and finally the state of charge at the end of charging and discharging should be between the minimum and maximum state of charge, the electric vehicle constraints are specifically

[0149]

[0150] S (t) = S (t - 1) + P * (1 - η) - P * η S (t) = S (t - 1) + P * (1 - η) - P * η = 1 if charging, otherwise 0 = 1 if discharging, otherwise 0 Smin, Smax are the minimum and maximum state of charge of the electric vehicle, η ch,ev η = η (t - 1) + η dis,ev η = η (t - 1) - η Sgrid, Sini are the state of charge of the electric vehicle and the initial state of charge, S is the daily mileage of the electric vehicle, and α is the power consumption per 100 kilometers of the EV.

[0151] The grid interaction constraints are:

[0152]

[0153] P (t) = P (t - 1) + P * (1 - η) - P * η line,t P (t) = P (t - 1) + P * (1 - η) - P * η t α is the tie line capacity between the virtual power plant and the grid at time t, and α t is the 0-1 constraint of grid interaction, which is used to limit the virtual power plant to only buy or sell electricity at the same time.

[0154] The virtual power plant balancing constraints are:

[0155]

[0156] P (t) = P (t - 1) + P * (1 - η) - P * η P (t) = P (t - 1) + P * (1 - η) - P * ηw,t P(t) is the wind turbine power output at time t pv,t P(t) is the wind turbine power output at time t gas,t P(t) is the gas turbine power output at time t dis,t P(t) is the energy storage discharge power at time t P(t) is the power bought or sold by the virtual power plant to the grid at time t P(t) is the electric vehicle discharge power at time t

[0157] S6, an optimal scheme of virtual power plant energy scheduling is formulated according to the data calculated by the model;

[0158] Specifically, according to the data of the multi-energy optimization scheduling model based on the mixed prediction algorithm constructed, the CPLEX solver in MATLAB is used to simulate and solve the multi-energy optimization scheduling model, and an optimal scheme of virtual power plant energy scheduling is formulated.

[0159] Further, other optimization algorithms can also be selected to simulate and solve the multi-energy optimization scheduling model, and since there are many other optimization algorithms in the related art, the present application will not be described in detail.

[0160] The present application adopts the random forest algorithm to screen the influencing factor indexes of wind and light turbine output, eliminate the indexes with low classification effect, and uses the artificial neural network algorithm to predict the wind and light turbine output, and trains the artificial neural network prediction model of wind and light turbine output based on random forest feature selection according to the data of wind and light turbine. The mixed prediction algorithm further makes up for the shortcomings of single algorithm to improve the prediction accuracy; the feature selection reduces the data set used in calculation, screens and simplifies the influencing factor indexes of wind and light turbine output, reduces the operation cost, simplifies the redundant analysis indexes of the prediction model in the prior art, and improves the model applicability.

[0161] The present application adopts the Monte Carlo method for simulation analysis for the load prediction of electric vehicles, establishes an optimization scheduling model of a multi-energy virtual power plant containing wind, light, electric vehicles, gas turbines and energy storage devices with the minimum virtual power plant operation cost as the objective function, solves the problem that only single distributed energy can be predicted and optimized in the prior art, improves the model applicability, reduces the cost of virtual power plant data collection and processing, and finally realizes the optimization scheduling of the multi-energy virtual power plant according to the model.

[0162] The present application also provides a multi-energy virtual power plant optimization scheduling system, comprising a data acquisition module, an influencing factor screening module, a prediction model construction module, a load simulation module, a scheduling model construction module and a scheduling scheme formulation module.

[0163] The data collection module is used for collecting wind turbine output influencing factor data, photovoltaic turbine output influencing factor data and electric vehicle related data, performing data cleaning on the collected data, processing abnormal values and missing values and normalizing;

[0164] The influencing factor screening module is used for screening wind and photovoltaic turbine output influencing factors respectively by using a random forest algorithm, and obtaining main factor indexes required for prediction respectively;

[0165] The prediction model construction module is used for establishing wind and photovoltaic turbine output prediction models based on artificial neural network respectively according to the screened main factor indexes;

[0166] The load simulation module is used for simulating electric vehicle unordered charging load by using a Monte Carlo method;

[0167] The scheduling model construction module is used for constructing each module constraint model and establishing a multi-energy virtual power plant optimal scheduling model by taking maximum virtual power plant economic benefit as an objective function;

[0168] The scheduling scheme formulation module is used for formulating a virtual power plant energy scheduling optimal scheme according to data calculated by the model.

[0169] The above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced equivalently; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for multi-energy virtual power plant optimal scheduling, characterized in that, It comprises the following steps: S1, collecting wind turbine output influencing factor data, photovoltaic turbine output influencing factor data, electric vehicle related data, data cleaning, processing abnormal values and missing values and normalizing the collected data; S2, using a random forest algorithm to screen wind and photovoltaic turbine output influencing factors respectively, and obtaining the main factor indexes required for prediction; S3, according to the main factor indexes screened in step S2, respectively establishing wind and photovoltaic turbine output prediction models based on artificial neural network; S4, using Monte Carlo method to simulate the unordered charging load of electric vehicles; S5, taking the maximum economic benefit of the virtual power plant as the objective function, constructing the constraint model of each module, and establishing the optimization scheduling model of the multi-energy virtual power plant; Specifically: S51, taking the minimum virtual power plant operation cost as the target, constructing the objective function; The objective function expression is: In the formula, is a target function, is a light and wind penalty cost, is a gas turbine power generation cost, is a wind turbine power generation cost, is a photovoltaic power generation cost, is an electric vehicle scheduling cost, is a storage charging and discharging cost, is a grid interaction cost; The abandoned light, abandoned wind penalty cost The calculation formula is: In the formula, and respectively abandoned wind, abandoned light power at the moment, abandoned wind, abandoned light penalty unit price, unit time; The electric vehicle scheduling cost is composed of charging benefits of the electric vehicles participating in demand response and discharging cost of the electric vehicles The calculation formula is: In the formula, for The cost of electric vehicle discharge compensation at all times for Benefits of electric vehicles participating in demand response at any time for Time of the first The discharge power of an electric vehicle for The price of compensation for electric vehicle discharge from a virtual power plant at any given time. for Number of electric vehicles discharging at any given time , They are respectively Real-time monitoring of charging load before and after demand response. , They are respectively Implement charging prices before and after demand response in real time; S52, establishing a constraint condition model, which includes wind and photovoltaic turbine output constraints, gas turbine constraints, energy storage constraints, electric vehicle constraints, grid interaction constraints, and balance constraints; S6, according to the data calculated by the model, formulating the optimal scheme of virtual power plant energy scheduling.

2. The multi-energy virtual power plant optimal scheduling method according to claim 1, characterized in that, The wind turbine output influencing factor data includes wind direction, wind speed, humidity, temperature, atmospheric pressure, and air density data; The photovoltaic turbine output influencing factor data includes wind speed, humidity, temperature, solar direct radiation intensity, scattered radiation intensity, total radiation intensity, and weather type; The electric vehicle related data includes departure time, home charging time, daily driving distance, charging and discharging power and time.

3. The method of claim 1, wherein, In step S2, the random forest algorithm is used to screen wind and photovoltaic turbine output influencing factors respectively, and the main factor indexes required for prediction are obtained, which specifically comprises: S21, respectively taking the historical data set of wind and photovoltaic turbine output influencing factors as the training set, using Bootstraping method for random sampling with replacement, selecting m samples, performing n times of sampling, generating n training sets, and training n decision tree models for n training sets respectively; S22, calculate out-of-bag data error for each decision tree corresponding to the out-of-bag data OOB ; add noise interference to the features X of all samples of the out-of-bag data OOB at random, and calculate the out-of-bag data error again , calculate the feature importance of the features X; S23, according to the feature importance, the corresponding proportion of features is removed to obtain a new feature set, and the above process is repeated until p features are left; S24, calculating the out-of-bag error rate corresponding to each feature set, and selecting the feature set with the lowest out-of-bag error rate as the main factor index; The out-of-bag error rate is the proportion of misclassified samples in the out-of-bag samples.

4. The multi-energy virtual power plant optimal scheduling method according to claim 3, characterized in that, The importance expression of feature X in step S22 is: Wherein, N is the number of decision trees in the random forest.

5. The multi-energy virtual power plant optimal scheduling method according to claim 4, characterized in that, In step S3, according to the main factor indexes screened in step S2, the wind and photovoltaic turbine output prediction models based on artificial neural network are established, which specifically comprises: Select the feature set screened in step S2 as the input variable of the artificial neural network model, and the output variable is the power generation of the corresponding unit. 80% of the data set is selected as the training set to train the neural network. The artificial neural network parameters are continuously adjusted through the back propagation algorithm. The gradient of the neural network parameters is calculated by the cost function to adjust the weights and biases of the neural network. 20% of the data set is selected as the validation set. The trained model is evaluated by the validation set or cross-validation method. The model is optimized according to the evaluation results.

6. The method of claim 1, wherein, The Monte Carlo method is used to simulate the unordered charging load of the electric vehicle in step S4. Specifically: S41, fitting the electric vehicle travel characteristics, including the distribution of electric vehicle arrival and departure time, electric vehicle charging and discharging power and time; S42, based on the fitted electric vehicle travel characteristics and data, the initial parameters of the electric vehicle are randomly extracted by the Monte Carlo method, and the scale of the electric vehicle is set; S43, according to the electric vehicle return home time density function, the return home and off-grid time is simulated, the electric vehicle power consumption is calculated by the daily mileage density function, and the charging time of a single electric vehicle is calculated; S44, repeat steps S42 to S43 until the number of simulated electric vehicles reaches the set scale value, then the final total charging load is obtained.

7. The method of claim 1, wherein, In step S52, the balance constraint expression is: wherein, is the virtual power plant day-ahead schedule output at time t, is the wind turbine generator output power at time t, is the wind turbine generator output power at time t, is the gas turbine generator output power at time t, is the energy storage discharge power at time t, , is the virtual power plant buy / sell power to / from the grid at time t, is the electric vehicle discharge power at time t.

8. A multi-energy virtual power plant optimal scheduling system, characterized in that, It includes a data acquisition module, an influence factor screening module, a prediction model construction module, a load simulation module, a scheduling model construction module and a scheduling scheme development module; The data acquisition module is used to collect wind turbine output influence factor data, photovoltaic turbine output influence factor data and electric vehicle related data, and to clean the collected data, process abnormal values and missing values and normalize them; The influence factor screening module is used to screen the wind and photovoltaic turbine output influence factors respectively using the random forest algorithm, and to obtain the main factor indexes required for prediction respectively; The prediction model construction module is used to establish wind and photovoltaic turbine output prediction models based on artificial neural networks according to the screened main factor indexes; The load simulation module is used to simulate the unordered charging load of the electric vehicle by the Monte Carlo method; The scheduling model construction module is used to construct a multi-energy virtual power plant optimization scheduling model with the maximum economic benefit of the virtual power plant as the objective function. Specifically: The objective function is constructed with the minimum virtual power plant operation cost as the target. The expression of the objective function is: In the formula, is a target function, is a light and wind penalty cost, is a gas unit generation cost, is a wind turbine generation cost, is a photovoltaic generation cost, is an electric vehicle scheduling cost, is a storage charging and discharging cost, is a grid interaction cost; The abandoned light, abandoned wind penalty cost The calculation formula is: In the formula, and respectively the abandoned wind and light power at the moment, the abandoned wind and light penalty unit price, is the unit time; The electric vehicle scheduling cost is composed of charging benefits of the electric vehicles participating in demand response and discharging cost of the electric vehicles The calculation formula is: In the formula, for The cost of electric vehicle discharge compensation at all times for Benefits of electric vehicles participating in demand response at any time for Time of the first The discharge power of an electric vehicle for The price of compensation for electric vehicle discharge from a virtual power plant at any given time. for Number of electric vehicles discharging at any given time , They are respectively Real-time monitoring of charging load before and after demand response. , They are respectively Implement charging prices before and after demand response in real time; The constraint condition model is established, which includes wind, light turbine output constraint, gas turbine constraint, energy storage constraint, electric vehicle constraint, grid interaction constraint and balance constraint; The scheduling scheme development module is used to develop the optimal scheme of virtual power plant energy scheduling according to the data calculated by the model.

Images