Electric vehicle flexibility prediction method considering demand response signal

Abstract

The invention discloses an electric vehicle flexibility prediction method considering demand response signals. The method comprises the following steps: S1, generating an input feature matrix with a unified format; s2, fuzzy feature data are output; s3, designing an improved fuzzy decision tree model structure according to the fuzzification feature data; s4, training the improved fuzzy decision tree model by using historical operation data and demand response signals of the electric vehicle; s5, inputting the real-time operation data and the real-time demand response signal into the trained and optimized improved fuzzy decision tree model, and classifying the flexible response capability of the electric vehicle according to a quantitative predicted value; and S6, based on the quantitative prediction value of the flexible response capability of the electric vehicles, calculating a charging and discharging behavior scheduling scheme of the electric vehicle group. According to the invention, accurate modeling of the complex interaction relationship between the electric vehicle and the power grid is realized.

Description

Technical Field

[0001] The invention relates to the technical field of electric vehicles, and in particular to a method for predicting the flexibility of electric vehicles taking into account demand response signals. Background Art

[0002] With the popularization of electric vehicles and large-scale access to renewable energy, the coordinated optimization of electric vehicles and smart grids has become an important research direction of modern energy systems. Due to their controllable charging and discharging behavior, electric vehicles can not only meet their own travel needs, but also participate in grid scheduling as distributed energy storage units to achieve peak shaving and valley filling and renewable energy utilization.

[0003] At present, the mainstream methods of flexibility prediction are mostly based on static parameter analysis, which usually relies on the battery capacity of electric vehicles, users' charging habits and simple historical electricity consumption data to build static models for flexibility assessment. For example, traditional methods are based on statistical models or simple machine learning algorithms, using fixed rules to classify or quantify the flexibility of electric vehicles. This can achieve certain results when the amount of data is small, but its limitations are also very obvious.

[0004] First, existing methods cannot fully consider the dynamic interaction process between electric vehicles and power grids. Demand response signals are the core input of power grid dispatch, and traditional methods usually only use signals at the static or approximate processing stage, making it difficult to capture the real-time dynamic changes of demand response signals and their complex impact on the flexibility of electric vehicles. Secondly, existing technologies are relatively simple in data processing and lack deep integration of electric vehicle operation data and demand response signals, resulting in poor sensitivity of the model to multi-dimensional inputs and inability to accurately reflect the true flexibility of electric vehicles.

[0005] In summary, the existing technologies have significant deficiencies in the dynamic adaptability, data fusion depth and uncertainty processing of flexibility prediction, and are unable to meet the precise needs of flexibility assessment in scenarios where electric vehicles are connected on a large scale. The defects of the existing technologies directly affect the collaborative optimization efficiency of electric vehicles and smart grids. An innovative flexibility prediction method is urgently needed to solve the above problems. Summary of the invention

[0006] One object of the present invention is to propose a method for predicting the flexibility of electric vehicles taking into account demand response signals, and the present invention realizes accurate modeling of the complex interactive relationship between electric vehicles and power grids.

[0007] According to an embodiment of the present invention, a method for predicting electric vehicle flexibility considering a demand response signal comprises the following steps:

[0008] S1. Collect real-time operation data and demand response signals of electric vehicles, perform data cleaning and standardization, and generate an input feature matrix in a unified format;

[0009] S2. Fuzzify the key parameters in the input feature matrix based on fuzzy logic and output fuzzy feature data;

[0010] S3. Design and improve the fuzzy decision tree model structure based on the fuzzy feature data, and set the classification output categories of the flexible response capabilities of electric vehicles under different working conditions at the leaf nodes of the improved fuzzy decision tree model;

[0011] S4. Using the historical operation data of electric vehicles and demand response signals to train the improved fuzzy decision tree model;

[0012] S5. Input the real-time operation data and the real-time demand response signal into the trained and optimized improved fuzzy decision tree model, use the improved fuzzy decision tree model to infer and calculate the input features, output the quantitative prediction value of the flexible response capability of the electric vehicle, and classify the flexible response capability of the electric vehicle according to the quantitative prediction value;

[0013] S6. Based on the quantitative prediction value of the flexible response capability of electric vehicles, the charging and discharging behavior scheduling plan of the electric vehicle group is calculated. Combined with the current operating status of the power grid, a peak shaving and valley filling strategy, load balancing plan and renewable energy access strategy are formulated and applied to the optimization decision of electric vehicle power grid.

[0014] Optionally, the S1 includes the following steps:

[0015] S11. Get battery status data D b (t), including the remaining battery power and charging and discharging power data, and collecting vehicle location data D l (t), including the real-time geographic location coordinates of the vehicle and the collection of user electricity usage habit data D u (t), including the user's average daily power consumption, charging frequency and charging time preference, constitutes the real-time operation data set of electric vehicles:

[0016] D EV (t) = {D b (t),D l (t),D u (t)};

[0017] Among them, t represents the acquisition time;

[0018] S12. Obtaining real-time electricity price signal P e (t), reflects the current power price of the power grid and collects the dynamic signal of regional power load L r(t), represents the real-time load changes in the power grid area, and constructs the demand response data set:

[0019] D DR (t) = {P e (t),L r (t)};

[0020] S13. Use interpolation methods to process missing battery status data and real-time electricity price signals for electric vehicle real-time operation data sets and demand response data sets, and reconstruct abnormal vehicle location data;

[0021] S14. Perform dimensionless processing on the electric vehicle real-time operation data set and the demand response data set, and map the data to the [-1,1] interval through standardization processing;

[0022] S15. The standardized electric vehicle real-time operation dataset D′ EV (t) and demand response dataset D′ DR (t) Merge by time step t to generate the input feature matrix:

[0023] X(t) = {D′ EV (t),D′ DR (t)}.

[0024] Optionally, S2 includes the following steps:

[0025] S21. According to the forecast demand of electric vehicle flexibility, a fuzzy set {F i}, where F i is the parameter x i (t)∈X(t) corresponds to the fuzzy set, and the membership function of each fuzzy set is defined To quantify x i (t) belongs to the fuzzy set F i The degree of membership;

[0026] S22. Aiming at the demand of electric vehicle flexibility prediction, combined with the actual physical meaning of each parameter in the input feature matrix X(t), a fuzzy rule set R is set. Each fuzzy rule r in the fuzzy rule set k It is expressed as:

[0027] r k :Ifx 1 ∈F 1 andx 2 ∈F 2 and…Theny∈F y ;

[0028] Among them, r krepresents the kth fuzzy rule, x 1 ,x 2 ,…are the key parameters in the input feature matrix, y is the fuzzy output category of flexibility prediction, F 1 ,F 2 ,F y are the fuzzy sets corresponding to the parameters respectively;

[0029] S23. Based on the input feature matrix X(t) and the defined fuzzy rule set R, for each fuzzy rule r k Calculation rule applicability α k :

[0030]

[0031] Among them, α k Indicates the applicability of the fuzzy rule, which is used to evaluate the matching degree of the fuzzy rule under the current input features and obtain the fuzzy output value through fuzzy reasoning

[0032] S24. Output value based on fuzzification Combined with the needs of flexibility prediction, the flexibility prediction result y(t) is divided into multiple levels, and the fuzzy level boundaries {b i}, the prediction results y(t) are divided into different flexibility levels:

[0033] y(t)∈Level i ifb i-1 ≤y(t) i ;

[0034] Among them, Level i is the i-th flexibility level, b i is the boundary value for dividing the levels;

[0035] The flexibility level y(t) is used as the fuzzy feature data and the corresponding membership value Output together.

[0036] Optionally, S3 includes the following steps:

[0037] S31. Based on the fuzzy feature data and electric vehicle flexibility forecasting needs, designing a multi-layer branching structure of an improved fuzzy decision tree model:

[0038] The first layer of nodes is used to segment the fuzzy features of demand response signals in the input data. The real-time electricity price signal P e and regional power load dynamics L r Make priority branches; ​

[0039] The second layer nodes are fuzzy about the real-time operation characteristics of electric vehicles Further refinement of branches, including battery status data D b , vehicle location data D l , User electricity usage habit data D u ;

[0040] The leaf nodes correspond to the flexibility prediction output categories, which are classified according to the comprehensive characteristics of the demand response signal and the real-time status of the electric vehicle;

[0041] S32. In the process of splitting decision tree nodes, fuzzy entropy H is used F As a splitting criterion, used to quantify the uncertainty of the input features in predicting the output class, the splitting rule is defined as:

[0042]

[0043] Among them, H F (x i ,F j ) represents the feature x i In the fuzzy set F j The fuzzy entropy of:

[0044]

[0045] Select the splitting path with the smallest fuzzy entropy to optimize and improve the ability of the fuzzy decision tree model to distinguish fuzzy feature data;

[0046] S33. Construct hierarchical adaptive branching paths based on the real-time dynamics of demand response signals and the time sensitivity of electric vehicle flexibility prediction:

[0047] Introducing a time weight factor ω into the branch path t :

[0048]

[0049] Among them, t is the current time step, t 0 is the reference time point, λ is the adjustment coefficient;

[0050] The time weight factor is used to adjust the selection priority of the fuzzy logic branch path, so that the improved fuzzy decision tree model gives priority to input features with recent dynamic changes higher than the preset value;

[0051] S34. Classify the flexibility in the tree leaf nodes of the decision tree and quantify and output the flexibility.

[0052] Optionally, the S34 includes the following steps:

[0053] S341. In the leaf nodes of the decision tree, according to the branch path of the input fuzzy feature data and the applicability of the fuzzy rules, define the classification output category set Class corresponding to each leaf node:

[0054]

[0055] Among them, Class j is the jth flexibility prediction category, including high flexibility, medium flexibility or low flexibility, τ is the set flexibility classification threshold, and n is the total number of candidate classification categories;

[0056] S342. According to the branch path P corresponding to the leaf node k , the applicability of each fuzzy rule is calculated by combining the fuzzy features of all path nodes By introducing dynamic fluctuations and weights of features, the improved fuzzy decision tree model can capture the real-time changes in the interaction between electric vehicles and the power grid:

[0057]

[0058] in, is feature x i Belongs to the fuzzy set F i The membership degree, w i is feature x i The weight of reflects the relative importance of the feature to flexibility prediction. is feature x i With the characteristic mean The deviation is used to quantify the dynamic fluctuation of the feature, β is the adjustment factor for adjusting the deviation, and m is the path P k The number of fuzzy features involved above;

[0059] S343. Combine the fuzzy rule applicability and the flexibility classification standardization value to calculate the leaf node flexibility quantization output value y(L k ), the flexibility quantification output value weighs the flexibility output of the leaf node so that it reflects the real-time operation characteristics of electric vehicles and combines the dynamic impact of demand response signals:

[0060]

[0061] Among them, Class j is the standardized value of the flexibility category, is the fuzzy characteristic membership related to the demand response signal, reflecting the impact of demand response on flexibility prediction, γ is the demand response signal weight adjustment factor, is the applicability of fuzzy rules.

[0062] Optionally, S4 includes the following steps:

[0063] S41. Using the historical operation data set of electric vehicles and demand response signal historical dataset Construct the initial training set, set the initial splitting rules and fuzzy logic rule parameters and flexibility classification standardization value Class j ;

[0064] S42. At each node N in the decision tree i , based on the fuzzy logic membership and the initial training set, calculate the gain value ΔG of the node splitting:

[0065]

[0066] Among them, H(N i ) represents node N i The fuzzy entropy of |N i | and |N i,j | respectively for node N i and child node N i,j The amount of data, split gain value ΔG(N i ) is used to select the optimal splitting path;

[0067] S43. Dynamically adjust the branching rules of the improved fuzzy decision tree model according to the dynamic changes of the demand response signal, and introduce real-time update parameters θ t , the dynamic update formula of branch rules is:

[0068] Rule new =Rule old ·(1+θ t );

[0069] Among them, Rule old is the initial branching rule, θ t To update parameters in real time:

[0070]

[0071] Where ΔP e (t) and ΔL r (t) respectively represent the change of real-time electricity price signal and regional electricity load, max(P e ,L r ) is the maximum value of the historical signal, which is used for normalization adjustment;

[0072] S44. According to the applicability of the fuzzy rules and the flexibility classification distribution of the initial training set, the leaf node output parameters are optimized, and the leaf node output parameters are updated to:

[0073]

[0074] in, is the classification value of the current leaf node, is the target classification value, η 1 is the learning rate, which is used to control the optimization step size;

[0075] S45. Using real-time demand response signals and electric vehicle operation data, the fuzzy rule parameters of the improved fuzzy decision tree model are dynamically updated based on the fuzzy logic rule optimization mechanism:

[0076]

[0077] in, is the original membership function value, Δx i Represents the dynamic change of the eigenvalue, λ 1 To update the weight factor;

[0078] S46. Perform global verification on the improved fuzzy decision tree model after training.

[0079] Optionally, S5 includes the following steps:

[0080] S51. Inputting the real-time operation data and the real-time demand response signal into the trained and optimized improved fuzzy decision tree model;

[0081] S52. Mapping the real-time input data to the fuzzy feature space, calculating the membership of each input feature under the corresponding fuzzy set through the defined membership function, and obtaining the fuzzy feature data;

[0082] S53. Select the optimal reasoning path P in the improved fuzzy decision tree model based on the fuzzy membership of real-time features k :

[0083]

[0084] in, For path P j The comprehensive membership score of

[0085] S54. Based on the selected optimal reasoning path P k The quantitative prediction value y(t) of the flexible response capability of electric vehicles is calculated based on the quantification rules of leaf nodes. According to the quantitative prediction value y(t) of the flexible response capability of electric vehicles and combined with the flexibility classification standard, the prediction value is mapped to the corresponding classification level.

[0086] Optionally, the S6 includes the following steps:

[0087] S61. Calculate the electric vehicle charging and discharging behavior scheduling plan based on the flexibility prediction result y(t) of the electric vehicle group and the real-time demand response signal:

[0088]

[0089] Among them, P EV (t) is the total charging and discharging power of the electric vehicle group, Flex i (t) is the predicted flexibility value of the i-th electric vehicle, N 1 The number of electric vehicles connected to the grid;

[0090] S62. Scheduling scheme combining charging and discharging behavior of electric vehicle groups P EV (t) and the load forecast value L of the power grid grid (t) Develop peak-shaving and valley-filling strategies:

[0091] Target(t)=min|L grid (t)+P EV (t)-L avg |;

[0092] Among them, Target(t) is the optimization target, L avg is the daily average value of the grid load, and the scheduling scheme P is adjusted by adjusting the charging and discharging behavior of electric vehicles. EV (t) Balance the grid load and achieve the effect of peak load reduction and valley load filling;

[0093] S63. Optimizing the scheduling scheme of electric vehicle charging and discharging behavior based on the peak shaving and valley filling strategy P EV (t) and adjustable load P adj (t), to balance the load:

[0094] L total (t) = L grid (t)+P EV (t)+P adj (t);

[0095] Among them, L total (t) is the total load of the power grid, P adj (t) other adjustable loads;

[0096] Adjust other adjustable loads P adj (t) and the electric vehicle charging and discharging behavior scheduling scheme P EV (t) Make the total load of the power grid L total (t) approaches the optimization target Target(t);

[0097] S64. Combined with the real-time renewable energy power generation of the power grid P RE (t) and the flexibility prediction result y(t) of the electric vehicle group, and formulate the renewable energy access strategy:

[0098] P RE-in(t) = min[P RE (t), P EV (t)·η eff ];

[0099] Among them, P RE-in (t) is the renewable energy power connected to the grid, η eff Optimize P for charging efficiency of electric vehicles RE-in (t) To maximize the utilization of renewable energy.

[0100] The beneficial effects of the present invention are:

[0101] (1) The present invention deeply integrates fuzzy logic with decision trees, models the uncertainty of input data through fuzzy logic, and combines fuzzy entropy to optimize the node splitting rules of decision trees, thereby solving the problem that traditional decision trees are unable to handle continuity and uncertainty data. The dynamic fuzzy rule adjustment mechanism is introduced, which can adaptively update the decision tree branch path and leaf node output value according to the real-time dynamic changes of demand response signals, thereby significantly improving the accuracy of flexibility prediction.

[0102] (2) The present invention conducts hierarchical analysis on demand response signals, integrates their dynamic characteristics into the flexibility prediction model, and proposes a path priority selection algorithm based on time weight factors. In the improved decision tree model, the applicability of fuzzy logic rules is adjusted by calculating the dynamic fluctuation coefficient of the demand response signal to ensure that the model can give priority to the impact of real-time signals on the flexibility of electric vehicles, thereby achieving accurate modeling of the complex interactive relationship between electric vehicles and power grids.

[0103] (3) The present invention deeply couples the flexibility prediction results with the optimal scheduling of electric vehicle charging and discharging behavior, and proposes a peak shaving and valley filling strategy and load balancing solution based on the quantified output of flexibility. By dynamically optimizing the charging and discharging behavior and the renewable energy access strategy, the real-time balance of the grid load and the efficient utilization of renewable energy are achieved, thus solving the problem of the disconnection between flexibility prediction and grid scheduling in traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0104] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0105] Figure 1 The present invention provides a flow chart of an electric vehicle flexibility prediction method considering demand response signals. DETAILED DESCRIPTION

[0106] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0107] refer to Figure 1 , a method for predicting electric vehicle flexibility considering demand response signals, comprising the following steps:

[0108] S1. Collect real-time operation data and demand response signals of electric vehicles, perform data cleaning and standardization, and generate an input feature matrix in a unified format;

[0109] S2. Fuzzify the key parameters in the input feature matrix based on fuzzy logic and output fuzzy feature data;

[0110] S3. Design and improve the fuzzy decision tree model structure based on the fuzzy feature data, and set the classification output categories of the flexible response capabilities of electric vehicles under different working conditions at the leaf nodes of the improved fuzzy decision tree model;

[0111] S4. Using the historical operation data of electric vehicles and demand response signals to train the improved fuzzy decision tree model;

[0112] S5. Input the real-time operation data and the real-time demand response signal into the trained and optimized improved fuzzy decision tree model, use the improved fuzzy decision tree model to infer and calculate the input features, output the quantitative prediction value of the flexible response capability of the electric vehicle, and classify the flexible response capability of the electric vehicle according to the quantitative prediction value;

[0113] S6. Based on the quantitative prediction value of the flexible response capability of electric vehicles, the charging and discharging behavior scheduling plan of the electric vehicle group is calculated. Combined with the current operating status of the power grid, a peak shaving and valley filling strategy, load balancing plan and renewable energy access strategy are formulated and applied to the optimization decision of electric vehicle power grid.

[0114] In this implementation, S1 includes the following steps:

[0115] S11. Get battery status data D b (t), including the remaining battery power and charging and discharging power data, and collecting vehicle location data D l (t), including the real-time geographic location coordinates of the vehicle and the collection of user electricity usage habit data D u (t), including the user's average daily power consumption, charging frequency and charging time preference, constitutes the real-time operation data set of electric vehicles:

[0116] D EV (t) = {D b (t), D l (t), Du (t)};

[0117] Among them, t represents the acquisition time;

[0118] S12. Obtaining real-time electricity price signal P e (t), reflects the current power price of the power grid and collects the dynamic signal of regional power load L r (t), represents the real-time load changes in the power grid area, and constructs the demand response data set:

[0119] D DR (t) = {P e (t), L r (t)};

[0120] S13. Use interpolation methods to process missing battery status data and real-time electricity price signals for electric vehicle real-time operation data sets and demand response data sets, and reconstruct abnormal vehicle location data;

[0121] S14. Perform dimensionless processing on the electric vehicle real-time operation data set and the demand response data set, and map the data to the [-1, 1] interval through standardization processing;

[0122] S15. The standardized electric vehicle real-time operation dataset D′ EV (t) and demand response dataset D′ DR (t) Merge by time step t to generate the input feature matrix:

[0123] X(t) = {D′ EV (t), D′ DR (t)}.

[0124] In this implementation, S2 includes the following steps:

[0125] S21. According to the forecast demand of electric vehicle flexibility, a fuzzy set {F i}, where F i is the parameter x i (t)∈X(t) corresponds to the fuzzy set, and the membership function of each fuzzy set is defined To quantify x i (t) belongs to the fuzzy set F i The degree of membership;

[0126] S22. Aiming at the demand of electric vehicle flexibility prediction, combined with the actual physical meaning of each parameter in the input feature matrix X(t), a fuzzy rule set R is set. Each fuzzy rule r in the fuzzy rule set k It is expressed as:

[0127] rk :Ifx 1 ∈F 1 and x 2 ∈F 2 and...Theny∈F y ;

[0128] Among them, r k represents the kth fuzzy rule, x 1 , x 2 , …are the key parameters in the input feature matrix, y is the fuzzy output category of flexibility prediction, F 1 , F 2 , F y are the fuzzy sets corresponding to the parameters respectively;

[0129] S23. Based on the input feature matrix X(t) and the defined fuzzy rule set R, for each fuzzy rule r k Calculation rule applicability α k :

[0130]

[0131] Among them, α k Indicates the applicability of the fuzzy rule, which is used to evaluate the matching degree of the fuzzy rule under the current input features and obtain the fuzzy output value through fuzzy reasoning

[0132] S24. Output value based on fuzzification Combined with the needs of flexibility prediction, the flexibility prediction result y(t) is divided into multiple levels, and the fuzzy level boundaries {b i}, the prediction results y(t) are divided into different flexibility levels:

[0133] y(t)∈Level i ifb i-1 ≤y(t) i ;

[0134] Among them, Level i is the i-th flexibility level, b i is the boundary value for dividing the levels;

[0135] The flexibility level y(t) is used as the fuzzy feature data and the corresponding membership value Output together.

[0136] In this implementation, S3 includes the following steps:

[0137] S31. Based on the fuzzy feature data ​and electric vehicle flexibility forecasting needs, designing a multi-layer branching structure of an improved fuzzy decision tree model:

[0138] The first layer of nodes is used to segment the fuzzy features of demand response signals in the input data. The real-time electricity price signal P e and regional power load dynamics L r Make priority branches;

[0139] The second layer nodes are fuzzy about the real-time operation characteristics of electric vehicles Further refinement of branches, including battery status data D b , vehicle location data D l , User electricity usage habit data D u ;

[0140] The leaf nodes correspond to the flexibility prediction output categories, which are classified according to the comprehensive characteristics of the demand response signal and the real-time status of the electric vehicle;

[0141] S32. In the process of splitting decision tree nodes, fuzzy entropy H is used F As a splitting criterion, used to quantify the uncertainty of the input features in predicting the output class, the splitting rule is defined as:

[0142]

[0143] Among them, H F (x i ,F j ) represents the feature x i In the fuzzy set F j The fuzzy entropy of:

[0144]

[0145] Select the splitting path with the smallest fuzzy entropy to optimize and improve the ability of the fuzzy decision tree model to distinguish fuzzy feature data;

[0146] S33. Construct hierarchical adaptive branching paths based on the real-time dynamics of demand response signals and the time sensitivity of electric vehicle flexibility prediction:

[0147] Introducing a time weight factor ω into the branch path t :

[0148]

[0149] Among them, t is the current time step, t 0 is the reference time point, λ is the adjustment coefficient;

[0150] The time weight factor is used to adjust the selection priority of the fuzzy logic branch path, so that the improved fuzzy decision tree model gives priority to input features with recent dynamic changes higher than the preset value;

[0151] S34. Classify the flexibility in the tree leaf nodes of the decision tree and quantify and output the flexibility.

[0152] In this implementation, S34 includes the following steps:

[0153] S341. In the leaf nodes of the decision tree, according to the branch path of the input fuzzy feature data and the applicability of the fuzzy rules, define the classification output category set Class corresponding to each leaf node:

[0154]

[0155] Among them, Class j is the jth flexibility prediction category, including high flexibility, medium flexibility or low flexibility, τ is the set flexibility classification threshold, and n is the total number of candidate classification categories;

[0156] S342. According to the branch path P corresponding to the leaf node k , the applicability of each fuzzy rule is calculated by combining the fuzzy features of all path nodes By introducing dynamic fluctuations and weights of features, the improved fuzzy decision tree model can capture the real-time changes in the interaction between electric vehicles and the power grid:

[0157]

[0158] in, is feature x i Belongs to the fuzzy set F i The membership degree, w i is feature x i The weight of reflects the relative importance of the feature to flexibility prediction. is feature x i With the characteristic mean The deviation is used to quantify the dynamic fluctuation of the feature, β is the adjustment factor for adjusting the deviation, and m is the path P k The number of fuzzy features involved above;

[0159] S343. Combine the fuzzy rule applicability and the flexibility classification standardization value to calculate the leaf node flexibility quantization output value y(L k ), the flexibility quantification output value weighs the flexibility output of the leaf node so that it reflects the real-time operation characteristics of electric vehicles and combines the dynamic impact of demand response signals:

[0160]

[0161] Among them, Class j is the standardized value of the flexibility category, is the fuzzy characteristic membership related to the demand response signal, reflecting the impact of demand response on flexibility prediction, γ is the demand response signal weight adjustment factor, is the applicability of fuzzy rules.

[0162] In this implementation, S4 includes the following steps:

[0163] S41. Using the historical operation data set of electric vehicles and demand response signal historical dataset Construct the initial training set, set the initial splitting rules and fuzzy logic rule parameters and flexibility classification standardization value Class j ;

[0164] S42. At each node N in the decision tree i , based on the fuzzy logic membership and the initial training set, calculate the gain value ΔG of the node splitting:

[0165]

[0166] Among them, H(N i ) represents node N i The fuzzy entropy of |N i | and |N i,j | respectively for node N i and child node N i,j The amount of data, split gain value ΔG(N i ) is used to select the optimal splitting path;

[0167] S43. Dynamically adjust the branching rules of the improved fuzzy decision tree model according to the dynamic changes of the demand response signal, and introduce real-time update parameters θ t , the dynamic update formula of branch rules is:

[0168] Rule new =Rule old ·(1+θ t );

[0169] Among them, Rule old is the initial branching rule, θ t To update parameters in real time:

[0170]

[0171] Where ΔP e (t) and ΔL r(t) respectively represent the change of real-time electricity price signal and regional electricity load, max(P e ,L r ) is the maximum value of the historical signal, which is used for normalization adjustment;

[0172] S44. According to the applicability of the fuzzy rules and the flexibility classification distribution of the initial training set, the leaf node output parameters are optimized, and the leaf node output parameters are updated to:

[0173]

[0174] in, is the classification value of the current leaf node, is the target classification value, η 1 is the learning rate, which is used to control the optimization step size;

[0175] S45. Using real-time demand response signals and electric vehicle operation data, the fuzzy rule parameters of the improved fuzzy decision tree model are dynamically updated based on the fuzzy logic rule optimization mechanism:

[0176]

[0177] in, is the original membership function value, Δx i Represents the dynamic change of the eigenvalue, λ 1 To update the weight factor;

[0178] S46. Perform global verification on the improved fuzzy decision tree model after training.

[0179] In this implementation, S5 includes the following steps:

[0180] S51. Inputting the real-time operation data and the real-time demand response signal into the trained and optimized improved fuzzy decision tree model;

[0181] S52. Mapping the real-time input data to the fuzzy feature space, calculating the membership of each input feature under the corresponding fuzzy set through the defined membership function, and obtaining the fuzzy feature data;

[0182] S53. Select the optimal reasoning path P in the improved fuzzy decision tree model based on the fuzzy membership of real-time features k :

[0183]

[0184] in, For path P j The comprehensive membership score of

[0185] S54. Based on the selected optimal reasoning path Pk The quantitative prediction value y(t) of the flexible response capability of electric vehicles is calculated based on the quantification rules of leaf nodes. According to the quantitative prediction value y(t) of the flexible response capability of electric vehicles and combined with the flexibility classification standard, the prediction value is mapped to the corresponding classification level.

[0186] In this implementation, S6 includes the following steps:

[0187] S61. Calculate the electric vehicle charging and discharging behavior scheduling plan based on the flexibility prediction result y(t) of the electric vehicle group and the real-time demand response signal:

[0188]

[0189] Among them, P EV (t) is the total charging and discharging power of the electric vehicle group, Flex i (t) is the predicted flexibility value of the i-th electric vehicle, N 1 The number of electric vehicles connected to the grid;

[0190] S62. Scheduling scheme combining charging and discharging behavior of electric vehicle groups P EV (t) and the load forecast value L of the power grid grid (t) Develop peak-shaving and valley-filling strategies:

[0191] Target(t)=min|L grid (t)+P EV (t)-L avg |;

[0192] Among them, Target(t) is the optimization target, L avg is the daily average value of the grid load, and the scheduling scheme P is adjusted by adjusting the charging and discharging behavior of electric vehicles. EV (t) Balance the grid load and achieve the effect of peak load reduction and valley load filling;

[0193] S63. Optimizing the scheduling scheme of electric vehicle charging and discharging behavior based on the peak shaving and valley filling strategy P EV (t) and adjustable load P adj (t), to balance the load:

[0194] L total (t) = L grid (t)+P EV (t)+P adj (t);

[0195] Among them, L total (t) is the total load of the power grid, P adj (t) other adjustable loads;

[0196] Adjust other adjustable loads Padj (t) and the electric vehicle charging and discharging behavior scheduling scheme P EV (t) Make the total load of the power grid L total (t) approaches the optimization target Target(t);

[0197] S64. Combined with the real-time renewable energy power generation of the power grid P RE (t) and the flexibility prediction result y(t) of the electric vehicle group, and formulate the renewable energy access strategy:

[0198] P RE-in (t) = min[P RE (t), P EV (t)·η eff ];

[0199] Among them, P RE-in (t) is the renewable energy power connected to the grid, η eff Optimize P for charging efficiency of electric vehicles RE-in (t) To maximize the utilization of renewable energy.

[0200] Embodiment 1:

[0201] Embodiment At 14:00 on May 15, 2024, the smart grid dispatch center of a large city received real-time operation data and demand response signals from 5,000 electric vehicles connected to the grid. The system showed that the current load of the grid was close to 950MW, exceeding 90% of the warning line of the day. The real-time demand response signal showed that the electricity price at that time fluctuated from 0.8 yuan / kWh to 1.2 yuan / kWh, and the regional power load dynamics showed a continuous upward trend. The dispatch center immediately enabled the method of the present invention to perform flexibility prediction and optimized dispatching.

[0202] The system first collects the operating data of some vehicles. In the embodiment, the electric vehicle with vehicle ID "EV-0001" has a current battery power of 65%, and the vehicle location is displayed in peak load area A. The user preset charging time is 20:00. Another electric vehicle with vehicle ID "EV-0374" shows a battery power of 30%, and is located in area C with lower load. The user preset charging time is 18:00. The data is input into the improved fuzzy decision tree model of the present invention together with the real-time electricity price signal and load fluctuation data.

[0203] The improved fuzzy decision tree model calculates the flexibility of each electric vehicle through the membership function. In the embodiment, the flexibility prediction value of "EV-0001" is 0.85 (high flexibility), and the improved fuzzy decision tree model determines that it can be used for immediate discharge regulation, while the flexibility prediction value of "EV-0374" is 0.55 (medium flexibility), which is suitable for valley filling tasks at night. The system classifies the flexibility of all 5,000 electric vehicles through the model, of which 1,750 are judged to be high flexibility, 2,250 are medium flexibility, and the rest are low flexibility.

[0204] At 14:15, the dispatch center formulated a peak shaving strategy based on the forecast results. The system used all high-flexibility electric vehicles in the higher-load area A for discharge. The total discharge power of area A was set to 80MW. In the embodiment, "EV-0001" began to discharge at 11kW through its bidirectional charging interface, and the total discharge time was 30 minutes. In area B, some medium-flexibility electric vehicles were selected to participate in the peak shaving task, with a total discharge power of 40MW. At the same time, the system coordinated the low-flexibility vehicles in area C to postpone the charging time to 22:00.

[0205] The system completed the above peak-shaving task between 14:30 and 15:00, and monitored that the grid load dropped to 920MW. Taking "EV-0001" as an example, the vehicle discharged a total of 5.5kWh in 30 minutes, and the remaining battery power dropped to 60%. The user's preset charging plan was not affected, and the vehicle discharge profit was 6.6 yuan (calculated based on the real-time electricity price).

[0206] During the off-peak period from 16:00 to 17:00, the dispatch center coordinated the wind farm to access 80MW of renewable energy power based on the renewable energy access strategy, and arranged for the medium-flexibility vehicles in area C to start charging. In the embodiment, "EV-0374" started charging at 16:15, with a charging power of 7kW, and the charging completion time was 17:45. The total charging power was 10.5kWh, and the user paid 6.3 yuan, saving 1.5 yuan compared with the conventional electricity price.

[0207] In order to verify the effectiveness of the method of the present invention, the dispatch center conducted a comparative test between the traditional method and the method of the present invention. The following Table 1 is the test data:

[0208] Table 1 Comparison data between the method of the present invention and the traditional method

[0209]

[0210] It can be seen from the data that the traditional method is unable to capture the dynamic changes of the demand response signal in real time, resulting in a low accuracy of flexibility prediction and only contributing 100MW of peak shaving power. The present invention not only improves the accuracy of flexibility prediction by dynamically adjusting fuzzy rules and path selection, but also increases the contribution of peak shaving power to 150MW, effectively alleviating the peak load pressure on the power grid and optimizing the utilization rate of wind farms.

[0211] This example verifies the practical application effect of the method of the present invention. In a dynamic electricity consumption scenario, the present invention can quickly respond to demand and perform flexible prediction and optimized scheduling based on the real-time data of electric vehicles and demand response signals. It not only solves the problem of insufficient prediction accuracy of traditional methods, but also significantly improves the efficiency of peak shaving and valley filling and renewable energy utilization, providing strong technical support for the intelligent operation of urban power grids.

[0212] The present invention deeply integrates fuzzy logic and decision trees, models the uncertainty of input data through fuzzy logic, and combines fuzzy entropy to optimize the node splitting rules of the decision tree, thereby solving the problem that traditional decision trees cannot handle continuity and uncertainty data sufficiently. The dynamic fuzzy rule adjustment mechanism is introduced, which can adaptively update the decision tree branch path and leaf node output value according to the real-time dynamic changes of the demand response signal, thereby significantly improving the accuracy of flexibility prediction.

[0213] The present invention conducts hierarchical analysis on demand response signals, integrates their dynamic characteristics into the flexibility prediction model, and proposes a path priority selection algorithm based on time weight factors. In the improved decision tree model, the applicability of fuzzy logic rules is adjusted by calculating the dynamic fluctuation coefficient of the demand response signal to ensure that the model can give priority to the impact of real-time signals on the flexibility of electric vehicles, thereby achieving accurate modeling of the complex interactive relationship between electric vehicles and power grids.

[0214] The present invention deeply couples the flexibility prediction results with the optimal scheduling of electric vehicle charging and discharging behavior, and proposes a peak shaving and valley filling strategy and load balancing solution based on flexibility quantification output. By dynamically optimizing charging and discharging behavior and renewable energy access strategy, real-time balancing of grid load and efficient utilization of renewable energy are achieved, solving the problem of disconnection between flexibility prediction and grid scheduling in traditional methods.

[0215] 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 technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for predicting electric vehicle flexibility considering demand response signals, characterized in that: The steps include: S1. Collect real-time operation data and demand response signals of electric vehicles, perform data cleaning and standardization, and generate an input feature matrix in a unified format; S2. Fuzzify the key parameters in the input feature matrix based on fuzzy logic and output fuzzy feature data; S3. Design and improve the fuzzy decision tree model structure based on the fuzzified feature data, and set the classification output categories of the flexible response capabilities of electric vehicles under different working conditions at the leaf nodes of the improved fuzzy decision tree model; S4. Using the historical operation data of electric vehicles and demand response signals to train the improved fuzzy decision tree model; S5. Input the real-time operation data and the real-time demand response signal into the trained and optimized improved fuzzy decision tree model, use the improved fuzzy decision tree model to infer and calculate the input features, output the quantitative prediction value of the flexible response capability of the electric vehicle, and classify the flexible response capability of the electric vehicle according to the quantitative prediction value; S6. Based on the quantitative prediction value of the flexible response capability of electric vehicles, the charging and discharging behavior scheduling plan of the electric vehicle group is calculated. Combined with the current operating status of the power grid, a peak shaving and valley filling strategy, load balancing plan and renewable energy access strategy are formulated and applied to the optimization decision of electric vehicle power grid.

2. The electric vehicle flexibility prediction method considering demand response signals according to claim 1, characterized in that: The S1 comprises the following steps: S11. Get battery status data D b (t), including the remaining battery power and charging and discharging power data, and collecting vehicle location data D l (t), including the real-time geographic location coordinates of the vehicle and the collection of user electricity usage habit data D u (t), including the user's average daily power consumption, charging frequency and charging time preference, constitutes the real-time operation data set of electric vehicles: D EV (t)={D b (t),D l (t),D u (t)}; Among them, t represents the acquisition time; S12. Obtain real-time electricity price signal P e (t), reflects the current power price of the power grid and collects the dynamic signal of regional power load L r (t), represents the real-time load changes in the power grid area, and constructs the demand response data set: D DR (t)={P e (t),L r (t)}; S13. Use interpolation methods to process missing battery status data and real-time electricity price signals for electric vehicle real-time operation data sets and demand response data sets, and reconstruct abnormal vehicle location data; S14. Perform dimensionless processing on the electric vehicle real-time operation data set and the demand response data set, and map the data to the [-1,1] interval through standardization processing; S15. The standardized electric vehicle real-time operation dataset D′ EV (t) and demand response dataset D′ DR (t) Merge by time step t to generate the input feature matrix: X(t)={D′ EV (t),D′ DR (t)}。 3. The electric vehicle flexibility prediction method considering demand response signals according to claim 1, characterized in that: The S2 comprises the following steps: S21. According to the forecast demand of electric vehicle flexibility, a fuzzy set {F i }, where F i is the parameter x i (t)∈X(t) corresponds to the fuzzy set, and the membership function of each fuzzy set is defined To quantify x i (t) belongs to the fuzzy set F i The degree of membership; S22. Aiming at the demand of electric vehicle flexibility prediction, combined with the actual physical meaning of each parameter in the input feature matrix X(t), a fuzzy rule set R is set. Each fuzzy rule r in the fuzzy rule set k It is expressed as: r k :Ifx1∈F1andx2∈F2and…Theny∈F y ; Among them, r k represents the kth fuzzy rule, x1, x2, ... are the key parameters in the input feature matrix, y is the fuzzy output category of flexibility prediction, F1, F2, F y are the fuzzy sets corresponding to the parameters respectively; S23. Based on the input feature matrix X(t) and the defined fuzzy rule set R, for each fuzzy rule r k Calculation rule applicability α k : a k =min{μ F1 (x1(t)),μ F2 (x2(t)),…}; Among them, α k Indicates the applicability of the fuzzy rule, which is used to evaluate the matching degree of the fuzzy rule under the current input features and obtain the fuzzy output value through fuzzy reasoning S24. Output value based on fuzzification Combined with the needs of flexibility prediction, the flexibility prediction result y(t) is divided into multiple levels, and the fuzzy level boundaries {b i }, the prediction results y(t) are divided into different flexibility levels: y(t)∈Level i ifb i-1 ≤y(t)<b i ; Among them, Level i is the i-th flexibility level, b i is the boundary value for dividing the levels; The flexibility level y(t) is used as the fuzzy feature data and the corresponding membership value Output together.

4. The electric vehicle flexibility prediction method considering demand response signals according to claim 1, characterized in that: The S3 comprises the following steps: S31. Based on the fuzzy feature data and electric vehicle flexibility forecasting needs, designing a multi-layer branching structure of an improved fuzzy decision tree model: The first layer of nodes is used to segment the fuzzy features of demand response signals in the input data. The real-time electricity price signal P e and regional power load dynamics L r Make priority branches; The second layer nodes are fuzzy about the real-time operation characteristics of electric vehicles Further refinement of branches, including battery status data D b , vehicle location data D l , User electricity usage habit data D u ; The leaf nodes correspond to the flexibility prediction output categories, which are classified according to the comprehensive characteristics of the demand response signal and the real-time status of the electric vehicle; S32. In the process of splitting decision tree nodes, fuzzy entropy H is used F As a splitting criterion, used to quantify the uncertainty of the input features in predicting the output class, the splitting rule is defined as: Among them, H F (x i ,F j ) represents the feature x i In the fuzzy set F j The fuzzy entropy of: Select the splitting path with the smallest fuzzy entropy to optimize and improve the ability of the fuzzy decision tree model to distinguish fuzzy feature data; S33. Construct hierarchical adaptive branching paths based on the real-time dynamics of demand response signals and the time sensitivity of electric vehicle flexibility prediction: Introducing a time weight factor ω into the branch path t : Among them, t is the current time step, t0 is the reference time point, and λ is the adjustment coefficient; The time weight factor is used to adjust the selection priority of the fuzzy logic branch path, so that the improved fuzzy decision tree model gives priority to input features with recent dynamic changes higher than the preset value; S34. Classify the flexibility in the tree leaf nodes of the decision tree and quantify and output the flexibility.

5. The electric vehicle flexibility prediction method considering demand response signals according to claim 1, characterized in that: The S34 comprises the following steps: S341. In the leaf nodes of the decision tree, according to the branch path of the input fuzzy feature data and the applicability of the fuzzy rules, define the classification output category set Class corresponding to each leaf node: Among them, Class j is the jth flexibility prediction category, including high flexibility, medium flexibility or low flexibility, τ is the set flexibility classification threshold, and n is the total number of candidate classification categories; S342. According to the branch path P corresponding to the leaf node k , the applicability of each fuzzy rule is calculated by combining the fuzzy features of all path nodes By introducing dynamic fluctuations and weights of features, the improved fuzzy decision tree model can capture the real-time changes in the interaction between electric vehicles and the power grid: in, is feature x i Belongs to the fuzzy set F i The membership degree, w i is feature x i The weight of reflects the relative importance of the feature to flexibility prediction. is feature x i With the characteristic mean The deviation is used to quantify the dynamic fluctuation of the feature, β is the adjustment factor for adjusting the deviation, and m is the path P k The number of fuzzy features involved above; S343. Combine the fuzzy rule applicability and the flexibility classification standardization value to calculate the leaf node flexibility quantization output value y(L k ), the flexibility quantification output value weighs the flexibility output of the leaf node so that it reflects the real-time operation characteristics of electric vehicles and combines the dynamic impact of demand response signals: Among them, Class j is the standardized value of the flexibility category, is the fuzzy characteristic membership related to the demand response signal, reflecting the impact of demand response on flexibility prediction, γ is the demand response signal weight adjustment factor, is the applicability of fuzzy rules.

6. The electric vehicle flexibility prediction method considering demand response signals according to claim 1, characterized in that: The S4 comprises the following steps: S41. Using the historical operation data set of electric vehicles and demand response signal historical dataset Construct the initial training set, set the initial splitting rules and fuzzy logic rule parameters and flexibility classification standardization value Class j ; S42. At each node N in the decision tree i , based on the fuzzy logic membership and the initial training set, calculate the gain value ΔG of the node splitting: Among them, H(N i ) represents node N i The fuzzy entropy of |N i | and |N i,j | respectively for node N i and child node N i,j The amount of data, split gain value ΔG(N i ) is used to select the optimal splitting path; S43. Dynamically adjust the branching rules of the improved fuzzy decision tree model according to the dynamic changes of the demand response signal, and introduce real-time update parameters θ t , the dynamic update formula of branch rules is: Rule new =Rule old ·(1+θ t ); Among them, Rule old is the initial branching rule, θ t To update parameters in real time: Among them, ΔP e (t) and ΔL r (t) respectively represent the change of real-time electricity price signal and regional electricity load, max(P e ,L r ) is the maximum value of the historical signal, which is used for normalization adjustment; S44. According to the applicability of the fuzzy rules and the flexibility classification distribution of the initial training set, the leaf node output parameters are optimized, and the leaf node output parameters are updated to: in, is the classification value of the current leaf node, is the target classification value, η1 is the learning rate, which is used to control the optimization step size; S45. Using real-time demand response signals and electric vehicle operation data, the fuzzy rule parameters of the improved fuzzy decision tree model are dynamically updated based on the fuzzy logic rule optimization mechanism: in, is the original membership function value, Δx i Represents the dynamic change of the eigenvalue, λ1 is the update weight factor; S46. Perform global verification on the improved fuzzy decision tree model after training.

7. The electric vehicle flexibility prediction method considering demand response signals according to claim 1, characterized in that: The S5 comprises the following steps: S51. Inputting the real-time operation data and the real-time demand response signal into the trained and optimized improved fuzzy decision tree model; S52. Mapping the real-time input data to the fuzzy feature space, calculating the membership of each input feature under the corresponding fuzzy set through the defined membership function, and obtaining the fuzzy feature data; S53. Select the optimal reasoning path P in the improved fuzzy decision tree model based on the fuzzy membership of real-time features k : in, For path P j The comprehensive membership score of S54. Based on the selected optimal reasoning path P k The quantitative prediction value y(t) of the flexible response capability of electric vehicles is calculated based on the quantification rules of leaf nodes. According to the quantitative prediction value y(t) of the flexible response capability of electric vehicles and combined with the flexibility classification standard, the prediction value is mapped to the corresponding classification level.

8. The electric vehicle flexibility prediction method considering demand response signals according to claim 1, characterized in that: The S6 comprises the following steps: S61. Calculate the electric vehicle charging and discharging behavior scheduling plan based on the flexibility prediction result y(t) of the electric vehicle group and the real-time demand response signal: Among them, P EV (t) is the total charging and discharging power of the electric vehicle group, Flex i (t) is the predicted flexibility value of the i-th electric vehicle, N1 is the number of electric vehicles connected to the grid; S62. Scheduling scheme combining charging and discharging behavior of electric vehicle groups P EV (t) and the load forecast value L of the power grid grid (t) Develop peak-shaving and valley-filling strategies: Target(t)=min|L grid (t)+P EV (t)-L avg |; Among them, Target(t) is the optimization target, L avg is the daily average value of the grid load, and the scheduling scheme P is adjusted by adjusting the charging and discharging behavior of electric vehicles. EV (t) Balance the grid load and achieve the effect of peak load reduction and valley load filling; S63. Optimizing the scheduling scheme of electric vehicle charging and discharging behavior based on the peak shaving and valley filling strategy P EV (t) and adjustable load P adj (t), so that the load is balanced: L total (t)=L grid (t)+P EV (t)+P adj (t); Among them, L total (t) is the total load of the power grid, P adj (t) other adjustable loads; Adjust other adjustable loads P adj (t) and the electric vehicle charging and discharging behavior scheduling scheme P EV (t) Make the total load of the power grid L total (t) approaches the optimization target Target(t); S64. Combined with the real-time renewable energy power generation of the power grid P RE (t) and the flexibility prediction result y(t) of the electric vehicle group, and formulate the renewable energy access strategy: P RE-in (t)=min[P RE (t),P EV (t)·η eff ]; Among them, P RE-in (t) is the renewable energy power connected to the grid, η eff Optimize P for charging efficiency of electric vehicles RE-in (t) To maximize the utilization of renewable energy.

Images