Short-term electric vehicle charging load prediction method and system under demand response

Abstract

The invention discloses a short-term electric vehicle charging load prediction method under demand response. The method comprises the following steps: acquiring travel historical data information of an electric vehicle in a target area, and processing to construct a training data set; constructing a short-term electric vehicle charging load prediction primary model based on an LSTM model and an attention mechanism, and training to obtain a short-term electric vehicle charging load prediction model; and adopting the short-term electric vehicle charging load prediction model to predict the actual short-term electric vehicle charging load of the target area. The invention also discloses a system for realizing the short-term electric vehicle charging load prediction method under the demand response. According to the method, the training data set is obtained by acquiring and processing the multi-dimensional travel historical data of the electric vehicle in the target area, and the prediction model is constructed and trained based on the LSTM model and the attention mechanism, so that the method not only can predict the charging load of the electric vehicle in a short term, but also is higher in reliability and better in accuracy.

Description

Technical Field

[0001] The present invention belongs to the field of electrical automation, and in particular relates to a method and system for predicting short-term electric vehicle charging load under demand response. Background Art

[0002] With the development of economy and technology and the improvement of people's living standards, electricity has become an indispensable secondary energy source in people's production and life, bringing endless convenience to people's production and life. Therefore, ensuring the stable and reliable supply of electricity has become one of the most important tasks of the power system.

[0003] At this stage, environmental problems are becoming more and more serious, so more and more people are beginning to choose electric vehicles as their first choice for travel. The rapid development of electric vehicles has also made the impact of electric vehicle charging load on the power system increasingly greater: the randomness and intermittency of electric vehicle charging load have brought great challenges to the safe and stable operation of the power system.

[0004] Short-term load forecasting is an important prerequisite for the power system to formulate operation plans and adjust operation modes. With the rapid growth of electric vehicle charging load, the prediction of short-term electric vehicle charging load is particularly important. At present, most traditional short-term electric vehicle load forecasting schemes only consider the impact of a single or a small number of influencing factors on the short-term electric vehicle charging load forecast. Therefore, this type of forecasting scheme has the problems of poor reliability and poor accuracy. Summary of the invention

[0005] One of the purposes of the present invention is to provide a short-term electric vehicle charging load forecasting method under demand response with high reliability and good accuracy.

[0006] A second objective of the present invention is to provide a system for realizing the short-term electric vehicle charging load prediction method under the demand response.

[0007] The short-term electric vehicle charging load forecasting method under the demand response provided by the present invention comprises the following steps:

[0008] S1. Obtaining historical data information of electric vehicle travel in the target area;

[0009] S2. Processing the data information obtained in step S1 to construct a training data set;

[0010] S3. Based on the LSTM model and attention mechanism, a primary model for short-term electric vehicle charging load prediction is constructed;

[0011] S4. Using the training data set obtained in step S2, the short-term electric vehicle charging load prediction primary model constructed in step S3 is trained to obtain a short-term electric vehicle charging load prediction model;

[0012] S5. Using the short-term electric vehicle charging load prediction model obtained in step S4, the actual short-term electric vehicle charging load of the target area is predicted.

[0013] The step S1 of obtaining the electric vehicle travel history data information of the target area specifically includes the following steps:

[0014] Obtaining historical data information on electric vehicle travel in the target area;

[0015] The data information includes electric vehicle travel data information, electric vehicle SOC data SOC EV , Charging pile load data PW cd 、Charging pile time-sharing electricity price data P cd , target area meteorological data information and electric vehicle charging load data information;

[0016] Among them, the electric vehicle travel data information includes the area data D of the electric vehicle before the trip before and the area data of the electric vehicle after the trip D after ; The target area meteorological data information includes temperature data T EV and precipitation data R EV .

[0017] Step S2 of processing the data information obtained in step S1 to construct a training data set includes the following steps:

[0018] Divide the target areas into types according to their functions;

[0019] Build an electric vehicle user travel model based on electric vehicle travel data;

[0020] Based on the time-sharing electricity price data of charging piles and the SOC data of electric vehicles, a user charging time-space diagram is constructed;

[0021] The training data set is constructed with the obtained electric vehicle user travel model, user charging time-space diagram and corresponding electric vehicle charging load data.

[0022] The method of classifying the target areas into types according to the functions of the areas specifically includes the following steps:

[0023] The target area is represented in the form of a graph: the nodes are the center points of each block in the target area, and the streets in the target area are the lines between the nodes;

[0024] In the figure, various areas are divided into residential areas, commercial areas and industrial areas according to their functions.

[0025] The construction of the electric vehicle user travel model based on the electric vehicle travel data specifically includes the following steps:

[0026] According to the travel destination and regional division, the travel starting point and end point of the electric vehicle user are expressed as

[0027]

[0028] Where D x is the travel purpose matrix of electric vehicle user x; D x mn represents the frequency of the route chosen by electric vehicle user x from travel area m to destination area n within one month, where m=j means the travel area is a residential area, m=s means the travel area is a commercial area, m=g means the travel area is an industrial area, n=j means the destination area is a residential area, n=s means the destination area is a commercial area, and n=g means the destination area is an industrial area;

[0029] According to the actual travel data of electric vehicle users, the travel probability matrix P of electric vehicle user x is constructed according to the travel period, travel probability and travel purpose. x :

[0030] P x =[P j P s P g ]

[0031]

[0032] Where P j is the travel probability matrix of electric vehicle users x with residential areas as their destinations; P s is the travel probability matrix of electric vehicle users x with the commercial area as the destination; P g is the travel probability matrix of electric vehicle users x with the industrial zone as the destination; P lnm represents the probability that electric vehicle user x travels from travel area m to destination area n during travel period l; the value of l ranges from 1 to 24;

[0033] According to the constructed travel purpose matrix and travel probability matrix, the travel time-space diagram of electric vehicle users is established, which is expressed as ξ cx ={t,D,P t}; where ξ cx is the space-time diagram of electric vehicle users’ travel, t is the travel time, D is the matrix composed of the user’s travel purpose matrix, P t It is the matrix composed of the travel probability matrix of the user at travel time t.

[0034] The construction of a user charging time-space diagram based on the charging pile time-sharing electricity price data and the electric vehicle SOC data specifically includes the following steps:

[0035] Based on the charging pile time-sharing electricity price data P cd and electric vehicle user charging area D cd , the charging scenarios of electric vehicle users are further divided and expressed as

[0036] P cd ∈{P cd_low ,P cd_flat ,P cd_high}

[0037] D cd ∈{D cd-j ,D cd-s ,D cd-g}

[0038] Ch={P cd ,D}

[0039] Where P cd is the time-sharing electricity price data of the charging pile, P cd Including off-peak electricity price P cd_low 、Usual electricity price P cd_flat and peak electricity price P cd_high ; Electric vehicle user charging area D cd Including residential areas D cd-j 、Business District D cd-s and Industrial Zone D cd-g ; Electric vehicle user charging scenario Ch is composed of P cd and D cd Decide;

[0040] The following formula is used to represent the user charging time-space diagram ξ cd :

[0041] ξ cd ={t,D cd ,P cd_t ,SOC EV_t}

[0042] Where P cd_t is the electricity price of the user at charging time t; SOC EV_t is the SOC value of the user's electric vehicle at charging time t.

[0043] The construction of a short-term electric vehicle charging load prediction primary model based on the LSTM model and the attention mechanism described in step S3 specifically includes the following steps:

[0044] The charging pile load data PW cd 、Electric vehicle user travel time and space diagramξcx 、User charging time-space diagramξ cd and target area meteorological data as the model input, and electric vehicle charging load data as the model output;

[0045] Set ω as the length of the time series window, and get the hidden state vector h corresponding to the input model at each moment t-ω ~h t ;

[0046] Set the hidden matrix H to H = (h t-ω ,h t-ω+1 ,…,h t-1 );

[0047] One-dimensional convolution is used to extract the temporal pattern matrix H θ , expressed as

[0048]

[0049] Where θ represents the TCN filter; is the jth TCN filter C j The eigenvalue extracted in the i-th row; len is the padding length in the convolution calculation; t' is the maximum length of the weight; T j,(t'-ω+len) is the jth TCN filter C j The convolution kernel.

[0050] The following formula is used as the evaluation function

[0051]

[0052] In the formula H θ is the vector of the i-th row of ; Α is the weight matrix;

[0053] Calculate the parameter weight δ i for

[0054] Attention A t The calculation formula is n is the characteristic quantity of input data;

[0055] Finally, the short-term electric vehicle charging load p predicted by the model is obtained t for

[0056] p t =Α'(Α h h t +Α v A t )

[0057] Wherein h ht The weight matrix of v A t ' is the weight matrix; Α' is the total weight matrix.

[0058] The present invention also provides a system for implementing the short-term electric vehicle charging load prediction method under the demand response, comprising a data acquisition module, a data processing module, a model construction module, a model training module and a load prediction module; the data acquisition module, the data processing module, the model construction module, the model training module and the load prediction module are connected in series in sequence; the data acquisition module is used to acquire the electric vehicle travel history data information of the target area, and upload the data information to the data processing module; the data processing module is used to process the acquired data information according to the received data information to construct a training data set, and upload the data information to the model construction module; the model construction module is used to construct a short-term electric vehicle charging load prediction primary model based on the LSTM model and the attention mechanism according to the received data information, and upload the data information to the model training module; the model training module is used to train the constructed short-term electric vehicle charging load prediction primary model according to the received data information, using the obtained training data set, to obtain the short-term electric vehicle charging load prediction model, and upload the data information to the load prediction module; the load prediction module is used to predict the actual short-term electric vehicle charging load of the target area according to the received data information and using the obtained short-term electric vehicle charging load prediction model.

[0059] The short-term electric vehicle charging load prediction method and system under demand response provided by the present invention obtains a training data set by acquiring and processing multi-dimensional electric vehicle travel history data in the target area, and constructs and trains a prediction model based on an LSTM model and an attention mechanism. Therefore, the present invention can not only realize the prediction of short-term electric vehicle charging load, but also has higher reliability and better accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 The figure is a schematic diagram of the method flow of the present invention.

[0061] Figure 2 Schematic diagram of the functional modules of the system of the present invention. DETAILED DESCRIPTION

[0062] like Figure 1 The method flow chart of the method of the present invention is shown as follows: The short-term electric vehicle charging load prediction method under the demand response disclosed by the present invention comprises the following steps:

[0063] S1. Obtaining the historical data information of electric vehicle travel in the target area; specifically including the following steps:

[0064] Obtaining historical data information on electric vehicle travel in the target area;

[0065] The data information includes electric vehicle travel data information, electric vehicle SOC data SOC EV , Charging pile load data PW cd 、Charging pile time-sharing electricity price data P cd , target area meteorological data information and electric vehicle charging load data information;

[0066] Among them, the electric vehicle travel data information includes the area data D of the electric vehicle before the trip before and the area data of the electric vehicle after the trip D after ; The target area meteorological data information includes temperature data T EV and precipitation data R EV ;

[0067] S2. Processing the data information obtained in step S1 to construct a training data set; comprising the following steps:

[0068] According to the functions of the regions, the target regions are divided into types; the specific steps include:

[0069] The target area is represented in the form of a graph: the nodes are the center points of each block in the target area, and the streets in the target area are the lines between the nodes;

[0070] In the figure, the various areas are divided into residential, commercial and industrial areas according to their functions;

[0071] Based on the travel data of electric vehicles, a travel model of electric vehicle users is constructed; the specific steps include:

[0072] According to the travel destination and regional division, the travel starting point and end point of the electric vehicle user are expressed as

[0073]

[0074] Where D x is the travel purpose matrix of electric vehicle user x; D x mn represents the frequency of the route chosen by electric vehicle user x from travel area m to destination area n within one month, where m=j means the travel area is a residential area, m=s means the travel area is a commercial area, m=g means the travel area is an industrial area, n=j means the destination area is a residential area, n=s means the destination area is a commercial area, and n=g means the destination area is an industrial area;

[0075] According to the actual travel data of electric vehicle users, the travel probability matrix P of electric vehicle user x is constructed according to the travel period, travel probability and travel purpose. x :

[0076] P x =[P j P s P g ]

[0077]

[0078]

[0079] Where P j is the travel probability matrix of electric vehicle users x with residential areas as their destinations; P s is the travel probability matrix of electric vehicle users x with the commercial area as the destination; P g is the travel probability matrix of electric vehicle users x with the industrial zone as the destination; P lnm represents the probability that electric vehicle user x travels from travel area m to destination area n during travel period l; the value of l ranges from 1 to 24;

[0080] According to the constructed travel purpose matrix and travel probability matrix, the travel time-space diagram of electric vehicle users is established, which is expressed as ξ cx ={t,D,P t}; where ξ cx is the space-time diagram of electric vehicle users’ travel, t is the travel time, D is the matrix composed of the user’s travel purpose matrix, P t is the matrix composed of the user's travel probability matrix at travel time t;

[0081] According to the time-sharing electricity price data of charging piles and the SOC data of electric vehicles, a user charging time-space diagram is constructed; specifically, the steps include:

[0082] Based on the charging pile time-sharing electricity price data P cd and electric vehicle user charging area D cd , the charging scenarios of electric vehicle users are further divided and expressed as

[0083] P cd ∈{P cd_low ,P cd_flat ,P cd_high}

[0084] D cd ∈{D cd-j ,D cd-s ,D cd-g}

[0085] Ch={Pcd ,D}

[0086] Where P cd is the time-sharing electricity price data of the charging pile, P cd Including off-peak electricity price P cd_low 、Usual electricity price P cd_flat and peak electricity price P cd_high ; Electric vehicle user charging area D cd Including residential areas D cd-j 、Business District D cd-s and Industrial Zone D cd-g ; Electric vehicle user charging scenario Ch is composed of P cd and D cd Decide;

[0087] The following formula is used to represent the user charging time-space diagram ξ cd :

[0088] ξ cd ={t,D cd ,P cd_t ,SOC EV_t}

[0089] Where P cd_t is the electricity price of the user at charging time t; SOC EV_t is the SOC value of the user's electric vehicle at charging time t;

[0090] The obtained electric vehicle user travel model, user charging time-space diagram and corresponding electric vehicle charging load data are used to construct a training data set;

[0091] S3. Based on the LSTM model and attention mechanism, a primary model for short-term electric vehicle charging load prediction is constructed; specifically, the following steps are included:

[0092] The charging pile load data PW cd 、Electric vehicle user travel time and space diagramξ cx 、User charging time-space diagramξ cd and target area meteorological data as the model input, and electric vehicle charging load data as the model output;

[0093] Set ω as the length of the time series window, and get the hidden state vector h corresponding to the input model at each moment t-ω ~h t ;

[0094] Set the hidden matrix H to H = (h t-ω ,h t-ω+1 ,…,h t-1 ), whose row vector represents the state of a single variable at different time steps, and the column vector represents the gate neuron parameters of LSTM at the same time step;

[0095] One-dimensional convolution is used to extract the temporal pattern matrix H θ , expressed as

[0096]

[0097] Where θ represents the TCN filter, and the hidden state of the LSTM model is extracted through multiple convolution operations of TCN on the time series; is the jth TCN filter C j The eigenvalue extracted in the i-th row; len is the padding length in the convolution calculation; t' is the maximum length of the weight, generally t' = ω; T j,(t'-ω+len) is the jth TCN filter C j The convolution kernel.

[0098] The following formula is used as the evaluation function

[0099]

[0100] In the formula H θ is the vector of the i-th row of ; Α is the weight matrix;

[0101] Calculate the parameter weight δ i for

[0102] Attention A t The calculation formula is n is the characteristic quantity of input data;

[0103] Finally, the short-term electric vehicle charging load p predicted by the model is obtained t for

[0104] p t =Α'(Α h h t +Α v A t )

[0105] Wherein h h t The weight matrix of v A t The weight matrix; Α' is the total weight matrix;

[0106] S4. Using the training data set obtained in step S2, the short-term electric vehicle charging load prediction primary model constructed in step S3 is trained to obtain a short-term electric vehicle charging load prediction model;

[0107] S5. Using the short-term electric vehicle charging load prediction model obtained in step S4, the actual short-term electric vehicle charging load of the target area is predicted.

[0108] The scheme of the present invention takes time-of-use electricity prices as economic influencing factors, electric vehicle users' travel needs and charging habits as user behavior influencing factors, and different regional functions as spatial influencing factors, extracts their spatiotemporal correlation, takes into account users' personal behavior and economic factors, and has a good performance in improving the accuracy of short-term electric vehicle charging load prediction.

[0109] like Figure 2 The functional module schematic diagram of the system of the present invention is shown as follows: the system of the present invention for realizing the short-term electric vehicle charging load prediction method under the demand response disclosed by the present invention comprises a data acquisition module, a data processing module, a model construction module, a model training module and a load prediction module; the data acquisition module, the data processing module, the model construction module, the model training module and the load prediction module are connected in series in sequence; the data acquisition module is used to obtain the electric vehicle travel history data information of the target area, and upload the data information to the data processing module; the data processing module is used to process the acquired data information according to the received data information to construct a training data set, and upload the data information to the model construction module; the model construction module is used to construct a short-term electric vehicle charging load prediction primary model based on the LSTM model and the attention mechanism according to the received data information, and upload the data information to the model training module; the model training module is used to train the constructed short-term electric vehicle charging load prediction primary model according to the received data information, using the obtained training data set, to obtain the short-term electric vehicle charging load prediction model, and upload the data information to the load prediction module; the load prediction module is used to predict the actual short-term electric vehicle charging load of the target area according to the received data information and using the obtained short-term electric vehicle charging load prediction model.

Claims

1. A short-term electric vehicle charging load forecasting method under demand response, comprising the following steps: S1. Obtaining historical data information of electric vehicle travel in the target area; S2. Processing the data information obtained in step S1 to construct a training data set; S3. Based on the LSTM model and attention mechanism, a primary model for short-term electric vehicle charging load prediction is constructed; S4. Using the training data set obtained in step S2, the short-term electric vehicle charging load prediction primary model constructed in step S3 is trained to obtain a short-term electric vehicle charging load prediction model; S5. Using the short-term electric vehicle charging load prediction model obtained in step S4, the actual short-term electric vehicle charging load of the target area is predicted.

2. The short-term electric vehicle charging load forecasting method under demand response according to claim 1 is characterized in that The step S1 of obtaining the electric vehicle travel history data information of the target area specifically includes the following steps: Obtaining historical data information on electric vehicle travel in the target area; The data information includes electric vehicle travel data information, electric vehicle SOC data SOC EV , Charging pile load data PW cd 、Charging pile time-sharing electricity price data P cd , target area meteorological data information and electric vehicle charging load data information; Among them, the electric vehicle travel data information includes the area data D of the electric vehicle before the trip before and the area data of the electric vehicle after the trip D after ; The target area meteorological data information includes temperature data T EV and precipitation data R EV .

3. The short-term electric vehicle charging load forecasting method under demand response according to claim 2 is characterized in that Step S2 of processing the data information obtained in step S1 to construct a training data set includes the following steps: Divide the target areas into types according to their functions; Build an electric vehicle user travel model based on electric vehicle travel data; Based on the time-sharing electricity price data of charging piles and the SOC data of electric vehicles, a user charging time-space diagram is constructed; The training data set is constructed with the obtained electric vehicle user travel model, user charging time-space diagram and corresponding electric vehicle charging load data.

4. The short-term electric vehicle charging load forecasting method under demand response according to claim 3 is characterized in that The method of classifying the target areas into types according to the functions of the areas specifically includes the following steps: The target area is represented in the form of a graph: the nodes are the center points of each block in the target area, and the streets in the target area are the lines between the nodes; In the figure, various areas are divided into residential areas, commercial areas and industrial areas according to their functions.

5. The short-term electric vehicle charging load forecasting method under demand response according to claim 4 is characterized in that The construction of the electric vehicle user travel model based on the electric vehicle travel data specifically includes the following steps: According to the travel destination and regional division, the travel starting point and end point of the electric vehicle user are expressed as Where D x is the travel purpose matrix of electric vehicle user x; D x mn represents the frequency of the route chosen by electric vehicle user x from travel area m to destination area n within one month, where m=j means the travel area is a residential area, m=s means the travel area is a commercial area, m=g means the travel area is an industrial area, n=j means the destination area is a residential area, n=s means the destination area is a commercial area, and n=g means the destination area is an industrial area; According to the actual travel data of electric vehicle users, the travel probability matrix P of electric vehicle user x is constructed according to the travel period, travel probability and travel purpose. x : P x =[P j P s P g ] Where P j is the travel probability matrix of electric vehicle users x with residential areas as their destinations; P s is the travel probability matrix of electric vehicle users x with the commercial area as the destination; P g is the travel probability matrix of electric vehicle users x with the industrial zone as the destination; P lnm represents the probability that electric vehicle user x travels from travel area m to destination area n during travel period l; the value of l ranges from 1 to 24; According to the constructed travel purpose matrix and travel probability matrix, the travel time-space diagram of electric vehicle users is established, which is expressed as ξ cx ={t,D,P t }; where ξ cx is the space-time diagram of electric vehicle users’ travel, t is the travel time, D is the matrix composed of the user’s travel purpose matrix, P t It is the matrix composed of the travel probability matrix of the user at travel time t.

6. The short-term electric vehicle charging load forecasting method under demand response according to claim 5 is characterized in that The construction of a user charging time-space diagram based on the charging pile time-sharing electricity price data and the electric vehicle SOC data specifically includes the following steps: Based on the charging pile time-sharing electricity price data P cd and electric vehicle user charging area D cd , the charging scenarios of electric vehicle users are further divided, expressed as P cd ∈{P cd_low ,P cd_flat ,P cd_high } D cd ∈{D cd-j ,D cd-s ,D cd-g } Ch={P cd ,D} Where P cd is the time-sharing electricity price data of the charging pile, P cd Including off-peak electricity price P cd_low 、Usual electricity price P cd_flat and peak electricity price P cd_high ; Electric vehicle user charging area D cd Including residential areas D cd-j 、Business District D cd-s and Industrial Zone D cd-g ; Electric vehicle user charging scenario Ch is composed of P cd and D cd Decide; The following formula is used to represent the user charging time-space diagram ξ cd : ξ cd ={t,D cd ,P cd_t ,SOC EV_t } Where P cd_t is the electricity price of the user at charging time t; SOC EV_t is the SOC value of the user's electric vehicle at charging time t.

7. The short-term electric vehicle charging load forecasting method under demand response according to claim 6 is characterized in that The construction of a short-term electric vehicle charging load prediction primary model based on the LSTM model and the attention mechanism described in step S3 specifically includes the following steps: The charging pile load data PW cd 、Electric vehicle user travel time and space diagramξ cx 、User charging time-space diagramξ cd and target area meteorological data as the model input, and electric vehicle charging load data as the model output; Set ω as the length of the time series window, and get the hidden state vector h corresponding to the input model at each moment t-ω ~h t ; Set the hidden matrix H to H = (h t-ω ,h t-ω+1 ,…,h t-1 ); One-dimensional convolution is used to extract the temporal pattern matrix H θ , expressed as Where θ represents the TCN filter; is the jth TCN filter C j The eigenvalue extracted in the i-th row; len is the padding length in the convolution calculation; t' is the maximum length of the weight; T j,(t'-ω+len) is the jth TCN filter C j The convolution kernel. The following formula is used as the evaluation function In the formula H θ is the vector of the i-th row of ; Α is the weight matrix; Calculate the parameter weight δ i for Attention A t The calculation formula is n is the characteristic quantity of input data; Finally, the short-term electric vehicle charging load p predicted by the model is obtained t for p t =A'(A h h t +A v A t ) Wherein h h t The weight matrix of v A t ' is the weight matrix; Α' is the total weight matrix.

8. A system for implementing the short-term electric vehicle charging load forecasting method under demand response as described in any one of claims 1 to 7, characterized in that It includes a data acquisition module, a data processing module, a model building module, a model training module and a load forecasting module; the data acquisition module, the data processing module, the model building module, the model training module and the load forecasting module are connected in series in sequence; the data acquisition module is used to obtain the historical data information of electric vehicle travel in the target area, and upload the data information to the data processing module; The data processing module is used to process the acquired data information according to the received data information to construct a training data set, and upload the data information to the model construction module; The model building module is used to build a primary model for short-term electric vehicle charging load prediction based on the received data information, based on the LSTM model and attention mechanism, and upload the data information to the model training module; The model training module is used to train the constructed short-term electric vehicle charging load prediction primary model according to the received data information and the obtained training data set, obtain the short-term electric vehicle charging load prediction model, and upload the data information to the load prediction module; The load forecasting module is used to forecast the actual short-term electric vehicle charging load in the target area based on the received data information and the obtained short-term electric vehicle charging load forecasting model.

Images