Charging demand prediction method based on multi-regression model fusion

Abstract

The invention discloses a charging demand prediction method based on multi-regression model fusion, and relates to the field of new energy automobiles, and the method comprises the steps: carrying out the clustering of the charging behaviors of a target vehicle through employing a trained behavior clustering model based on the historical charging data of the target vehicle, and obtaining n types of charging behavior types; based on the historical charging data of the target vehicle and the n category labels, calculating the occurrence probability of each charging behavior category by using a trained behavior recognition model; based on historical charging data of the target vehicle, utilizing a trained single charging demand prediction model corresponding to each category label to respectively predict the charging demand of the target vehicle under each charging behavior category; and performing weighted summation on the charging demand of the target vehicle under the n charging behavior categories and the occurrence probability of each charging behavior category by using the trained weighting model to obtain the future charging demand of the target vehicle. According to the invention, the accuracy of new energy vehicle charging demand prediction is improved.

Description

Technical Field

[0001] The present application relates to the field of new energy vehicle technology, and in particular to a charging demand prediction method based on multi-regression model fusion. Background Art

[0002] In recent years, the number of new energy vehicles in my country has shown a trend of rapid growth, and the demand for charging has increased significantly, which has also brought about a sharp increase in charging load. The large-scale charging load has brought certain challenges to the normal operation of the distribution network. Therefore, it is necessary to formulate a reasonable and effective charging strategy for the interaction between new energy vehicles and distribution networks, so that it can not only ensure the charging service quality of new energy vehicle users, but also ensure the power safety of the distribution network.

[0003] In the prior art, the formulation of new energy vehicle charging strategies relies on accurate prediction of charging demand, and the prediction of charging demand is mainly reflected in the prediction of charging energy and access to the power grid. In order to effectively predict the charging demand of new energy vehicles, statistical characteristics are usually used to establish models. The statistical model based on travel characteristics extracts the statistical parameters and empirical distribution of travel modes such as vehicle ownership, departure time, arrival time and travel time, and uses these parameters to generate deterministic or random vehicle travel and charging modes for charging demand statistics.

[0004] Although the above method can predict the charging demand of new energy vehicles, it assumes that all vehicles follow a unified travel pattern and predefined charging scenarios, ignoring the heterogeneity and randomness of individual users and their specific travel patterns. This may cause the statistical results of charging demand to deviate from reality, making the final predicted charging demand inaccurate and biased. Summary of the invention

[0005] The purpose of this application is to provide a charging demand prediction method based on multi-regression model fusion, which improves the accuracy of new energy vehicle charging demand prediction.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a charging demand prediction method based on multi-regression model fusion, and the charging demand prediction method based on multi-regression model fusion includes:

[0008] Acquire historical charging data of a target vehicle; the target vehicle is a new energy vehicle;

[0009] Based on the historical charging data of the target vehicle, the charging behavior of the target vehicle is clustered using the trained behavior clustering model to obtain n charging behavior categories; one charging behavior category corresponds to one category label; n is an integer greater than or equal to 1;

[0010] Based on the historical charging data of the target vehicle and n category labels, use the trained behavior recognition model to calculate the occurrence probabilities of each charging behavior category;

[0011] Based on the n category labels, determine the trained single charging demand prediction models corresponding to each category label; each trained single charging demand prediction model integrates multiple regression models;

[0012] Based on the historical charging data of the target vehicle, use the trained single charging demand prediction models corresponding to each category label to predict the charging demands of the target vehicle under each charging behavior category respectively; the charging demands include charging energy and grid connection duration;

[0013] Use the trained weighted model to perform weighted summation of the charging demands of the target vehicle under n charging behavior categories and the occurrence probabilities of each charging behavior category to obtain the future charging demand of the target vehicle.

[0014] In a second aspect, the present application also provides a computer system, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the charging demand prediction method based on multi-regression model fusion described in the first aspect.

[0015] According to the specific embodiments provided by the present application, the following technical effects are disclosed:

[0016] The present application does not uniformly define the travel patterns and charging scenarios of new energy vehicles, but directly analyzes and processes the historical charging data of the target vehicle, fully considering the heterogeneity and randomness of different users and their travel patterns. In view of the characteristics of high complexity and numerous categories of the charging behaviors of new energy vehicles, the present application first uses the trained behavior clustering model and behavior recognition model to cluster and divide the charging behaviors of the target vehicle, and calculates the occurrence probabilities of each charging behavior category according to the clustering results. Secondly, the present application also uses the trained single charging demand prediction models respectively according to the clustering results to predict the charging demands of the target vehicle under each charging behavior category. Finally, the future charging demand of the target vehicle is determined by weighted summation of the above prediction results. It is precisely because the present application can cluster based on the historical charging data of the target vehicle and use the single charging demand prediction models respectively according to the clustering results that the accuracy of predicting the future charging demand of the target vehicle can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for use in the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0018] Figure 1 It is a flowchart of the charging demand prediction method based on multi - regression model fusion provided by the embodiment of the present application;

[0019] Figure 2 It is an execution framework diagram of the charging demand prediction method based on multi - regression model fusion provided by the embodiment of the present application;

[0020] Figure 3 It is a clustering tree diagram of new energy vehicle charging behavior provided by the embodiment of the present application;

[0021] Figure 4 It is an internal structure diagram of the computer system provided by the embodiment of the present application. Detailed implementation manners

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of them. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0023] The purpose of the present application is to provide a charging demand prediction method based on multi - regression model fusion, which improves the accuracy of new energy vehicle charging demand prediction.

[0024] To make the above - mentioned objects, features, and advantages of the present application more obvious and understandable, the following will further describe the present application in detail in conjunction with the accompanying drawings and specific implementation manners.

[0025] Embodiment 1

[0026] This embodiment provides a charging demand prediction method based on multi - regression model fusion. As Figure 1 shown, the charging demand prediction method based on multi - regression model fusion includes:

[0027] Step S1: Obtain the historical charging data of the target vehicle.

[0028] In this embodiment, the target vehicle is a new energy vehicle, and its historical charging data includes: charging vehicle model, charging start time, charging end time, charging start mileage, charging end mileage, charging start state of charge (SOC), charging end SOC, charging duration, total power of the power battery, pure electric cruising range, maximum charging power, minimum charging power, longitude and latitude of the charging location, etc. In addition, the above historical charging data is obtained from the national monitoring and management platform for new energy vehicles.

[0029] Step S2: Based on the historical charging data of the target vehicle, use the trained behavior clustering model to cluster the charging behavior of the target vehicle to obtain n types of charging behavior categories; n is an integer greater than or equal to 1.

[0030] In this embodiment, first construct a behavior clustering model based on the variational Bayesian Gaussian mixture model (VB-GMM) of the Gaussian mixture model, and then obtain training data from the national monitoring and management platform for new energy vehicles. The training data is the historical charging data of different vehicles within a set time period, and this training data can consider the heterogeneous charging needs of users from the complete time chain, and use this training data to effectively train the VB-GMM.

[0031] Since new energy vehicles have the characteristics of complex and uncertain charging behaviors, it is difficult to accurately describe them with a single category. In order to reduce the determination of the number of categories by the artificial intervention clustering algorithm, the VB-GMM is used to adaptively adjust the clustering quantity. Among them, the expression of the overall probability density function is:

[0032]

[0033] In the formula, f VB-GMM (·) is the overall probability density function, z is the set of variables obtained by reconstructing the given features through L1 regularization, and satisfies z = (z1, z2,..., z i ) = {[x1, y1], [x1, y1],..., [x i , y i}, [x i , y i represents the i-th input data and output result, G represents the set of VB-GMM, π g is the weight of the g-th sub-distribution model, is a multi-Gaussian distribution model with estimated parameters θ g , including the mean μ g and the covariance matrix Σ g , and d is the dimension of the feature tuple.

[0034] Furthermore, VB - GMM is defined as a full - rank covariance matrix, and this setting can enable each feature to have a different standard covariance matrix. At the same time, the parameter θ in the multi - Gaussian distribution model g is estimated using the Expectation - Maximization (EM) algorithm, and its solution process can be divided into two steps: In the E - step, according to Bayes' rule, the multi - Gaussian distribution model needs to use the parameter θ g to calculate the probability of x i from the Gaussian distribution g; in the M - step, the multi - Gaussian distribution model needs to seek to maximize the likelihood of the estimated parameters until sufficient convergence. The specific solution process is as follows:

[0035] (1) Define the expression for estimating model parameters:

[0036]

[0037] where θ is the parameter to be estimated, and z i is the i - th latent variable.

[0038] (2) Determine the input information: The input data x=(x1, x2,..., x i ), the joint distribution p(x i , z i ; θ), the conditional distribution p(z i |x i , θ j ), and the maximum number of iterations J.

[0039] (3) Randomly initialize the initial value θ0 of the model parameter θ.

[0040] (4) Start the EM algorithm iteration process j = 1, 2,..., J.

[0041] (5) In the E - step, calculate the conditional probability expectation of the joint distribution:

[0042] Q i (z i ) = p(z i |x i , θ j );

[0043]

[0044] where Q i (z i ) is a newly introduced distribution and is the posterior probability of the sample.

[0045] (6) In the E - step, maximize l(θ, θ j ) to obtain θj+1 :

[0046] θ j+1 = argmaxl(θ, θ j );

[0047] If convergence has been achieved, the algorithm ends; if not, execute (5), (6) until convergence.

[0048] Based on the above-trained behavior clustering model, a clustering tree for new energy vehicle charging behaviors as shown in Figure 3 is constructed. The clustering tree for new energy vehicle charging behaviors first clusters the charging behaviors of vehicles according to noon, afternoon, and night, and then clusters them according to the parking duration after charging being less than 10 minutes, the parking duration after charging being greater than 10 minutes and less than 60 minutes, and the parking duration after charging being greater than 60 minutes. Finally, a total of 9 charging behavior categories are obtained, and one charging behavior category corresponds to one category label.

[0049] Step S3: Based on the historical charging data of the target vehicle and n category labels, use the trained behavior recognition model to calculate the occurrence probabilities of each charging behavior category.

[0050] In this embodiment, in order to improve the accuracy of charging demand prediction, the recognition probability is incorporated into the prediction model, and a behavior recognition model is established using the charging behavior clustering results. Due to the uncertainty of charging behaviors, the Light Gradient Boosting Machine (LightGBM) is used to provide a "soft classification" result for each charging behavior category, that is, the classification result is output in the form of probability, increasing the categories describing charging behaviors. LightGBM improves the computational efficiency of histogram optimization of charging behavior data, and its objective function is as follows:

[0051]

[0052] In the formula, l L (·) is the loss function, y i is the true label, is the predicted value of the model, is the sum of the complexities of all the trees in the model, which is set in advance as a regularization term in the optimization objective, and f j is the regression equation of the jth tree. In addition, since this problem is a multi-classification model, the clustering results of VB-GMM should be converted into a matrix consisting of only 0 and 1 using binary encoding, and this process can support the performance evaluation of the model. Finally, applying the "soft classification" of LightGBM can output the probability of each category, and this result will be applied in each single charging demand prediction model.

[0053] Step S4: Based on the n category labels, determine the trained single charging demand prediction models corresponding to each category label.

[0054] In this embodiment, each single charging demand prediction model is constructed based on the Stacking model and uses grid search and k-fold cross-validation. Each trained single charging demand prediction model integrates multiple regression models. The trained single charging demand prediction model includes two layers of learners, namely the first layer of learners and the second layer of learners. The first layer of learners includes multiple base learners, and the second layer of learners includes a meta-learner. Since the model structure applied in the base learners of the first layer is relatively complex, when establishing the meta-learner in the second layer, a machine learning model with a simple structure is preferred to reduce the risk of model overfitting. Finally, the overall training process can be described by the following formula:

[0055]

[0056] In the formula, is the set of base learners, and each base learner is represented by the number r, P r is the training set, T r is the test set, and Y pred is the prediction result of the test set.

[0057] Furthermore, the multiple base learners are respectively the LightGBM regression model, the Ridge Regression (RR) model, and the Random Forest (RF) regression model, and the meta-learner is the Logistic Regression (LR) model.

[0058] Among them, the RR regression model is a LR model with an L2 criterion. Its regression function is basically the same as that of the LR model. Its regression function and loss function are specifically as follows:

[0059] H RR (x i ) = ω T x i + b;

[0060]

[0061] In the formula, ω is the weight coefficient vector, b is the intercept term, and λ is used to control the degree of penalty.

[0062] The RF regression model is an ensemble learning algorithm of the Bagging technique, which can reduce the prediction difference by calculating the average value of different trees. The expression of the RF regression model is as follows:

[0063]

[0064]

[0065] Wherein, H RF (·) is the prediction result of the RF regression model, N is the number of weak learners, and f n (·) is the regression result of the nth tree, I(·) is the state function, and if x is in R m , then this value is 1, is the average value of y in R m space, and M is the number of spaces.

[0066] Taking into account the calculation efficiency, calculation accuracy, and data volume, the LightGBM regression model is finally selected, and its expression is as follows:

[0067]

[0068] Wherein, Φ k is the weight of the kth tree, and f k (·) is the regression equation of the kth tree.

[0069] Step S5: Based on the historical charging data of the target vehicle, use the trained single charging demand prediction models corresponding to various category labels to predict the charging demands of the target vehicle under each charging behavior category respectively.

[0070] In this embodiment, the charging demand includes charging energy and grid connection duration.

[0071] Step S6: Use the trained weighted model to perform weighted summation on the charging demands of the target vehicle under n charging behavior categories and the occurrence probabilities of each charging behavior category to obtain the future charging demand of the target vehicle.

[0072] In this embodiment, the formula for calculating the future charging demand of the target vehicle is:

[0073]

[0074] Wherein, y i is the future charging demand of the ith target vehicle, x i is the historical charging data of the ith target vehicle, and H stacking,c (x i ) is the charging demand of the ith target vehicle under the cth charging behavior category, is the occurrence probability of the cth charging behavior category, and C is the number of charging behavior categories.

[0075] In summary, the charging demand prediction method based on multi-regression model fusion has the following advantages:

[0076] (1) In view of the characteristics of high complexity and variety of new energy vehicle charging behaviors, a behavior clustering model based on variational Bayesian Gaussian mixture model is established, and a charging behavior clustering tree of new energy vehicles is constructed.

[0077] (2) Based on the stacking technology of multi-regression model fusion, a charging demand prediction model for new energy vehicles considering behavior recognition probability is established. By comparing with the traditional ensemble learning model without behavior classification and the single model with behavior classification, accurate prediction of charging energy and vehicle grid connection duration can be achieved.

[0078] Example 2

[0079] This embodiment provides a computer system, which can be a server or a terminal, and its internal structure diagram can be as shown in Figure 4 the figure. The computer system includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of this computer system is used to provide computing and control capabilities. The memory of this computer system includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of this computer system is used to store the historical charging data of the target vehicle. The input / output interface of this computer system is used to exchange information between the processor and external devices. The communication interface of this computer system is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, it realizes a charging demand prediction method based on multi-regression model fusion.

[0080] Those skilled in the art can understand that Figure 4 the structure shown in is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0081] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0082] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memories can include read-only memory (ROM), magnetic tapes, floppy disks, flash memories, optical memories, high-density embedded non-volatile memories, resistive random access memories (ReRAM), magnetoresistive random access memories (MRAM), ferroelectric random access memories (FRAM), phase change memories (PCM), graphene memories, etc. Volatile memories can include random access memory (RAM) or external cache memories, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0083] The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logics, data processing logics based on quantum computing, etc., without limitation.

[0084] All actions of obtaining signals, information, or data in the present application are carried out on the premise of complying with the corresponding data protection regulations and policies of the country where it is located and with the authorization given by the owner of the corresponding device.

[0085] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other.

[0086] In this article, specific examples are used to illustrate the principle and implementation of this application. The description of the above embodiments is only for helping to understand the method and its core idea of this application; at the same time, for those of ordinary skill in the art, according to the idea of this application, there will be changes in the specific implementation and application scope. In summary, the content of this specification should not be construed as a limitation to this application.

Claims

1. A charging demand prediction method based on multi-regression model fusion, characterized in that: The charging demand prediction method based on multi-regression model fusion includes: Acquire historical charging data of a target vehicle; the target vehicle is a new energy vehicle; Based on the historical charging data of the target vehicle, the charging behavior of the target vehicle is clustered using the trained behavior clustering model to obtain n charging behavior categories; one charging behavior category corresponds to one category label; n is an integer greater than or equal to 1; Based on the historical charging data of the target vehicle and n category labels, the occurrence probability of each charging behavior category is calculated using the trained behavior recognition model; Based on n category labels, determine the trained single charging demand prediction model corresponding to each category label; each trained single charging demand prediction model integrates multiple regression models; Based on the historical charging data of the target vehicle, the charging demand of the target vehicle under each charging behavior category is predicted using the trained single charging demand prediction model corresponding to each category label; the charging demand includes charging energy and grid access time; Using the trained weighted model, the charging demand of the target vehicle under n charging behavior categories and the occurrence probability of each charging behavior category are weighted and summed to obtain the future charging demand of the target vehicle.

2. The charging demand prediction method based on multi-regression model fusion according to claim 1 is characterized in that: The historical charging data includes at least: charging vehicle model, charging start time, charging end time, charging start mileage, charging end mileage, charging start SOC, charging end SOC, charging maximum power, charging minimum power, charging location longitude and latitude, and total power battery energy.

3. The charging demand prediction method based on multi-regression model fusion according to claim 1 is characterized in that: The charging behavior of the target vehicles is first clustered according to noon, afternoon and night, and then clustered according to the parking time after charging being less than 10 minutes, the parking time after charging being greater than 10 minutes and less than 60 minutes, and the parking time after charging being greater than 60 minutes.

4. The charging demand prediction method based on multi-regression model fusion according to claim 1 is characterized in that: The behavioral clustering model is constructed based on VB-GMM.

5. The charging demand prediction method based on multi-regression model fusion according to claim 1 is characterized in that: The behavior recognition model is built based on LightGBM.

6. The charging demand prediction method based on multi-regression model fusion according to claim 1 is characterized in that: The single charging demand prediction model is based on the Stacking model and constructed using grid search and k-layer cross validation.

7. The charging demand prediction method based on multi-regression model fusion according to claim 6 is characterized in that: The single charging demand prediction model includes two layers of learners, namely a first layer learner and a second layer learner; the first layer learner includes multiple base learners; and the second layer learner includes a meta learner.

8. The charging demand prediction method based on multi-regression model fusion according to claim 7 is characterized in that: The multiple base learners are respectively a LightGBM regression model, a RR regression model and a RF regression model; the meta learner is an LR regression model.

9. The charging demand prediction method based on multi-regression model fusion according to claim 1 is characterized in that: The formula for calculating the future charging demand of the target vehicle is: In the formula, y i is the future charging demand of the i-th target vehicle, x i is the historical charging data of the i-th target vehicle, H stacking,c (x i ) is the charging demand of the i-th target vehicle under the c-th charging behavior category, is the occurrence probability of the cth charging behavior category, and C is the number of charging behavior categories.

10. A computer system comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the charging demand prediction method based on multi-regression model fusion as described in any one of claims 1 to 9.

Images