A vehicle speed prediction method based on two-level decomposition

Abstract

The present invention relates to the technical field of vehicle speed prediction, and in particular to a vehicle speed prediction method based on two-level decomposition. The method of the present invention comprises: first, using wavelet decomposition to decompose an original vehicle speed sequence into multiple high- and low-frequency components to reduce the non-stationarity of the original vehicle speed sequence; second, performing a secondary decomposition on the low-frequency components to further reduce the non-stationarity of the low-frequency components; and finally, using an RBF neural network to establish prediction models for each vehicle speed component, and superimposing the results of each prediction model to obtain a final vehicle speed prediction result. The vehicle speed prediction method of the present invention performs a secondary decomposition on the low-frequency components based on the primary decomposition, further reducing the non-stationarity of the vehicle speed and making the prediction more accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle speed prediction, and in particular to a vehicle speed prediction method based on two-level decomposition. Background Art

[0002] Speed ​​prediction accuracy plays a crucial role in the energy-saving effects of predictive energy management strategies for hybrid electric vehicles. Among current speed prediction methods, data-driven neural network prediction models demonstrate a certain advantage in accuracy. While these methods leverage the generalization capabilities of neural network models to predict vehicle speed in the short term, achieving some success, they fail to consider the impact of speed nonstationarity on prediction results. Decomposition followed by integration is a method that effectively reduces data nonstationarity and combines decomposition techniques with prediction models to improve model prediction accuracy.

[0003] Although the decomposition method can improve the non-stationarity of the sequence and improve the prediction accuracy, if the non-stationarity of the original sequence is high, the components after one decomposition are highly volatile, and the irregularity and nonlinearity are high, then the overall prediction effect will be affected.

[0004] In order to further reduce the non-stationarity of vehicle speed and obtain better vehicle speed prediction results, we propose a vehicle speed prediction method based on two-level decomposition of the present invention. Summary of the Invention

[0005] The purpose of the present invention is to provide a vehicle speed prediction method based on two-level decomposition, which is used to solve the problem of reduced prediction accuracy caused by the non-stationarity of vehicle speed when predicting vehicle speed in current technology.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A vehicle speed prediction method based on two-level decomposition, the steps are as follows:

[0008] S1. Collect historical vehicle speed data and establish the original vehicle speed sequence based on the time axis;

[0009] S2. Perform n-layer wavelet decomposition on the original vehicle speed sequence to obtain n high-frequency subsequences and one low-frequency subsequence;

[0010] S3. Perform m-layer wavelet decomposition on the low-frequency subsequence obtained in S2 to obtain one low-frequency subsequence and m high-frequency subsequences.

[0011] S4. Based on the n+1+m subsequences decomposed in S2 and S3, an RBF neural network multi-step vehicle speed prediction model is established through the RBF neural network;

[0012] S5. Using the RBF neural network multi-step speed prediction model and a direct multi-step prediction strategy, the vehicle speed is predicted in multiple steps, and the speed prediction output for each subsequence is obtained;

[0013] The vehicle speed prediction outputs of each subsequence are superimposed to obtain the final multi-step vehicle speed prediction result.

[0014] Furthermore, after S1, the following steps are also included:

[0015] S11. Divide the original vehicle speed sequence into a training set and a test set;

[0016] And what is decomposed in S2 is the training set.

[0017] Furthermore, in S4, the RBF neural network is defined as follows:

[0018]

[0019] Among them, y kj is the output corresponding to the kth input; i = 1, 2, ..., m, m is the number of hidden layer nodes; w ij is the weight from the hidden layer to the output layer; x k is the kth input of the neural network; c i is the center of the radial basis function; σ i is the standard deviation of the Gaussian function; ||x k -c i || is the Gaussian norm;

[0020] The input of the vehicle speed prediction model established by the RBF neural network is the vehicle speed over the past period of time, which is specifically represented by the vehicle speed at the current moment and the past four sampling moments. Its output is the vehicle speed at the next sampling point, and its sampling period is 1s.

[0021] Furthermore, in S4, an RBF neural network is used to establish an RBF neural network multi-step vehicle speed prediction model. When training each subsequence prediction model, the vehicle speed in the past period is defined as the RBF neural network input, and the future vehicle speed is defined as the expected output. The training target is the mean square error between the actual vehicle speed and the expected future vehicle speed. The calculation formula is as follows:

[0022]

[0023] Where, MSE is the mean square error of the entire working condition; p is the prediction time; v r (j+i) is the predicted vehicle speed at time j+i within the prediction time domain at time j; v(j+i) is the actual vehicle speed at time j+i within the prediction time domain at time j; n is the total length of the working condition;

[0024] When the MSE value reaches the preset training target or the maximum number of training times, the training is stopped and a multi-step vehicle speed prediction model is obtained.

[0025] Furthermore, in S5, the direct multi-step prediction strategy is expressed as follows:

[0026] [v t+1 ,v t+2 ,…,v t+p ]=f(v t ,v t-1 ,v t-2 ,…,v t-d-1 )

[0027] Among them, t is the current time, v t-1 ,v t-2 ,…,v t-d-1 and v t+1 ,v t+2 ,…,v t+p are the input and output of the vehicle speed prediction model respectively; v t-1 ,v t-2 ,…,v t-d-1 and v t+1 ,v t+2 ,…,v t+p The sequence length is 5.

[0028] Furthermore, after S4, the following steps are also included:

[0029] S41. Verify the prediction results of the RBF neural network multi-step vehicle speed prediction model through the test set.

[0030] The present invention has at least the following beneficial effects:

[0031] 1. Based on the primary decomposition, the present invention adopts the secondary decomposition to further reduce the non-stationarity of vehicle speed and further improve the vehicle speed prediction accuracy;

[0032] 2. The present invention is mainly used in energy management strategies based on model predictive control. It uses two-level decomposition and RBF neural network to predict vehicle speed in real time, effectively improving the speed prediction accuracy and thus improving the fuel economy of the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0034] Figure 1Flowchart of the vehicle speed prediction method of the present invention.

[0035] Figure 2 Schematic diagram of the training set and test set working condition data of the present invention.

[0036] Figure 3 This is a schematic diagram of the multi-step prediction of vehicle speed using the RBF neural network of the present invention.

[0037] Figure 4 This is a schematic diagram of the primary decomposition result of the present invention.

[0038] Figure 5 This is a schematic diagram of the secondary decomposition result of the low-frequency signal after the primary decomposition according to the present invention.

[0039] Figure 6 Schematic diagram of the prediction results of the two prediction methods of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0041] See also Figure 1 The present invention provides a vehicle speed prediction method based on two-level decomposition, as follows:

[0042] 1) Collect historical vehicle speed data and establish the vehicle original speed sequence based on the time axis.

[0043] For the convenience of data acquisition, the historical vehicle speed data of the present invention can be directly extracted from an existing database, which does not affect the data acquisition quality. Of course, it is also possible to actually collect a large amount of historical vehicle speed data for each vehicle.

[0044] The entire historical driving data is divided into two parts: training set and test set. The training set data set includes five standard driving conditions: China_urban, HWFET, LA92, UDDS and WLTC, and the test set data includes NEDC and CTLCP.

[0045] Please refer to the following for details: Figure 2 , which is a time series diagram of splitting the original vehicle speed data into a training set and a test set in a specific embodiment of the present invention.

[0046] 2) See Figure 4 , perform three-layer wavelet decomposition on the training set working conditions, and obtain a low-frequency subsequence with similar amplitude and change trend to the original sequence and three high-frequency subsequences with smaller amplitudes; refer to Figure 5, perform three-layer wavelet decomposition on the low-frequency subsequence to obtain one low-frequency subsequence and three high-frequency subsequences; a total of 7 subsequences are obtained after two decompositions.

[0047] 3) The two-level decomposition in step 2) is also performed on the training set conditions to obtain 7 subsequence data, and an RBF neural network multi-step vehicle speed prediction model is established. The RBF neural network calculation definition is as follows:

[0048]

[0049] Among them, y kj is the output corresponding to the kth input; i = 1, 2, ..., m, m is the number of hidden layer nodes; w ij is the weight from the hidden layer to the output layer; x k is the kth input of the neural network; c i is the center of the radial basis function; σ i is the standard deviation of the Gaussian function; ||x k -c i || is the Gaussian norm.

[0050] The input of the RBF neural network speed prediction model is the speed of the vehicle over the past period of time, which is specifically represented by the speed of the vehicle at the current moment and the past four sampling moments, a total of five sampling points, recorded as [v t ,v t-1 ,…,v t-4 ], and its output is the vehicle speed at the next five sampling points, recorded as [v t+1 ,v t+2 ,…,v t+5 ], where the sampling period is 1s.

[0051] Figure 3 This is the schematic diagram of the RBF neural network prediction model mentioned above.

[0052] 4) For RBF neural network, use MATLAB's RBF neural network toolbox, and its calling format is as follows:

[0053] net=newrb(P,T,goal,spread,MN,DF)

[0054] Among them, P is the input matrix; T is the target matrix; goal is the mean square error; spread is the expansion coefficient of the basis function; MN is the maximum number of neurons; DF is the number of neurons added between two displays during training.

[0055] When the present invention uses the RBF neural network to train the seven subsequences, the vehicle speed in the past period is defined as the RBF neural network input, the future vehicle speed is defined as the expected output, and the training target is the mean square error between the actual vehicle speed and the expected output speed. The calculation formula is as follows:

[0056]

[0057] Where, MSE is the mean square error of the entire working condition; p is the prediction time; v r (j+i) is the predicted vehicle speed at time j+i within the prediction time domain at time j; v(j+i) is the actual vehicle speed at time j+i within the prediction time domain at time j; n is the total length of the working condition;

[0058] When the MSE value reaches a preset training target or reaches a maximum number of training times, the training is stopped to obtain a multi-step vehicle speed prediction model.

[0059] 5) Use the direct multi-step prediction strategy to perform multi-step prediction, the expression is as follows:

[0060] [v t+1 ,v t+2 ,…,v t+p ]=f(v t ,v t-1 ,v t-2 ,…,v t-d-1 )

[0061] Among them, t is the current time, v t-1 ,v t-2 ,…,v t-d-1 and v t+1 ,v t+2 ,…,v t+p are the input and output of the vehicle speed prediction model respectively; v t-1 ,v t-2 ,…,v t-d-1 and v t+1 ,v t+2 ,…,v t+p The length of the sequence is 5;

[0062] The outputs of each subsequence vehicle speed prediction model are superimposed to obtain the final multi-step vehicle speed prediction result.

[0063] 6) Experimental verification

[0064] The data of the test set are respectively used to predict the vehicle speed using the combined prediction method based on two-level decomposition + RBF neural network and the single RBF neural network prediction method of the present invention, and compared with the actual data of the test set. The prediction results are compared as shown in the figure below. Figure 6 shown.

[0065] In order to quantitatively evaluate the effect of the vehicle speed prediction method proposed in this invention, the root mean square error (RMSE) is used for quantitative evaluation, and its calculation formula is as follows:

[0066]

[0067] The comparison results of the vehicle speed prediction results are shown in the following table:

[0068] Table 1: Vehicle speed prediction results of single RBF prediction method and the method of the present invention

[0069]

[0070] pass Figure 6 As can be seen from Table 1, based on the given test conditions, the combined prediction method based on two-level decomposition and RBF neural network proposed in the present invention improves the RMSE of the vehicle speed prediction results by 60.22% over the entire test conditions compared with the single RBF neural network prediction method.

[0071] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A vehicle speed prediction method based on two-level decomposition, characterized by: Here are the steps: S1. Collect historical vehicle speed data and establish the original vehicle speed sequence based on the time axis; S2. Perform n-layer wavelet decomposition on the original vehicle speed sequence to obtain n high-frequency subsequences and one low-frequency subsequence; S3. Perform m-layer wavelet decomposition on the low-frequency subsequence obtained in S2 to obtain one low-frequency subsequence and m high-frequency subsequences. S4. Based on the n+1+m subsequences decomposed in S2 and S3, an RBF neural network multi-step vehicle speed prediction model is established through the RBF neural network; S5. Using the RBF neural network multi-step speed prediction model and a direct multi-step prediction strategy, the vehicle speed is predicted in multiple steps, and the speed prediction output for each subsequence is obtained; The vehicle speed prediction outputs of each subsequence are superimposed to obtain the final vehicle speed multi-step prediction result; The input of the RBF neural network speed prediction model is the vehicle speed over the past period of time, specifically the vehicle speed at the current moment and the past four sampling moments, a total of five sampling points. Its output is the vehicle speed at the next five sampling points, where the sampling period is 1s.

2. The vehicle speed prediction method based on two-level decomposition according to claim 1, characterized in that: After S1, the following steps are also included: S11. Divide the original vehicle speed sequence into a training set and a test set; And what is decomposed in S2 is the training set.

3. The vehicle speed prediction method based on two-level decomposition according to claim 1, characterized in that: In S4, the RBF neural network is defined as follows: Among them, y kj is the output corresponding to the kth input; i = 1, 2, ..., m, m is the number of hidden layer nodes; w ij is the weight from the hidden layer to the output layer; x k is the kth input of the neural network; c i is the center of the radial basis function; σ i is the standard deviation of the Gaussian function; ||x k -c i || is the Gaussian norm.

4. The vehicle speed prediction method based on two-level decomposition according to claim 1, characterized in that: In S4, an RBF neural network is used to establish an RBF neural network multi-step vehicle speed prediction model. When training each subsequence prediction model, the vehicle speed in the past period is defined as the RBF neural network input, and the future vehicle speed is defined as the expected output. The training target is the mean square error between the actual vehicle speed and the expected future vehicle speed. The calculation formula is as follows: Where, MSE is the mean square error of the entire working condition; p is the prediction time; v r (j+i) is the predicted vehicle speed at time j+i within the prediction time domain at time j; v(j+i) is the actual vehicle speed at time j+i within the prediction time domain at time j; n is the total length of the working condition; When the MSE value reaches the preset training target or the maximum number of training times, the training is stopped and a multi-step vehicle speed prediction model is obtained.

5. The vehicle speed prediction method based on two-level decomposition according to claim 1, characterized in that: In S5, the direct multi-step prediction strategy is expressed as follows: [v t+1 ,v t+2 ,…,v t+p ]=f(v t ,v t-1 ,v t-2 ,…,v t-d-1 ) Among them, t is the current time, v t-1 ,v t-2 ,…,v t-d-1 and v t+1 ,v t+2 ,…,v t+p are the input and output of the vehicle speed prediction model respectively; v t-1 ,v t-2 ,…,v t-d-1 and v t+1 ,v t+2 ,…,v t+p The sequence length is 5.

6. The vehicle speed prediction method based on two-level decomposition according to claim 2, characterized in that: After S4, the following steps are also included: S41. Verify the prediction results of the RBF neural network multi-step vehicle speed prediction model through the test set.

Images