CNN-BiLSTM neural network vehicle speed prediction method, device and equipment based on ISSA

Abstract

The invention relates to the technical field of vehicle speed prediction, and provides a CNN-BiLSTM neural network vehicle speed prediction method, device and equipment based on ISSA, and the method comprises the steps: collecting a vehicle operation data set, extracting original time-vehicle speed data, and forming a training sample data set; the SSA algorithm is improved to obtain an ISSA algorithm; constructing a CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and performing learning training on the training sample data set by using the model; and vehicle speed prediction is carried out by using a CNN-BiLSTM neural network vehicle speed prediction model based on ISSA. According to the invention, the prediction precision of the vehicle speed prediction neural network model is improved, so that the application of the vehicle speed prediction neural network model in practice is improved.

Description

Technical Field

[0001] The embodiments of the present invention generally relate to the field of vehicle speed prediction technology, and more particularly to a CNN-BiLSTM neural network vehicle speed prediction method, apparatus, and device based on ISSA. Background Art

[0002] In the field of modern transportation, vehicle energy management systems (EMS) have become an important means to improve vehicle dynamic performance, reduce energy consumption, and optimize energy utilization efficiency. The control effect of predictive energy management strategies is closely related to the accuracy of the prediction model, and the accuracy of vehicle speed prediction directly determines the vehicle's energy-saving effect. Currently, common methods for vehicle speed prediction include prediction based on physical models, statistical models, and machine learning models. Among these methods, machine learning-based speed prediction models, such as long short-term memory networks (LSTMs), bidirectional LSTMs (BiLSTMs), and support vector machines (SVMs), have gradually become mainstream because neural networks achieve higher-precision vehicle speed predictions due to their robustness and nonlinear mapping capabilities.

[0003] Currently, the neural network models commonly used for speed prediction, while capable of prediction, suffer from low accuracy, with the predicted speed often differing significantly from the actual speed. To improve the accuracy of speed prediction models, some researchers have enhanced the neural network model's architecture, thereby enhancing its fitting and generalization capabilities and ultimately improving speed prediction accuracy. However, simply increasing network complexity fails to fully consider the impact of complex architecture on prediction results, resulting in limited improvements in model accuracy. Ultimately, the error between predicted and actual speeds is large, and speed prediction accuracy remains insufficient. Summary of the Invention

[0004] To address the above problems, the present invention combines a convolutional neural network with a bidirectional long short-term memory network. This not only captures the temporal dependence and local spatial features in vehicle speed data, but also enables in-depth analysis of vehicle speed sequences from different perspectives. The ISSA algorithm is used to optimize the L2 regularization coefficient, number of neurons, and learning rate hyperparameters of the CNN-BiLSTM model, significantly improving the model's fitting and generalization capabilities.

[0005] According to an embodiment of the present invention, a method, apparatus and device for vehicle speed prediction using a CNN-BiLSTM neural network based on ISSA are provided.

[0006] In a first aspect of the present invention, a method for vehicle speed prediction based on a CNN-BiLSTM neural network using ISSA is provided. The method comprises:

[0007] Step S01: collecting a vehicle operation dataset for ISSA-based CNN-BiLSTM neural network speed prediction, extracting original time-speed data using a sliding time window method to form a training sample dataset;

[0008] Step S02: using Circle chaos mapping, sine and cosine and Gaussian chaos strategies to improve the SSA algorithm to obtain the ISSA algorithm;

[0009] Step S03: constructing a CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and using the CNN-BiLSTM neural network vehicle speed prediction model to train the training sample data set;

[0010] Step S04: Use Matlab software to use the historical vehicle speed of n time steps after the sliding time window operation as the input of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and the vehicle speed of the next m time steps as the output of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA to perform vehicle speed prediction.

[0011] Furthermore, the steps of the ISSA algorithm described in step S02 are:

[0012] Step S021: Initialize parameters: sparrow population size, maximum number of iterations, dimension of each sparrow position, individual discoverer ratio, and warning threshold;

[0013] Step S022: Evaluate the fitness value;

[0014] Step S023: Initialize the Circle chaotic map population, and its expression is:

[0015]

[0016] Among them, x n+1 is the chaotic variable of the next iteration, x n is the chaotic variable of the current iteration, 3.85 is the control parameter of the chaotic mapping, which is used to adjust the chaotic behavior, sin(3.85πx n ) is the nonlinear term of the sine function, which increases the complexity of the system. Mod(·,1) is the modulo operation, which ensures that the variable is always in the interval [0,1].

[0017] Step S024: Update the discoverer's location: The sine and cosine strategies are introduced into the discoverer's location update formula. The improved discoverer's location update formula is:

[0018]

[0019] in, Update the position of the jth dimension for the i-th sparrow individual in the t+1th iteration, Update the position of the i-th sparrow in the j-th dimension in the t-th iteration, X best is the optimal individual in the current population, r2 is a random number in [0, 2π], which determines the moving distance of the sparrow; r3 is a random number in [0, 2π], which controls the influence of the optimal individual on the next position of the sparrow; ST is a constant in [0.5, 1], which represents the safety value; R2 is a random number in [0, 1], which represents the warning value; α is the cosine decreasing weight factor;

[0020] Step S025: Follower position update: The Gaussian perturbation strategy is introduced into the follower position update formula. The improved follower position update formula is:

[0021]

[0022] in, is the updated position of the jth dimension for the i-th sparrow individual in the t+1th iteration, Update the position of the i-th sparrow in the j-th dimension in the t-th iteration, X best is the optimal individual in the current population, r1 is a random factor, which is a uniformly distributed random number between [0,1], σ is the Gaussian perturbation coefficient, which is set to [0.01,0.1], and γ(0,1) is the standard normal distribution noise with a mean of 0 and a standard deviation of 1;

[0023] Step S026: Update the position of the sentinel.

[0024] Furthermore, the expression of the cosine decreasing weight factor α in step S024 is:

[0025]

[0026] Among them, α max is the initial weight value, α min is the final weight value, t is the current number of iterations, T is the maximum number of iterations, and cos is the cosine function.

[0027] Furthermore, the process of the ISSA-based CNN-BiLSTM neural network vehicle speed prediction model described in step S03 is as follows:

[0028] The local features of the input historical vehicle speed sequence are extracted through a one-dimensional convolutional layer;

[0029] Downsampling is performed through the maximum pooling layer to obtain local temporal features, reducing the feature dimension while retaining key feature information;

[0030] The local time series features are passed into the BiLSTM structure. BiLSTM introduces two LSTM units, forward and reverse, to process the forward and backward dependency information of the time series respectively.

[0031] The BiLSTM layer outputs high-dimensional time series features, which are passed to the fully connected layer and mapped into the predicted vehicle speed value for the next m time steps.

[0032] Furthermore, the steps of using the ISSA-based CNN-BiLSTM neural network vehicle speed prediction model to learn and train the training sample data set in step S03 are as follows:

[0033] Step S031: importing the training sample data set described in step S01, dividing it into a training set and a test set, and performing normalization processing;

[0034] Step S032: setting the upper and lower bounds of the initialization parameters of the ISSA algorithm, including the size of the sparrow population, the maximum number of iterations, the amount of optimization parameters, and the optimization parameters;

[0035] Step S033: using the ISSA algorithm to optimize the L2 regularization expression parameters, the number of neurons, and the learning rate parameters of the CNN-BiLSTM neural network vehicle speed prediction model to obtain the ISSA-based CNN-BiLSTM neural network vehicle speed prediction model;

[0036] Step S034: Train the ISSA-based CNN-BiLSTM neural network vehicle speed prediction model. When the loss function value reaches a set value or no longer decreases significantly, stop training.

[0037] Furthermore, the vehicle speed prediction expression in step S04 is:

[0038] [v t+1 ,v t+2 ,v t+3 ,…,v t+p ]=f CNN-BiLSTM (v t ,v t-1 ,v t-2 ,…v t-h+1 )

[0039] Among them, t is the current time, v t+1 ,v t+2 ,v t+3 ,…,v t+p is the predicted output speed of the speed prediction model, v t ,v t-1 ,v t-2 ,…v t-h+1 is the input speed of the speed prediction model, v t+1 ,vt+2 ,v t+3 ,…,v t+p is the length of the output speed sequence, v t ,v t-1 ,v t-2 ,…v t-h+1 is the length of the input vehicle speed sequence, f CNN-BiLSTM It is a CNN-BiLSTM neural network vehicle speed prediction model based on ISSA.

[0040] Furthermore, the root mean square error is used as an indicator to evaluate the accuracy of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and the average root mean square error is used to measure the overall prediction effect of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA. The expression is:

[0041]

[0042] Among them, RMSE is the root mean square error, ARMSE is the average root mean square error, p is the prediction time, T is, v j To predict the actual vehicle speed in the time domain, is the predicted vehicle speed in the prediction time domain, and T is the total number of vehicle speed sample sequences.

[0043] In a second aspect of the present invention, a device for predicting vehicle speed using a CNN-BiLSTM neural network based on ISSA is provided. The device comprises:

[0044] Data acquisition module: used to collect vehicle operation data sets for ISSA-based CNN-BiLSTM neural network speed prediction, and use the sliding time window method to extract the original time-speed data to form a training sample data set;

[0045] ISSA algorithm module: used to improve the SSA algorithm using Circle chaos mapping, sine-cosine and Gaussian chaos strategies to obtain the ISSA algorithm;

[0046] Model building module: used to build a CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and use the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA to learn and train the training sample data set;

[0047] Vehicle speed prediction module: This module uses Matlab software to use the historical vehicle speed of n time steps after the sliding time window operation as the input of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and the vehicle speed of m time steps in the future as the output of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA to perform vehicle speed prediction.

[0048] In a third aspect of the present invention, an electronic device is provided, comprising: a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the program, the method according to the first aspect of the present invention is implemented.

[0049] In a fourth aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the method according to the first aspect of the present invention is implemented.

[0050] It should be understood that the contents described in the summary of the invention are not intended to limit the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become readily understood through the following description.

[0051] The beneficial effects of the present invention are:

[0052] 1. By combining a convolutional neural network with a bidirectional long short-term memory network, this invention not only captures the temporal dependence and local spatial characteristics of vehicle speed data, but also conducts in-depth analysis of vehicle speed sequences from different perspectives. This combined model can better capture the complexity and diversity of vehicle speed data, thereby providing more accurate vehicle speed prediction results.

[0053] 2. This paper uses the ISSA algorithm to optimize the L2 regularization coefficient, number of neurons, and learning rate hyperparameters of the CNN-BiLSTM model. The ISSA algorithm introduces sine, cosine, and Gaussian chaos strategies to effectively search for the optimal solution, avoid local optimal solutions, and find the globally optimal hyperparameter combination, significantly improving the model's fitting and generalization capabilities and the accuracy of vehicle speed prediction.

[0054] 3. In the field of predictive energy management strategies, the present invention uses the ISSA algorithm-optimized CNN-BiLSTM neural network speed prediction model to accurately predict historical vehicle speeds, thereby improving the fuel economy of the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The above and other features, advantages and aspects of the embodiments of the present invention will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings, in which:

[0056] Figure 1 A flow chart of a method for vehicle speed prediction using a CNN-BiLSTM neural network based on ISSA according to an embodiment of the present invention is shown;

[0057] Figure 2 A flow chart of a vehicle speed prediction method according to an embodiment of the present invention is shown;

[0058] Figure 3A BiLSTM structure diagram according to an embodiment of the present invention is shown;

[0059] Figure 4 A diagram showing the working principle of CNN-BiLSTM according to an embodiment of the present invention is shown;

[0060] Figure 5 A diagram showing a vehicle speed prediction effect of a vehicle speed prediction model according to an embodiment of the present invention is shown;

[0061] Figure 6 A diagram showing a vehicle speed prediction root mean square error of a vehicle speed prediction model according to an embodiment of the present invention is shown;

[0062] Figure 7 A block diagram of a device for predicting vehicle speed using a CNN-BiLSTM neural network based on ISSA according to an embodiment of the present invention is shown;

[0063] Figure 8 A schematic diagram of a device for vehicle speed prediction using a CNN-BiLSTM neural network based on ISSA according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0064] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0065] According to an embodiment of the present invention, a method, device and equipment for vehicle speed prediction using a CNN-BiLSTM neural network based on ISSA are proposed. By combining a convolutional neural network and a bidirectional long short-term memory network, not only can the temporal dependency and local spatial features in vehicle speed data be captured, but also in-depth analysis of vehicle speed sequences can be performed from different angles. The ISSA algorithm is used to optimize the L2 regularization coefficient, number of neurons and learning rate hyperparameters of the CNN-BiLSTM model, significantly improving the model's fitting and generalization capabilities.

[0066] The principles and spirit of the present invention are explained in detail below with reference to several representative embodiments of the present invention.

[0067] Figure 1 1 is a flow chart of a method for predicting vehicle speed using a CNN-BiLSTM neural network based on ISSA according to an embodiment of the present invention. The method includes:

[0068] Step S01: collecting a vehicle operation dataset for ISSA-based CNN-BiLSTM neural network speed prediction, extracting original time-speed data using a sliding time window method to form a training sample dataset;

[0069] Step S02: using Circle chaos mapping, sine and cosine and Gaussian chaos strategies to improve the SSA algorithm to obtain the ISSA algorithm;

[0070] Step S03: constructing a CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, using the ISSA algorithm to optimize the parameters of the CNN-BiLSTM neural network vehicle speed prediction model, and using the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA to learn and train the training sample data set;

[0071] Step S04: Use Matlab software to use the historical vehicle speed of n time steps after the sliding time window operation as the input of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and the vehicle speed of the next m time steps as the output of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA to perform vehicle speed prediction.

[0072] It should be noted that although the operations of the method of the present invention are described in a specific order in the above embodiments and drawings, this does not require or imply that these operations must be performed in this specific order, or that all illustrated operations must be performed to achieve the desired results. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0073] In order to more clearly explain the above-mentioned ISSA-based CNN-BiLSTM neural network vehicle speed prediction method, a specific embodiment is used for illustration below. However, it is worth noting that this embodiment is only for better illustrating the present invention and does not constitute an improper limitation to the present invention.

[0074] The following example further illustrates the ISSA-based CNN-BiLSTM neural network speed prediction method:

[0075] like Figure 2 FIG. 1 is a flow chart of a vehicle speed prediction method according to an embodiment of the present invention, which is specifically as follows:

[0076] Step S01: Collect a vehicle operation dataset for ISSA-based CNN-BiLSTM neural network speed prediction, and use the sliding time window method to extract the original time-speed data to form a training sample dataset.

[0077] In this embodiment, the vehicle speed data is directly obtained from an existing database, wherein the training set data includes data under WVUCITY, UDDS, WVUINTER and HWFET standard driving conditions, and the test set data is data under the China light-duty vehicle test cycle-passenger (CLTC_P) standard conditions.

[0078] Specifically, the sliding time window method for extracting the original time-vehicle speed data includes the following settings: setting the historical input window length to 10, that is, each input sample contains the vehicle speed values ​​at the current and previous 9 sampling moments, setting the output window length to 5, that is, predicting the vehicle speed values ​​for 5 consecutive moments in the future, and the sample construction step size to 1 to generate a training sample data set.

[0079] Step S02: using Circle chaos mapping, sine and cosine and Gaussian chaos strategies to improve the Sparrow Search Algorithm (SSA) algorithm to obtain the Improved Sparrow Search Algorithm (ISSA) algorithm.

[0080] ISSA algorithm: The ISSA algorithm introduces Circle chaos mapping, sine and cosine, and Gaussian chaos strategies to the SSA algorithm, effectively searching for the optimal solution. This can avoid the problem of local optimal solutions and find the globally optimal hyperparameter combination, thereby improving the model's fitting and generalization capabilities.

[0081] The ISSA algorithm of the present invention comprises the following steps:

[0082] Step S021: Initialize parameters: sparrow population size, maximum number of iterations, dimension of each sparrow position, individual discoverer ratio, and warning threshold;

[0083] Step S022: Evaluate the fitness value;

[0084] Step S023: Initialize the Circle chaotic map population, and its expression is:

[0085]

[0086] Among them, x n+1 is the chaotic variable of the next iteration, x n is the chaotic variable of the current iteration, 3.85 is the control parameter of the chaotic mapping, which is used to adjust the chaotic behavior, sin(3.85πx n ) is the nonlinear term of the sine function, which increases the complexity of the system. Mod(·,1) is the modulo operation, which ensures that the variable is always in the interval [0,1].

[0087] Step S024: Leader position update: Leaders are responsible for exploring the global optimal solution. Their position update method is crucial to the algorithm's global search capability. To enhance the search capability and prevent falling into local optimality, the Sine-Cosine Strategy can be introduced into the leader position update formula. The improved position update formula is as follows:

[0088]

[0089] in, Update the position of the jth dimension for the i-th sparrow individual in the t+1th iteration, Update the position of the i-th sparrow in the j-th dimension in the t-th iteration, X best is the optimal individual in the current population, r2 is a random number in [0, 2π], which determines the moving distance of the sparrow; r3 is a random number in [0, 2π], which controls the influence of the optimal individual on the next position of the sparrow; ST is a constant in [0.5, 1], which represents the safety value; R2 is a random number in [0, 1], which represents the warning value; α is the cosine decreasing weight factor, which is expressed as follows:

[0090]

[0091] Among them, α max is the initial weight value, α min is the final weight value, t is the current number of iterations, T is the maximum number of iterations, and cos is the cosine function.

[0092] Step S025: Follower Position Update: During the foraging process, followers typically forage around the optimal finder and may compete with the finder for food, thereby becoming a new finder. To prevent the algorithm from falling into a local optimum, the follower position update is improved by introducing a Gaussian perturbation strategy to enhance global search capabilities. The improved follower position update formula is as follows:

[0093]

[0094] in, is the updated position of the jth dimension for the i-th sparrow individual in the t+1th iteration, Update the position of the i-th sparrow in the j-th dimension in the t-th iteration, X best is the optimal individual in the current population, r1 is a random factor, which is a uniformly distributed random number between [0,1], σ is the Gaussian perturbation coefficient, which is set to [0.01,0.1], and γ(0,1) is the standard normal distribution noise with a mean of 0 and a standard deviation of 1;

[0095] Step S026: Update the position of the sentinel.

[0096] Step S03: constructing a CNN-BiLSTM neural network vehicle speed prediction model, using the ISSA algorithm to optimize the parameters of the CNN-BiLSTM neural network vehicle speed prediction model, and using the CNN-BiLSTM neural network vehicle speed prediction model to learn and train the training sample data set;

[0097] The Bidirectional Long Short Term Memory (BiLSTM) network combines a forward LSTM network and a backward LSTM network, such as Figure 3 As shown in the figure, the forward LSYM network is used to capture forward information, and the backward LSTM is used to capture backward information, thereby enhancing the model's ability to model complex time data. The calculation process is as follows:

[0098]

[0099] in, is the hidden state of the forward LSTM at time t, LSTM forward is the forward LSTM calculation process, x t To input historical vehicle speed data, is the hidden state of the forward LSTM at time t-1, is the hidden state of the backward LSTM at time t, LSTM backward For the backward LSTM calculation process, is the hidden state of the backward LSTM at time t+1, W T is the forward propagation weight, W V is the back propagation weight, b is the bias term, W x is the residual weight.

[0100] The CNN-BiLST network speed prediction process is as follows Figure 4As shown. The CNN-BiLSTM network is achieved by integrating the advantages of the convolutional neural network (CNN) and the bidirectional long short-term memory network (BiLSTM) structure. The model first extracts local features from the input historical speed sequence through a one-dimensional convolution layer. The convolution layer usually uses multiple convolution kernels of different sizes to slide and extract features of different scales in the sequence. It then performs downsampling through a maximum pooling layer to reduce the feature dimension while retaining key feature information. After obtaining the local time series features, these features are passed into the BiLSTM structure. BiLSTM introduces two LSTM units, forward and reverse, to process the forward and backward dependency information of the time series respectively. Finally, the high-dimensional time series features output by the BiLSTM layer are passed to the fully connected layer and mapped to the speed prediction value for the next 5 time steps. Through end-to-end training, the network learns to automatically extract deep features that can characterize the future speed change trend from the historical 10-second speed sequence, thereby achieving high-precision speed prediction for the next few seconds.

[0101] The steps for training the training sample data set using the ISSA-based CNN-BiLSTM neural network speed prediction model are as follows:

[0102] Step S031: importing the training sample data set described in step S01, dividing it into a training set and a test set, and performing normalization processing;

[0103] Step S032: setting the upper and lower bounds of the initialization parameters of the ISSA algorithm, including the size of the sparrow population, the maximum number of iterations, the amount of optimization parameters, and the optimization parameters;

[0104] Step S033: using the ISSA algorithm to optimize the L2 regularization expression parameters, the number of neurons, and the learning rate parameters of the CNN-BiLSTM neural network vehicle speed prediction model to obtain the ISSA-based CNN-BiLSTM neural network vehicle speed prediction model;

[0105] Step S034: Train the CNN-BiLSTM neural network vehicle speed prediction model. When the loss function value reaches the set value or no longer decreases significantly, stop training.

[0106] Step S04: Use Matlab software to use the historical vehicle speed of n time steps after the sliding time window operation as the input of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and the vehicle speed of the next m time steps as the output of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA to perform vehicle speed prediction.

[0107] In this example, the original time-speed sequence test data is imported into Matlab. The speed sample data for the current and previous nine sampling moments extracted using a sliding time window method are used as input parameters for the trained ISSA-based CNN-BiLSTM model. The speed values ​​for the next five consecutive moments are used as the output results of the trained ISSA-based CNN-BiLSTM speed prediction model to predict the speed of the entire test sample. The speed prediction expression is as follows:

[0108] [v t+1 ,v t+2 ,v t+3 ,…,v t+p ]=f CNN-BiLSTM (v t ,v t-1 ,v t-2 ,…v t-h+1 )

[0109] Among them, t is the current time, v t+1 ,v t+2 ,v t+3 ,…,v t+p is the predicted output speed of the speed prediction model, v t ,v t-1 ,v t-2 ,…v t-h+1 is the input speed of the speed prediction model, v t+1 ,v t+2 ,v t+3 ,…,v t+p is the length of the output speed sequence, v t ,v t-1 ,v t-2 ,…v t-h+1 is the length of the input vehicle speed sequence, f CNN-BiLSTM It is a CNN-BiLSTM neural network vehicle speed prediction model based on ISSA.

[0110] To evaluate the accuracy of the ISSA-based CNN-BiLSTM neural network speed prediction model, the root mean square error (RMSE) is used as an indicator to reflect the deviation of single-step prediction. The average root mean square error (ARMSE) is used to measure the overall prediction effect of the ISSA-based CNN-BiLSTM neural network speed prediction model. The expression is:

[0111]

[0112] Among them, RMSE is the root mean square error, ARMSE is the average root mean square error, p is the prediction time, T is, v j To predict the actual vehicle speed in the time domain, is the predicted vehicle speed in the prediction time domain, and T is the total number of vehicle speed sample sequences.

[0113] To validate the effectiveness of this invention, the China Light-Duty Vehicle Test Cycle-Passenger (CLTC_P) driving cycle, which covers urban, suburban, and highway driving conditions, was selected as the test driving cycle. The optimized CNN-BiLSTM neural network speed prediction model was used to verify its accuracy in predicting the vehicle's future short-term speed. Key information from the CLTC_P driving cycle in this example includes a total driving cycle duration of 1800 seconds, a cumulative driving distance of 14.48 km, and a maximum speed of 31.6 m / s (114 km / h).

[0114] According to the simulation results, it can be seen that the vehicle speed prediction method based on the combined deep learning model CNN-BiLSTM based on the improved SSA proposed in this invention has the following advantages:

[0115] The comparison results of vehicle speed prediction results are shown in the table below:

[0116] Table 1 Vehicle speed prediction results of the benchmark method and the method of the present invention

[0117] Prediction Methods ARMSE / [Km / h] ARMSE improvement ratio [%] CNN-BiLSTM 3.5321 - CNN-BiLSTM based on ISSA (this invention) 2.7233 29.70

[0118] The data in Table 1 demonstrates that, under given test conditions, the ISSA-based CNN-BiLSTM neural network speed prediction method proposed in this invention reduces the ARMSE of speed prediction by 29.70% across the entire test condition compared to the unoptimized CNN-BiLSTM neural network speed prediction model. Therefore, the prediction model of this invention far surpasses the unoptimized CNN-BiLSTM neural network speed prediction model in accuracy, enabling more accurate speed prediction and thus effectively improving vehicle fuel economy.

[0119] according to Figure 5 It can be seen that, overall, the prediction trends of both methods can track the actual vehicle speed well, but the ISSA-based CNN-BiLSTM method can more accurately fit the speed changes at certain moments of drastic fluctuations, reducing prediction errors and improving the robustness of the model. From the locally enlarged image, the prediction curve of the ISSA-based CNN-BiLSTM vehicle speed prediction model (below) is closer to the actual vehicle speed than the CNN-BiLSTM vehicle speed prediction model (above), especially in the final high-speed stage. The ISSA-based CNN-BiLSTM vehicle speed prediction model is significantly more accurate than the unoptimized CNN-BiLSTM vehicle speed prediction model, indicating that the ISSA-based CNN-BiLSTM vehicle speed prediction model has improved the prediction accuracy of the model by optimizing parameters. Figure 6 As shown in the figure, there is a comparison between the ISSA-based CNN-BiLSTM vehicle speed prediction model of the present invention and the CNN-BiLSTM vehicle speed prediction model. The ISSA-optimized model shows lower prediction errors at most sample points, especially in the error surge interval, showing better suppression ability, indicating that the proposed model has higher prediction accuracy and stronger robustness.

[0120] Based on the same inventive concept, the present invention also proposes a device for predicting vehicle speed based on ISSA CNN-BiLSTM neural network. The implementation of the device can refer to the implementation of the above method, and the repeated parts will not be repeated. Figure 7 As shown, the device 100 includes:

[0121] Data acquisition module 101: used to collect vehicle operation data sets for ISSA-based CNN-BiLSTM neural network speed prediction, and extract original time-speed data using a sliding time window method to form a training sample data set;

[0122] ISSA algorithm module 102: used to improve the SSA algorithm by using Circle chaos mapping, sine-cosine and Gaussian chaos strategies to obtain the ISSA algorithm;

[0123] Model construction module 103: used to construct a CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and use the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA to learn and train the training sample data set;

[0124] Vehicle speed prediction module 104: Used to use Matlab software to use the historical vehicle speed of n time steps after the sliding time window operation as the input of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and the vehicle speed of m time steps in the future as the output of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA to perform vehicle speed prediction.

[0125] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the described module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0126] like Figure 8As shown, the device includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) or loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for the operation of the device can also be stored. The CPU, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0127] Many components in a device are connected to the I / O interface, including: input units, such as a keyboard and mouse; output units, such as various types of displays and speakers; storage units, such as magnetic disks and optical disks; and communication units, such as network cards, modems, and wireless communication transceivers. The communication unit allows the device to exchange information / data with other devices via computer networks such as the Internet and / or various telecommunication networks.

[0128] The processing unit performs the various methods and processes described above, such as method steps S01 to S04. For example, in some embodiments, method steps S01 to S04 can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program can be loaded and / or installed on the device via a ROM and / or a communication unit. When the computer program is loaded into the RAM and executed by the CPU, one or more of the method steps S01 to S04 described above can be executed. Alternatively, in other embodiments, the CPU can be configured to execute method steps S01 to S04 in any other appropriate manner (for example, by means of firmware).

[0129] The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, and without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), and the like.

[0130] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. Such program code can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0131] In the context of the present invention, machine-readable medium can be a tangible medium that can contain or store a program for use with an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. Machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. More specific examples of machine-readable storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0132] In addition, although adopting specific order to describe each operation, this should be understood as requiring such operation to be carried out in the specific order shown or in sequential order, or requiring all illustrated operations to be carried out to obtain desired result.Under certain environment, multitasking and parallel processing may be advantageous.Similarly, although comprising some specific implementation details in the above discussion, these should not be construed as limiting the scope of the present invention.Some features described in the context of independent embodiment can also be realized in single realization in combination.On the contrary, the various features described in the context of independent realization also can be realized in multiple realizations individually or in the mode of any suitable subcombination.

[0133] Although the subject matter has been described in language specific to structural features and / or methodological logical acts, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are merely example forms of implementing the claims.

Claims

1. A method for vehicle speed prediction based on ISSA CNN-BiLSTM neural network, characterized in that: The method includes: Step S01: collecting a vehicle operation dataset for ISSA-based CNN-BiLSTM neural network speed prediction, extracting original time-speed data using a sliding time window method to form a training sample dataset; Step S02: using Circle chaos mapping, sine and cosine and Gaussian chaos strategies to improve the SSA algorithm to obtain the ISSA algorithm; Step S03: constructing a CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and using the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA to train the training sample data set; Step S04: Use Matlab software to use the historical vehicle speed of n time steps after the sliding time window operation as the input of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and the vehicle speed of the next m time steps as the output of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA to perform vehicle speed prediction.

2. The method for vehicle speed prediction based on ISSA CNN-BiLSTM neural network according to claim 1 is characterized in that: The steps of the ISSA algorithm described in step S02 are: Step S021: Initialize parameters: sparrow population size, maximum number of iterations, dimension of each sparrow position, individual discoverer ratio, and warning threshold; Step S022: Evaluate the fitness value; Step S023: Initialize the Circle chaotic map population, and its expression is: Among them, x n+1 is the chaotic variable of the next iteration, x n is the chaotic variable of the current iteration, 3.85 is the control parameter of the chaotic mapping, which is used to adjust the chaotic behavior, sin(3.85πx n ) is the nonlinear term of the sine function, which increases the complexity of the system. Mod(·,1) is the modulo operation, which ensures that the variable is always in the interval [0,1]. Step S024: Update the discoverer's location: The sine and cosine strategies are introduced into the discoverer's location update formula. The improved discoverer's location update formula is: in, Update the position of the jth dimension for the i-th sparrow individual in the t+1th iteration, Update the position of the i-th sparrow in the j-th dimension in the t-th iteration, X best is the optimal individual in the current population, r2 is a random number in [0, 2π], which determines the moving distance of the sparrow; r3 is a random number in [0, 2π], which controls the influence of the optimal individual on the next position of the sparrow; ST is a constant in [0.5, 1], which represents the safety value; R2 is a random number in [0, 1], which represents the warning value; α is the cosine decreasing weight factor; Step S025: Follower position update: The Gaussian perturbation strategy is introduced into the follower position update formula. The improved follower position update formula is: in, is the updated position of the jth dimension for the i-th sparrow individual in the t+1th iteration, Update the position of the i-th sparrow in the j-th dimension in the t-th iteration, X best is the optimal individual in the current population, r1 is a random factor, which is a uniformly distributed random number between [0,1], σ is the Gaussian perturbation coefficient, which is set to [0.01,0.1], and γ(0,1) is the standard normal distribution noise with a mean of 0 and a standard deviation of 1; Step S026: Update the position of the sentinel.

3. The method for vehicle speed prediction based on ISSA CNN-BiLSTM neural network according to claim 2 is characterized in that: The expression of the cosine decreasing weight factor α in step S024 is: Among them, α max is the initial weight value, α min is the final weight value, t is the current number of iterations, T is the maximum number of iterations, and cos is the cosine function.

4. The method for vehicle speed prediction based on ISSA CNN-BiLSTM neural network according to claim 1, characterized in that: The process of the ISSA-based CNN-BiLSTM neural network vehicle speed prediction model described in step S03 is as follows: The local features of the input historical vehicle speed sequence are extracted through a one-dimensional convolutional layer; Downsampling is performed through the maximum pooling layer to obtain local temporal features, reducing the feature dimension while retaining key feature information; The local time series features are passed into the BiLSTM structure. BiLSTM introduces two LSTM units, forward and reverse, to process the forward and backward dependency information of the time series respectively. The BiLSTM layer outputs high-dimensional time series features, which are passed to the fully connected layer and mapped into the predicted vehicle speed value for the next m time steps.

5. The method for vehicle speed prediction based on ISSA CNN-BiLSTM neural network according to claim 1, characterized in that: The steps of using the ISSA-based CNN-BiLSTM neural network vehicle speed prediction model to learn and train the training sample data set in step S03 are as follows: Step S031: importing the training sample data set described in step S01, dividing it into a training set and a test set, and performing normalization processing; Step S032: setting the upper and lower bounds of the initialization parameters of the ISSA algorithm, including the size of the sparrow population, the maximum number of iterations, the amount of optimization parameters, and the optimization parameters; Step S033: using the ISSA algorithm to optimize the L2 regularization expression parameters, the number of neurons, and the learning rate parameters of the CNN-BiLSTM neural network vehicle speed prediction model to obtain the ISSA-based CNN-BiLSTM neural network vehicle speed prediction model; Step S034: Train the ISSA-based CNN-BiLSTM neural network vehicle speed prediction model. When the loss function value reaches a set value or no longer decreases significantly, stop training.

6. The method for vehicle speed prediction based on ISSA CNN-BiLSTM neural network according to claim 1, characterized in that: The vehicle speed prediction expression in step S04 is: [v t+1 ,v t+2 ,v t+3 ,…,v t+p ]=f CNN-BiLSTM (v t ,v t-1 ,v t-2 ,…v t-h+1 ) Among them, t is the current time, v t+1 ,v t+2 ,v t+3 ,…,v t+p is the predicted output speed of the speed prediction model, v t ,v t-1 ,v t-2 ,…v t-h+1 is the input speed of the speed prediction model, v t+1 ,v t+2 ,v t+3 ,…,v t+p is the length of the output speed sequence, v t ,v t-1 ,v t-2 ,…v t-h+1 is the length of the input vehicle speed sequence, f CNN-BiLSTM This is the optimized CNN-BiLSTM neural network vehicle speed prediction model.

7. The method for vehicle speed prediction based on ISSA CNN-BiLSTM neural network according to claim 1, characterized in that: The root mean square error is used as an indicator to evaluate the accuracy of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and the average root mean square error is used to measure the overall prediction effect of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA. The expression is: Among them, RMSE is the root mean square error, ARMSE is the average root mean square error, p is the prediction time, T is, v j To predict the actual vehicle speed in the time domain, is the predicted vehicle speed in the prediction time domain, and T is the total number of vehicle speed sample sequences.

8. A device for predicting vehicle speed based on ISSA CNN-BiLSTM neural network, characterized in that: The device implements the method according to any one of claims 1 to 7, comprising: Data acquisition module: used to collect vehicle operation data sets for ISSA-based CNN-BiLSTM neural network speed prediction, and use the sliding time window method to extract the original time-speed data to form a training sample data set; ISSA algorithm module: used to improve the SSA algorithm using Circle chaos mapping, sine-cosine and Gaussian chaos strategies to obtain the ISSA algorithm; Model building module: used to build a CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and use the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA to learn and train the training sample data set; Vehicle speed prediction module: This module uses Matlab software to use the historical vehicle speed of n time steps after the sliding time window operation as the input of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA, and the vehicle speed of m time steps in the future as the output of the CNN-BiLSTM neural network vehicle speed prediction model based on ISSA to perform vehicle speed prediction.

9. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images