Research On Prediction Method Of Idle Cycle

Idle cycle is a condition of the vehicle stops but not stopping the engine in the vehicle driving cycle,which mainly reflects the condition that the vehicle encounters red light or traffic jam during the driving process.The prediction and analysis of idle cycle are mainly used in the fields of engine idle start-stop system,path planning and intelligent driving.The existing idling start-stop technology only considers the driver’s intention and driving parameters,ignoring the actual idling time.This will cause the engine to stop for a short time and start frequently,which greatly reduces the fuel-saving and emission-reduction efficiency of the start-stop system.Aiming at the problem of obtaining the actual idling time,this paper proposes a prediction method for idling cycle.(1)Driving cycle construction.In order to make the prediction and analysis of idle cycle in this paper more in line with domestic road conditions,the vehicle driving data in a domestic city is selected to construct driving cycle.Firstly,the linear densification method and the boundary value method are used to preprocess the original data such as densification and smoothing.The processed data is divided into kinematic segments according to the kinematics division standard,and feature parameters are extracted according to the vehicle kinematics characteristics.Then,through the analysis of the principal component eigenvalues and the load matrix,the eigenparameters that can describe most of the kinematic information are retained,and at the same time,it plays the role of dimensionality reduction.Using the fuzzy C-means clustering method for the reserved characteristic parameters,the working conditions are divided into four categories: congested working conditions,relatively congested working conditions,relatively unobstructed working conditions and unobstructed working conditions.Finally,the comprehensive driving conditions are synthesized by the deviation optimization method,and compared with the conditions synthesized by the K-means and Gaussian mixture clustering methods.Through the comparison of the MATLAB synthesis condition curve,it can be seen that the parameter error and synthesis time of the synthesis condition using the method in this paper are reduced by 0.05 and 1.9s,respectively.(2)Research on the prediction method of idle cycle.Through the correlation analysis of the vehicle speed in the working conditions,it can be known that the idle cycle is a Markov process.Based on the four types of driving conditions obtained by the fuzzy C-means clustering method proposed in this paper,the prediction method of idle cycle is studied based on Markov and support vector machine theory.Since the vehicle driving data is a nonlinear discrete sequence,the working condition prediction is related to the explicit state quantities of the driving speed and the implicit state quantities such as the engine speed.In order to overcome the single state transition process of Markov chain,this paper establishes a hidden-Markov model.The hidden state and the observed state are taken together as input to predict the idle condition.Since the results of the hidden Markov model are unstable when the prediction step is long,this paper proposes the least squares support vector machine model to predict the idle speed condition.Particle swarm optimization is introduced to optimize the penalty factor and kernel parameters to avoid the influence of artificial subjectivity on the selection of prediction model parameters.Aiming at the problem of low prediction accuracy of a single model,this paper proposes a hybrid model of least squares support vector machine and Markov to predict idle speed conditions.By analyzing the residuals of the least squares support vector machine prediction results,the Markov property of the residual sequence is shown.Then use the Markov model to correct the residual sequence to obtain the optimized working condition prediction result.The simulation results show that the MSE values of the hybrid model are 3.1,2.06,0.7,and 0.11 lower than that of the single model under the four operating conditions,indicating that the overall prediction effect of the hybrid model on idle speed conditions is better than that of Markov or least squares support vector machine alone.(3)Verify the validity of the idle cycle prediction.MATLAB and VISSIM co-simulation method is used to verify the validity of idle cycle prediction.The composition of the idling start-stop system is briefly introduced,the influence of idling conditions on the start-stop strategy is analyzed,and the start-stop control strategy is optimized by using the hybrid model proposed in this paper to predict idling cycle.In the construction of the driving model,several factors that have a great influence on the simulation of the start-stop strategy,such as the safe following distance and the change of signal lights,are considered.Build Wiedemann following modules,four-phase signal light control and other modules respectively.The prediction results of idle speed in MATLAB are combined with the vehicle driving model in VISSIM through the Excel Link interface,and the vehicle driving model is used to simulate the start-stop control strategy before and after optimization.The experimental results show that the optimized strategy reduces idling times by 9.21% and idling stop time by 6.96% respectively under congested conditions,saves fuel consumption by 2.3%,and the pollutant emissions are basically the same;under relatively congested and relatively unobstructed conditions Compared with the strategy before optimization,the number of idle speeds of the optimized strategy decreased by 18.4% and 13.21%,the idle stop time was decreased by 6.83% and 4.01%,respectively,the fuel consumption was saved by 4.7%and 3.7%,and the pollutant emissions were reduced by 5.6% and 2.2%.Under relatively congested and relatively unobstructed conditions with good idling duration prediction effect,the optimized strategy has significantly improved fuel-saving and emission-reduction effect,while the smaller improvement in fuel-saving and emission-reduction effect under congested conditions is mainly affected by the accuracy of idling duration prediction.On the whole,the optimized start-stop strategy based on the prediction of idle speed can reduce short-term idle speed and improve the efficiency of fuel saving and emission reduction.The driving condition construction method and the idling cycle method studied in this paper provide effective input information for the optimization of the idling stop strategy,which is feasible and practical in reducing short-term idling and overcoming frequent starts.