Short-time Energy Optimization Control Method For Connected Commercial Vehicles Based On Desired Speed Prediction

With the vigorous development of Internet economy and e-commerce,the research on energy-saving technology of commercial logistics vehicles becomes more and more important;the further development of intelligently connected vehicles has provided convenient conditions for breakthroughs in vehicle energy saving technology.This paper studies the energy-saving technology of commercial vehicles with network connection conditions.Based on the vehicle-cloud platform,this study comprehensively analyzes the historical data of vehicles,fully evaluates the personal behavior characteristics of drivers,predicts their behavior selection according to their driving style,and finally establishes an energy-saving control model to improve the economy of vehicles.In this paper,the research focus on the online logistics vehicles operated by drivers.Based on the control system of vehicle-cloud interaction,the torque of the engine output in the future short time domain is optimized and adjusted,and the effect of fuel saving is finally achieved.The framework of this study can be divided into two parts: data processing and prediction module based on machine learning and energy saving module based on model predictive control(MPC).According to different research conditions and targets,the data processing uses unsupervised learning method,that is,the driver’s driving style is classified and identified by clustering the unlabeled driving data;the prediction task uses the supervised deep learning model based on the recurrent neural network to accurately estimate the desired speed of the driver.The energy-saving control model mainly includes the establishment of vehicle dynamics and vehicle fuel consumption models and the construction of optimization problems.More specifically,the main research content of this paper includes three parts: classification of driving style,prediction of driver’s desired speed and establishment of vehicle energy-saving controller based on model predictive control:1.The driving style database is established based on the natural driving data on the real road.The whole driving cycle is divided into different typical driving scenarios according to the speed and acceleration of the vehicle,and the main typical driving scenarios corresponding to different driving conditions are determined.This study selects the classification characteristics based on typical driving scenarios,and realizes the identification of drivers’ style under different driving conditions based on unsupervised learning model to verify the consistency of drivers’ style in the whole driving process.Finally,the type of driver’s style in the whole driving cycle is obtained.2.According to the characteristics of the vehicle speed sequence prediction task,the long and short term memory neural network(LSTM)is selected as the basis to establish the deep learning model.By comparing the variation of prediction accuracy under different input information,the optimal model input type is selected.After the completion of model construction and debugging,the influence of driving style information on model prediction is further studied,and it was proved that driving style information is helpful to improve the ability of predicting the driver’s desired speed.3.After the driver’s desired speed has been obtained,this study established a multiobjective optimization problem,including meeting the driver’s power command,improving vehicle’s economy,comfort and other indicators.The simplified vehicle model and the mathematical model of fuel consumption are also established in this study,and the optimization problem is solved under the framework of model predictive control.Finally,a closed-loop simulation model is built to verify the energy-saving effect of the controller,which proves that the introduction of energy-saving control method is effective in improving the economy of the vehicle.