Research On Driving Style Economy Prediction Model And Method Based On Big Data From The Internet Of Vehicles

The development of the national economy has been extensively promoted by the continuous growth of motor vehicle retention,but it also raises social and environmental issues.According to related statistics,the fuel consumption of heavy-duty commercial vehicles accounts for more than 50% of the total consumption of the whole society,and carbon emission accounts for more than 80% of motor vehicles.Many studies show that destructive driving behaviors will increase automobile fuel consumption to a certain extent.However,fuel consumption,as the most critical evaluation index for measuring the fuel economy of vehicles,is faced with the problem of inaccurate and inefficient current prediction models.Therefore,this paper focused on the hot issues of energy saving and emission reduction of heavy goods vehicles,researched short-term high-precision fuel prediction model,established the connection between driving style and fuel consumption,and provided data support for improving vehicle fuel efficiency,regulating driving behavior,and optimizing the control strategy of the automobile power system.It plays a positive role in effectively reducing vehicle fuel consumption,promoting energy conservation and emission reduction in the transportation industry,and boosting the national “carbon peaking and neutrality” strategy.In allusion to the above problems,this paper divided the massive data of vehicle networking,sampled typical trips,used K-means clustering algorithm,quantitatively analyzed the number of optimal clusters based on relevant evaluation indexes,and then modeled the driving style.At the same time,in order to overcome the existing problems of fuel consumption prediction models,this paper innovatively proposed a Shared-LSTM model and quantitatively proved the effectiveness of this model by comparing with multiple models.Secondly,the control variable method was used to constrain the load and driving style so as to predict the vehicle fuel economy.The main research work of this paper is as follows:(1)Modeling and analysis of driving styles based on vehicle networking big data.Based on the heavy-duty truck CAN bus and GPS big data,this study divided the vehicle trips and completed sample preprocessing as well as parameter calculation.The KMO test and Bartlett sphericity test were used simultaneously to examine whether there was a correlation among the original variables.Moreover,quantitative analysis of the structural validity of the samples was conducted.In order to reduce the complexity of computation and eliminate miscellaneous data,PCA method was used to reduce the dimension of data set.Finally,the driving styles were divided into three categories: aggressive,conservative,and conventional types according to silhouette coefficient,CH score,SSE,and other related parameters,laying the foundation for subsequent research.(2)Efficient prediction model for instantaneous fuel consumption of large-scale vehicles.In order to address the issues of low efficiency and poor accuracy of existing fuel consumption prediction methods,this paper innovatively proposed a Shared-LSTM model combining input gate and output gate based on the LSTM model.The efficiency and accuracy of the fuel consumption prediction of Shared-LSTM,LSTM,GRU,and BP neural network models were compared and analyzed under the same vehicle type,working conditions,multiple driving styles,and typical trips.As shown by the results,compared to LSTM,GRU,and BP neural networks,the prediction efficiency of the Shared-LSTM model was increased by more than 3%.Regarding prediction accuracy,its mean absolute error,mean squared error,and root mean squared error were all reduced by more than 6%.(3)Analysis of influencing factors of driving style and validation of economic forecasting methods.A typical trip sample(multiple styles and multiple loads)was sampled.An extensive sample data of 30 drivers in one month was read through the sliding window.The fuel consumption of different drivers with the same load and multiple styles and the same styles and multiple loads was focused under the same type,the same working condition,and other constraints.According to the final experimental results,the dead cargo weight and driving style combined affected the fuel economy of vehicles.Under no-load,half-load,and full-load conditions,the driving styles with the highest fuel economy of vehicles were respectively aggressive,conventional,and conservative,and the predicted results had a little deviation from the actual value,which has relatively strong engineering practical value.