| At present,road transportation is the industry with the highest oil consumption in China and the world,and it also puts more pressure on the growing energy supply.As the leader in the driving process of the vehicle,the driver’s poor driving style is the main reason for the increase in fuel consumption of the vehicle.Therefore,it is of great significance to study the changes in vehicle fuel consumption based on the driving style during the driving process.In this paper,a clustering algorithm and neural network classification method are used to mine and analyze the driving data,and an evaluation model of driving style is established.Based on the clustering results of driving style,a fuel consumption prediction model during vehicle driving is constructed.The main work of this article is as follows:(1)According to the characteristics of truck driving data,the data is preprocessed and feature extraction of abnormal driving behavior is performed.Through the pre-processing work of filtering and duplicate value deletion,the data with abnormal frequency,repeated records,and abnormal latitude and longitude ranges are filtered and cleaned,and the behavioral characteristics of abnormal driving behaviors are extracted based on the filtered data;then the principal component analysis(PCA)The statistics of speed,acceleration and driving behavior parameters are used to reduce dimensionality.Finally,19 parameters are reduced to 7 main components,which effectively reduces the time complexity of clustering data dimension and data processing.(2)A driving style evaluation model based on clustering and BP neural network is proposed.First,build a two-layer clustering model based on the BIRCH algorithm and k-means algorithm,and use the BIRCH clustering algorithm to perform the first layer of rough clustering on the seven principal component factors to create a clustering feature tree(CF-Tree);Then use the k-means algorithm to perform second-level fine clustering on the leaf nodes of the CFTree;based on the driving style clustering results,use the BP network to learn,and finally use the trained BP neural network classifier to classify the driving style Evaluate and use the confusion matrix to verify the effectiveness of the model.(3)A fuel consumption prediction model based on driving style is proposed.In this paper,two fuel consumption prediction models are proposed based on driving style data during fuel consumption analysis,including: one-way stacked LSTM model and two-way stacked LSTM model.These two models use driving data under different driving styles as input variables,fuel consumption As an output variable;Finally,the effectiveness of the model is verified by comparing the mean absolute error MAE and root mean square error RMSE of the algorithm in this paper with RNN and GRU.The driving style evaluation model and fuel consumption prediction model proposed in this paper are tested by GPS data of real trucks.The research results show that the driving style evaluation model eliminates the subjective factors of people and achieves accurate and efficient evaluation of driving style.The model accuracy rate is as high as 96.75%,the simulation experiment on the fuel consumption model found that the fuel consumption model proposed in this paper is more accurate than the prediction results of RNN and GRU,and the evaluation error is within.Finally,based on the evaluation results of the fuel consumption model,some eco-driving strategies that are easy to implement are proposed to achieve the purpose of reducing resource consumption and pollutant emissions. |
