Energy parameter modeling in plug-in hybrid electric vehicles using supervised machine learning approaches

Abstract

In this study, supervised machine learning techniques like random forest (RF), K-nearest neighbour (KNN), multiple linear regression (MLR), and artificial neural networks (ANNs) were used to predict energy parameters in plug-in hybrid electric vehicles. The primary objective of the study is to propose a machine learning model for estimating combined energy consumption for city and highway driving. For the prediction of combined, city, and highway fuel consumption, accuracy is found to range from 3.40 to 6.13, 0.029 to 0.0625, 0.030 to 0.091, and 0.022 to 0.038 for KNN, MLR, RF, and ANN, respectively, in terms of mean square error (MSE). The accuracy of the artificial neural network model is found to be higher than that of the random forest, K-nearest neighbour, and multiple linear regression for the estimations of the selected energy parameters. For the prediction of the combined fuel consumption, the accuracy of the artificial neural network is 183.51, 42.53, and 1.85 times higher than that of the K-nearest neighbour, random forest, and multiple linear regression, respectively. For the prediction of city fuel consumption (C), the accuracy of the ANN is 113.33, 19.75, and 2.06 times greater than that of the KNN, RF, and MLR, respectively. The predicted accuracy of highway fuel consumption (H) is 34.58, 16.62, and 1.17 times higher for ANN than that of the KNN, RF, and MLR, respectively. The accuracy of the artificial neural network is 1.046 times higher in terms of MSE when three hidden layers are used than when two hidden layers are used for the prediction of combined fuel consumption. The study suggests that supervised machine learning approaches like artificial neural networks and random forests can predict or model the energy parameters of plug-in hybrid electric vehicles with a significant level of accuracy.