Temperature Prediction Algorithms Using Machine Learning for Electric Vehicles

The development of precise and efficient temperature prediction algorithms to maintain ambient charging and operational environment of electric vehicles (EVs) has grown to be a critical concern for automakers due to steep rise in their demand. By utilizing machine learning (ML) based temperature prediction techniques, the energy wastage due to high temperature can be minimized as well as better driving power and enhanced passenger comfort may be obtained. For this work the dataset has been generated in real-time pseudo-environment using a temperature sensor-based hardware which can be effectively clipped-on the casing of the vehicle for monitoring and recording the temperature of both charging and normal running. Machine learning based temperature prediction using Polynomial Regression algorithm has been compared with Linear Regression and Time Series analysis (ARIMA method) based on their Mean Absolute Error (MAE), Coefficient of Determination (R2 Score), Root Mean Square Error (RMSE), for computational effectiveness. This methodology may also further be utilized in enhancing the air conditioning, cabin cooling of EVs. The advantage of this methodology is that it is platform independent and future values of critically high temperature may be predicted before it occurs so that corrective measures may be applied and the damages may be prevented.