Electric Truck Energy Consumption Prediction: A Comparison of Models based on Physics, Linear Regression, and Artificial Neural Networks
E-Force One AG is an electric truck company based in central Switzerland. Predicting the energy consumption on a given trip is one of the challenges they face. This report shows a method based on the sub-systems: route predictor, speed profile predictor, and truck energy model. Knowledge- and data- driven truck energy models were developed and compared by its performance. Physical models were found to have the best generalization. Linear regression has shown to be a good tool to extract physical parameters of a truck. Additionally, certain artificial neural network models show also a promising use case, especially when its structure is based on physics and its number of parameters is in balance to the available training data. The flexible number of parameters and the capability of neural networks to learn non-linear relationships gives it an edge over other models, in the way it is able to incorporate relationships that have not previously been thought of or are hard to predict.
Studienbetreuer: Thierry Prud’homme