Forecasting of energy consumption rate and battery stress under real-world traffic conditions using ANN model with different learning algorithms

Estimating energy consumption rates is a necessary step when building infrastructure for charging and schedule optimization of battery-powered vehicles utilized in public urban driving patterns. This study examined several input factors for the prediction of vehicle performance. Input conditions were energy management controls, State of Charge (SOC) power train batteries, and ultra-capacitor vehicle models; output metrics included consumption rates, battery loads, and trip distances. To examine the experimental design, an L9 design was used with four control factors at three different levels each. Artificial neural network (ANN) models were developed employing four learning algorithms: quick propagation (QuP), batch backpropagation (BBaP), Levenberg-Marquardt backpropagation (LMBaP), and incremental backpropagation (IBaP). Post-simulation results were summarized and validated using the root mean square error (RMSE), which indicated that the values collected experimentally were close to those predicted by the models. This paper built an ANN-based prediction model and accurately predicted vehicle performance and potential energy shortfalls in public transportation networks. These insights can be applied to interventions like charging stations or reshaping bus timings to avoid power loss.