Charging station placement optimization based on the location significance prediction

One of the key challenges in charging infrastructure planning is ensuring optimal charging station placement. To address this issue, we introduce a novel approach to optimize the placement of electric vehicle charging stations, integrating a novel location-based charging station significance prediction model with a lion optimization algorithm (LOA) where the significance is defined as the combination of charging energy and the number of sessions. First, the data recorded on the existing charging stations are analyzed and preprocessed. Subsequently, we introduced the modified support vector regression (SVR) model for significance prediction and compared it with eight existing models showing its superiority over others. The SVR modification is related to the kernel function, where the standard Gaussian kernel is adapted to better suit location-based significance predictions. Following this, we utilize LOA to optimize the placement of additional charging stations in established charging infrastructure based on the prediction model trained on data congregated at the existing charging stations. The optimization is conducted for Zagreb and Split to evaluate the performance on small and large datasets. The results are assessed using significance prediction and pseudo-simulation. The new charging stations in Zagreb have a significance prediction of 10.53% greater than the calculated significance of existing charging stations. Furthermore, the significance prediction for new stations in Split is 0.42% greater than the calculated significance for existing stations. Pseudo-simulation proves that new charging stations have 38.15% greater significance than the existing stations in Zagreb and 31.24% in Split. Both methods confirm that infrastructure significance, in terms of charging energy and session count, is improved.