A semi-SPO (smart predict then optimize) method for electric bus scheduling optimization Available to Purchase

The state of charge (SoC) is very important for the operation of electric buses (EBs). At present, many EB scheduling plans fail to consider the arrangement of buses with different SoC in different operating periods, and different periods often have different passenger demands (PDs). In this work, three prediction–optimization methods to match PD at different periods with the most suitable EBs were established, aiming to reduce the number of buses that cannot operate due to insufficient SoC, thereby improving the robustness of EB systems. The first method is to predict the PD at each station by minimizing mean square error (MSE). The predicted PD is then substituted into an optimization model of the scheduling scheme. The second method directly predicts the operation benefits of each bus at different departure times in a way that minimizes MSE, and then substitutes the predicted benefit into the optimization model. The third method directly predicts the operation benefits of each bus at different departure times by minimizing the MSE and the prediction error of operating benefits between each two different buses. Then, the predicted benefits are substituted into the optimization model. Analysis showed that the bus scheduling scheme obtained using the proposed method can increase operational benefits by more than 39%.