Battery Maintenance of Pedelec Sharing System: Big Data Based Usage Prediction and Replenishment Scheduling

Pedelecs are an alternative of traditional share bikes by applying the battery-powered motor to assist pedaling and accordingly extend the riding coverage. The large scale deployment of pedelecs, however, requires a careful design of maintenance system to replace the batteries regularly that can be costly. This paper investigates the maintenance of a city-wide pedelec system by developing an offline solution in two steps. First, we develop an optimal and efficient hybrid prediction model which predicts the usage demand of pedelecs in every 48 h on a scale of millions of pedelecs. Our proposal predicts the future usage increment of pedelecs by combining a local predictor, a global predictor, and an inflection predictor, which captures both the short-term and long-term factors affecting the pedelec usage. Second, based on the developed predictor and results of big data analytics, an optimal path planning scheme for the replenishment of pedelec batteries is developed. As compared to other schemes, our scheme can save 40% of the maintenance cost. To verify our proposal, extensive real-data driven simulations are performed which show that the accuracy of the prediction process is high enough than each traditional method and our proposal solves the maintenance problem efficiently.