Bike sharing systems: The impact of precise trip demand forecasting on operational efficiency in different city structures

Increasing environmental concerns drive interest in sustainable solutions across various fields. Vehicle sharing systems such as one-way station-based bike sharing systems (BSSs) offer one such solution in transportation, although they pose operational challenges like vehicle imbalance. While many studies focus on optimizing rebalancing operations using trip demand forecasting, the added value of precise trip demand forecasting remains unexplored. This study assesses the added value of collecting detailed trip demand data and developing trip demand forecasting models. To achieve this, we create a simulation–optimization framework representing a city BSS in operation during the day. We use a discrete-event simulator representing the system dynamics and an enhanced mathematical model optimizing the rebalancing operations. We employ clustering in the optimization module to manage larger case studies, dividing the problem into smaller sub-problems. Our computational experiments compare two main scenarios, perfect demand forecast and unknown future demand, as well as several intermediate scenarios where partial future trip demand information is available. These scenarios allow us to determine the trade-off between lost trip demand and rebalancing operations costs, assess demand forecasting benefits, and identify the budget’s upper limit for precise trip demand forecasting. Subsequently, we conduct experiments on one synthetic (35 stations) and four real-life case studies, ranging from small systems (21 and 298 stations) to large systems (681 and 1361 stations). Results reveal that precise trip demand forecasting has varying impacts on small and large BSSs, with larger BSSs benefiting the most without significantly increasing the rebalancing operations costs. We also observe that the most significant improvements occur between 0% and 40% trip demand knowledge while beyond 60%, the increase in returns diminishes. The findings of this study offer valuable insights for operators in enhancing service levels and optimizing resource allocation. • Unknown, partial, and perfect knowledge scenarios assess demand forecasting impact. • Data from Sarajevo, Berlin, Chicago, and NYC validate the framework’s applicability. • The larger the bike sharing system, the more it benefits from demand forecasting. • Simple forecasting methods already reduce lost trips in large bike sharing systems. • Forecasting cuts lost trips in large systems without increasing rebalancing costs.