•A Markov-chain-based approach for modeling the system states of the EBSS containing e-bike stations and battery cabinet stations.•Integrated battery swapping and rebalancing strategies considering both e-bike number and battery power level.•A fast and adaptive one-step Markovian strategy and a globally optimal rolling-horizon Markovian strategy.•Modeling of virtual station for realizing optimized allocation of limited resources. A new generation of the e-bike sharing system (EBSS) is emerging, where the e-bikes are dockless but need to be parked in designated zones defined by electric fences. The operation of the EBSS relies on the efficient swapping of batteries, in addition to e-bike rebalancing. The replaced power-deficient batteries can be charged in a central depot or street-side cabinets. This paper proposes an approach for modeling the system states of the EBSS containing e-bike stations and battery cabinet stations based on Markov chain dynamics considering both e-bike number and battery power level. Utilizing the prediction of e-bike inventory levels at e-bike stations and batteries’ status at e-bike stations/battery cabinets, a fast and adaptive one-step Markovian strategy is introduced, informing the operating staff of the next station to visit and detailed operations for battery swapping at e-bike/cabinet stations and rebalancing at e-bike stations. Besides, a rolling-horizon Markovian strategy is proposed to make a globally optimal plan of battery swapping and e-bike rebalancing by solving an integer nonlinear programming problem, which gives an approximate upper bound for integrated operations. By considering future demand and making optimal decisions, the proposed Markovian strategies can achieve a profit improvement of over 20% on average, compared to the current strategy used by the industry, as illustrated by the numerical simulations on a real-world EBSS.