Recent years have witnessed the growing popularity of e-bike sharing systems (EBSS) due to their enhanced riding comfort and suitability for longer rides, compared to conventional bike sharing systems (BSS). However, similar to BSS, EBSS also suffer from the spatiotemporal supply-demand imbalance. Additionally, the demand loss in EBSS could be caused by not only the shortage of e-bikes but also the insufficiency of battery power. Currently, operators address this imbalance by employing staff-driven trucks to relocate e-bikes and replace batteries, which is costly and environmentally unfriendly. This paper proposes a user-based relocation policy as a complement to the staff-based relocation. This policy incentivizes users to choose e-bikes with more systematically favored battery power levels under the premise of meeting the riding requirements. Consequently, a better distribution of e-bikes can be achieved, and e-bikes at various power levels in the system can be better utilized without additional efforts from users or operators. The proposed method consists of three parts, including (a) a Markov model for estimating relocation utilities, (b) an approximate state reduction method for simplifying high-dimensional calculations, and (c) an optimization model for determining incentives and e-bike recommendations. Numerical experiments on a real-world EBSS show that the proposed user-based relocation policy can achieve a nearly 15% profit increase when adopted alone or as a supplement to staff-based operations.