Data-driven control for orbit transition of piezoelectric energy harvesters under stochastic excitation

To address the tendency of bistable piezoelectric energy harvesting systems to remain trapped in low-energy intrawell oscillations and the difficulty of sustaining high-energy interwell motion under stochastic disturbances, this study proposes a data-driven event-triggered jump–drift hybrid control method. The proposed method takes the neighborhood of a high-energy reference orbit as the control target and constructs a complete electromechanical state representation using displacement, velocity, and piezoelectric voltage. A signed orbit error is further defined through local radial projection to characterize both the magnitude and direction of the state deviation, thereby enabling direction-selective event triggering and control release. During the active control phase, a locally corrective control input is generated using an offline-identified one-step prediction model. Once the high-energy state is established, the controller switches to the drift mode and intervenes again only when a pronounced degradation of the system state is detected. Numerical results demonstrate that the proposed method can drive the system from a low-energy state toward the high-energy orbit while achieving a favorable trade-off among orbit-transition speed, maintenance of high-energy motion, and control effort. The method also retains a certain degree of adaptability under stochastic disturbances and mismatches in load resistance and mechanical damping. Further analyses of the sampling period and control delay indicate that the proposed method exhibits reasonable tolerance to nonidealities associated with digital implementation.