To break the energy bottleneck of wireless sensor networks and considering the sensitivity of wireless charging efficiency to charging distance, a machine learning-based adaptive dual-mode device co-scheduling scheme is proposed. First, an adaptive threshold selection charging algorithm is proposed to take into account the node states based on residual energy, energy consumption and charging efficiency, and then an improved genetic algorithm is designed to select the appropriate charging mode for each node with the goal of maximizing energy use. In addition, to further reduce the time complexity of the charging algorithm, a support vector machine-based machine learning method is used to learn the above charging mode switching mechanism and construct an intelligent node state prediction model. Simulation results show that the proposed algorithm can effectively improve energy utility of multiple wireless charging devices and enhance the sustainability of the sensing network while ensuring low charging latency.