Mobile energy storage systems (MESS) are crucial technologies for alleviating the spatiotemporal mismatch between distributed renewable energy supply and demand

while promoting its local consumption. In systems characterized by multiple energy sources and loads

the efficient scheduling of MESS is challenged by the spatiotemporal coupling between mobile node selection and charge/discharge decisions

as well as multiple conflicting constraints. Traditional optimization approaches relying on precise mathematical models often fail to derive effective scheduling strategies under such conditions. To address this

this paper proposes a spatiotemporal scheduling strategy based on an adaptive reward function and multi-agent deep reinforcement learning. The problem is formulated as a multi-agent Markov decision process and cooperative multi-agent framework is used to decouple the high-dimensional problem with strong spatiotemporal coupling. Through agent-environment interaction

the system autonomously learns adaptive scheduling strategies in complex scenarios. Furthermore

this paper introduces an adaptive weight adjustment mechanism to guide MESS in developing rational decision-making preferences under dynamic conditions

significantly improving learning efficiency and model convergence. Simulation results demonstrate that the proposed strategy enables MESS to achieve efficient scheduling in complex scenarios

demonstrating strong stability and robustness

and providing theoretical support and practical reference for the intelligent scheduling of MESS in distributed energy systems.