This paper proposes a scheduling optimization strategy for multi-type mobile emergency resources based on a data-model hybrid-driven approach to enhance distribution network resilience. Firstly

a multi-type mobile emergency resource optimization scheduling model is developed to minimize the total loss cost of the coupled power transportation network

considering the impact of dynamic changes in traffic road conditions on mobile energy storage systems (MESS) and repair crews (RC) strategies. Secondly

a data-model hybrid-driven approach is proposed to solve the complicated nonlinear stochastic model in real time and precisely determine the optimal routing and scheduling strategies of MESS and RC. In the data-driven part

a novel graph attention network multi-agent reinforcement learning algorithm is proposed to solve the optimal routing strategy for MESS and RC

considering uncertain factors such as the road repair time of the traffic network and dynamic changes in the adjacency relationship of mobile emergency resources. This algorithm effectively combines multiple improvement strategies and priority experience replay strategy to improve the sampling efficiency and training effect. In the model-driven part

the optimization scheduling problem of multi-type mobile emergency resources is constructed as a mixed integer second-order cone programming model using the second-order cone relaxation and big M method to solve the optimal scheduling strategy of MESS and RC under the influence of renewable energy and load changes. Finally

the proposed method's effectiveness

generalization ability

and scalability are validated in two coupled power transportation networks.