For the dynamic weapon target assignment (DWTA) problem, a memetic algorithm (MA) based on evolutionary algorithm and local search algorithm is proposed. In order to maximize the target damage, the DWTA model considering the constraints of capacity, strategy, resource and interception feasibility was established. A virtual permutation was introduced to encode the DWTA model to meet the requirement of interception feasibility, and the construction method of transforming the permutation into the actual allocation scheme was designed, and the method of dealing with random events in the operation process of the algorithm was given. Compared with genetic algorithm (GA) and memetic algorithm global local search (MA-GLS) in solving different test cases, the simulation results show that MA algorithm has the advantages of fast optimization speed, strong optimization ability and good stability.