Due to the fluctuation of renewable energy output and load in multi-area integrated energy system, as well as the coupling relationship among multi-energy, it brings many challenges to the real-time optimal scheduling of multi-zone integrated energy system. To this end, this paper proposes a data-driven based multi-agent proximal policy optimization (MAPPO) algorithm for economic dispatch method of multi-area integrated energy system. Considering the energy trading and carbon market trading between areas, a real-time optimal scheduling model of multi-area integrated energy system is established to minimize the daily operating cost of the area. The optimization problem is modeled as a Markov decision process, and the state space, action space and reward function are designed. Through a large number of historical data training, the optimization scheduling neural network model of multi-area integrated energy system is obtained to realize multi- area decentralized real-time optimal scheduling. The results show that, under the influence of random fluctuations of new energy output and load, the proposed method can reduce the operating cost of each area, as well as the information interaction, which helps to improve the security of private information in each area.