There are problems such as surplus water and breach of contract caused by the mismatch between the bid winning electricity quantity and the generated water quantity in the upstream and downstream cascade power stations in hydropower enrichment areas. Thus a medium and long-term transaction model of hydropower enrichment grid with the consideration of incentive compatibility is proposed based on such a model of the Brazilian electricity market. The model takes the maximum benefit of individual power generation and the minimum cost of power purchase at the grid side as the dual goals of incentive compatibility. First, the minimum power generated of hydropower system is maximized to optimize the operation of the hydropower system, and the monthly generating capacity of each hydropower station is defined as the guaranteed electricity of this month. This is divided into basic, provincial and west-to-east electricity. Then, market trading methods at all levels are established. Basic electricity is purchased according to the average price of previous years with similar inflow, the provincial electricity participates in two rounds of auction, and the west-to-east electricity is purchased successively according to the proportion of remaining electricity. A reinforcement learning method is used to determine the share and price of each power producer. Finally, the improved electricity redistribution mechanism is used to reallocate the power generation of each hydropower station in the alliance. The results for 52 hydropower stations in 5 river basins in Yunnan Province show that the proposed model can effectively alleviate the problems of surplus water and breach of contract, and promote rational market competition and the full utilization of water resources.