Abstract: In order to promote the low-carbon transformation of the power system, reduce energy consumption and carbon dioxide emissions, rationally plan the ratio of installed capacity between renewable energy and thermal power, reduce system costs, and reduce the occurrence of wind and light abandonment, the cost model of fire, wind, light and water power system is first established. Non-dominated Sorting Genetic Algorithm II (NSGA-II) was constructed to optimize the installed capacity allocation of the power system with the goal of lowest total system cost and maximum renewable energy generation. The rationality of the model is verified. The results show that the NSGA-II algorithm is reasonable to optimize the ratio allocation of thermal power and new energy capacity. In a northwest region of China, thermal power∶new energy = 1∶1.5. The flexible transformation of thermal power units has a certain improvement in the carrying capacity and absorption capacity of new energy installed capacity, but the long-term effect is limited. Under the current situation, excessive increase in the proportion of installed capacity of new energy will increase the amount of abandoned wind and light and the total system cost, among which, the total system cost will increase more when the installed capacity of wind power is large; Considering the flexible transformation of thermal power units and the addition of energy storage installations, the proportion of thermal power is reduced to 40%.