Abstract: In order to solve the current thermal power plant coal supply shortages, coal prices remain high status quo. In this paper, the optimal coal blending for pre-fired coal is studied through the actual use of coal in the power plant. The plan establishes a coal blending mathematical model that uses the minimum cost of blended coal as the objective function and the unit's requirements for the industrial components of the blended coal as a constraint to use the particle swarm. The local fast convergence feature of the algorithm is optimized by genetic algorithm to solve the model. The non-linear mapping relationship between the industrial components of the single coal blend coal quality is predicted by establishing a GA-BP neural network prediction model. Through the analysis of examples and error results, it is proved that this method can predict and solve the reliability of coal blending with the lowest cost.