In coal blending optimization, due to the lag of the calculation of power supply coal consumption, the power supply coal consumption of the blended coal can not be obtained, so usually the lowest price of blended coal can only be taken as the objective function ,but this does not guarantee the cost reduction and efficiency increase of power plant. Therefore, this paper uses the digital twin model of a 320 MW power plant based on Thermoflow to simulate the power plant burning different kinds of coal, establishes the database of coal quality and power supply coal consumption, and BP neural network is applied to establish the relationship model of coal quality and power supply coal consumption. On this basis, taking the coal quality index as the constraint condition, combined with the blended coal price and transportation expenses, and taking the lowest power supply coal consumption cost as the objective function, the coal blending optimization model is established. According to the actual demand of the power plant, the optimal blending scheme is given. The model can help the power plant quickly obtain the power supply cost of burning the coal or blending the coal, and provide guidance for the coal source selection and blending optimization of the power plant.