Temperature sampling data of microwave drying lignite has multimodal non-stationary characteristics and using wavelet threshold filtering can keep details of original data very well. The temperature prediction model for microwave heating of lignite via back propagation（ BP） neural network has low precision of prediction

slow convergence speed and easy to fall into local minima and other shortcomings. The mind evolutionary algorithm has strong global optimization capability to optimize initial weights and thresholds of BP neural network. The experimental results show that the BP neural network optimized by the mind evolutionary algorithm has higher prediction precision and generalization ability. The prediction performance has been significantly improved.