Accurate prediction of coal prices is of great significance to defuse energy price risks. Aiming at the problem of coal price forecasting, a multi-step forecasting study of coal price based on ensemble model is carried out. This study analyzes the main controlling factors affecting coal prices, and establishes a data set; effectively integrates Particle swarm optimization (PSO) and Long Short-Term Memory (LSTM), and establishes a multi-parameter multi-step prediction model based on PSO-LSTM. Using the multi-parameter and multi-step forecasting model to call the data set to forecast the coal price of Caofeidian Port, the results show that the forecasting effect of the multi-parameter and multi-step forecasting model based on PSO-LSTM is better than the forecasting model based on BP and LSTM. It’s MAPE and R2 values of predicted price and actual price are 0.025 and 0.908 respectively, which can provide help for scientific control of coal market.