Short Term Coal Price Prediction Based on VMD-Informer-LSTM Model Considering Error Compensation

The sustainable and healthy development of coal-fired power enterprises plays an important role in building a new type power system. The higher the proportion of installed renewable energy, the more prominent the supporting role of thermal power. The price of coal is will directly affect the generating willingness of power generation company, but its nonlinear and abrupt characteristics will make short-term prices difficult to predict. To address this issue, a VMD Informer LSTM short-term coal price prediction method is proposed, which takes into account error compensation. Firstly, the factors with high impact are selected through grey correlation analysis and Pearson correlation coefficient calculation. Secondly, the original coal price time series is decomposed into a series of relatively stable IMF sub signals through VMD decomposition to enhance the recognizability of temporal features. Then, each IMF is sequentially input into the Informer neural network for time series prediction, and the preliminary prediction results are obtained by stacking them. Finally, the prediction error is calculated and applied to the LSTM neural network to complete error compensation. The example shows that the proposed method can effectively improve prediction accuracy.