EMD-LSTM Based Demand Forecasting Research on Power Engineering Banded Framework Procurement Materials

More and more electric power enterprises apply the framework agreement procurement mode for procurement, and accurate power material demand forecasting is a key link to guarantee the material response speed and stable execution of the framework. The power material demand sequence usually contains both linear time series and nonlinear time series components, and the traditional time series analysis methods are not reasonable for non-smooth and nonlinear data. Based on this, the method of forecasting using LSTM network model on the basis of EMD algorithm decomposition of material demand dataset is proposed. The sequences are decomposed by EMD according to a certain law to extract different frequency features in the complex sequences; secondly, the LSTM model is used for the different frequency features to obtain the time-series information and make predictions. Finally, through the experimental analysis, it is proved that the established model has strong feature extraction ability and can predict the material demand better.