Power Plant Coal Inventory Forecasting Based on LightGBM with Bayesian Optimization

Abstract

Accurate prediction of coal inventory in power plants is an important prerequisite for the smooth implementation of power plant supply guarantee policies. It is of great significance to accurately predict the supply risk of coal in advance. The coal inventory of hydropower plants has the characteristics of periodicity, randomness and seasonality, and is affected by coal input, coal consumption, meteorology and special policy events. Aiming at the problems that the prediction results of the existing power plant coal inventory prediction algorithm are not accurate enough, and the prediction is only based on historical data such as power generation and coal consumption, this paper proposes a new coal inventory prediction method for hydropower plants, which uses the learning advantage of the LightGBM model based on Bayesian optimization to the regression problem, determines the local optimal hyperparameter configuration, and uses the histogram algorithm and the gradient unilateral sampling algorithm, Reduce the number of training samples and features in the iteration to avoid over fitting. In this paper, the experimental analysis is carried out on the historical data of a hydropower plant. Finally, the accuracy of the model is evaluated by statistical indicators such as RMSE, Mae and correlation coefficient. Experiments show that compared with other algorithms, the method proposed in this paper has higher accuracy and is of great significance for the prediction of coal inventory in hydropower plants.