Research On The Daily Electricity Forecast Model Based On LightGBM

Abstract

Short-term electricity demand forecast is of great significance for guiding the Power Grid Company production operation, but the existing forecast mainly concentrated on load forecast, and most of them only use meteorological and date factors as input features, ignoring the economic factors as a basis to support the growth of electricity, also doesn’t take into account the influence of recent significant events. To improve the accuracy and rationality of short-term electricity demand forecast, based on the Light Gradient Boosting Machine (LightGBM) model, in addition to the common meteorological and date factors, all kinds of economic and significant events are introduced into the model to forecast the daily electricity in this paper. The results show that economic and significant events provide useful information, making our model more accurate than previous short-term electricity demand forecasts. Moreover, due to the adoption of machine learning algorithms, our model is more stable and accurate than traditional measurement models with the same factors.