Transformer Machine Learning Models for Time Series Electricity Power Forecasting

Time series forecasting is a cornerstone of predictive analytics in diverse domains, such as electricity power production (power plants) and transmission (power grids). Predicting electricity power is essential to renewable energy management. It helps optimize energy grids, reduce operational costs, and enhance power sustainability. The paper compares a few transformer and recurrent neural network models applied to time series data to estimate the wind farm's active power generation. The following machine learning models are described in detail: Patch Time Series Transformers (PatchTST), Time Series Transformers (TST), Long Short-Term Memory (LSTM), and Gated Recurrent Units (GRU). The prediction results of a known big data set (time series) are presented and analyzed in detail.