Powering Electricity Forecasting with Transfer Learning
Firuz Kamalov, Hana Sulieman, Sherif Moussa, Jorge Avante Reyes, Murodbek Safaraliev
Abstract
Accurate forecasting is one of the keys to the efficient use of the limited existing energy resources and plays an important role in sustainable development. While most of the current research has focused on energy price forecasting, very few studies have considered medium-term (monthly) electricity generation. This research aims to fill this gap by proposing a novel forecasting approach based on zero-shot transfer learning. Specifically, we train a Neural Basis Expansion Analysis for Time Series (NBEATS) model on a vast dataset comprising diverse time series data. Then, the trained model is applied to forecast electric power generation using zero-shot learning. The results show that the proposed method achieves a lower error than the benchmark deep learning and statistical methods, especially in backtesting. Furthermore, the proposed method provides vastly superior execution time as it does not require problem-specific training.