Location-aware green energy availability forecasting for multiple time frames in smart buildings: The case of Estonia

Abstract

In recent years, renewable energies (RE) have gained more attention as they provide clean and sustainable energy. Affordable and clean energy is one of the sustainable development goals (SDG-7) of the UN. Solar energy is considered to be one of the world's most abundant renewable natural resources and can certainly contribute to meeting the SDGs. The photovoltaic (PV) panel converts solar energy into electrical energy without emitting greenhouse gases. PV panels generate power based on solar radiation received at a specific location over time. As a result, PV output power is difficult to forecast. Many public/private institutes generate green energy, and balancing demand and supply is vital. This research aims to forecast PV system output power based on weather and derived features using different machine learning (ML) models. A best-fitting model is developed based on the data to accurately predict output power. Moreover, different performance metrics are used to compare and evaluate the accuracy under different ML models such as Random Forest (RF), Extreme Gradient Boosting (XGBoost), K-Nearest Neighbour (KNN), Multilayer Perceptron (MLP) and Support Vector Regression (SVR). Compared to the recent research, our results demonstrate that the obtained prediction accuracy is superior, with performance gains reaching up to 14% (R-squared) in the case of the RF model and 25% (MAPE) in the case of the SVR model. The comparison result inferred a similar performance when RF and XGBoost were applied. On the other hand, the ML models MLP, SVR, and KNN give a similar prediction pattern.