Day-Ahead Parametric Probabilistic Forecasting of Wind and Solar Power Generation Using Bounded Probability Distributions and Hybrid Neural Networks
Theodoros Konstantinou, Nikos Hatziargyriou
Abstract
The penetration of renewable energy sources in modern power systems increases at an impressive rate. Due to their intermittent and uncertain nature, it is important to forecast their generation including its uncertainty. In this article, an ensemble artificial neural network is applied for day ahead solar and wind power generation parametric probabilistic forecasting. The proposed architecture includes two components: a sub-models component and a Meta-Learner component. The first component includes an ensemble of artificial neural networks that have the ability to estimate the parameters of an underlying probability distribution. The Meta-Learner is responsible for grouping the training samples based on the estimated level of generation, through a classification-clustering process and use the output of the corresponding sub-models to calculate the final parametric probabilistic estimation. The proposed model is compared to both parametric and non-parametric state of the art probabilistic techniques for solar and wind power generation forecasting, exhibiting superior performance.