Adaptive wavelet neural network for wind speed and solar power forecasting for Italian data

Abstract

Conventional energy sources are nowadays exhausting and that is the reason why renewable energy sources are so important in current situation. In addition renewables are non-pollutant and freely available in nature. Wind and solar power are the fastest growing renewable energy sources for the past few decades, especially according to the 2020 energy strategy in Europe. They are having enough scope in the power market. The main problem with these renewable energy sources is their unpredictability and, in this context, issues like power quality and power system grid stability arise. In order to limit the effects of these issues, power market needs information about power generation at least one day in advance. This problem can be addressed by proper forecasting of Renewable Energy Sources (RES). Forecasting helps to schedule power properly. Adaptive Wavelet Neural Network (AWNN), a technique already assessed in literature for wind speed forecasting, is here applied also to solar power prediction. After forecasting each individual signal, the Mean Absolute Percentage Error (MAPE) is calculated in different time horizons.