Solar Radiation Time-Series Prediction Based on Empirical Mode Decomposition and Artificial Neural Networks

Abstract

This paper presents a new model for daily solar radiation prediction. In order to capture the hidden knowledge of existing data, a time-frequency analysis on past measurements of the solar energy density is carried out. The Hilbert-Huang transform (HHT) is employed for the representation of the daily solar irradiance time series. A set of physical measurements and simulated signals are selected for the time series analysis. The empirical mode decomposition is applied and the adaptive basis of each raw signal is extracted. The decomposed narrow-band amplitude and frequency modulated signals are modelled by using dynamic artificial neural networks (ANNs). Nonlinear autoregressive networks are trained with the average daily solar irradiance as exogenous (independent) input. The instantaneous value of solar radiation density is estimated based on previous values of the time series and previous values of the independent input. The results are promising and they reveal that the proposed system can be incorporated in intelligent systems for better load management in photovoltaic systems.

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.