Photovoltaic forecast based on hybrid PCA-LSSVM using dimensionality reducted data

The power forecasting plays a significant role in the electrical systems. Furthermore the high-dimensional data reduction without losing essential information represents an important advantage in the forecasting models. Low computational costs and short execution time together with high predicted performance are the main goals to be reached in the development of a prediction method. In this paper a hybrid method based on an active selection of the support vectors, using the quadratic Renyi entropy criteria in combination with the principal component analysis (PCA), is shown to dimensionally reduce the training data in the forecasting models. The reduced data have been used to implement the Least Squares Support Vector Machines (LS-SVM) in order to predict the photovoltaic (PV) power in the day-ahead time horizon. The model has been validated using historical data of a PV system in the Mediterranean climate. Additionally the weather variations have been taken into account to evaluate the outcome of the sunny and cloudy condition in the PV forecasting models. The proposed technique gives fulfill results. A training data size same as 30% original dimension allows to improve the forecasting accuracy and reduces the computational time of 70% respect to an implementation without dimensionality reduction data.