Prediction of photovoltaic system output using hybrid least square support vector machine

The electrical system photovoltaic (PV) modules required special design considerations due to unpredictable and sudden changes in weather conditions such as the solar irradiation level as well as the cell operating temperature. Therefore, this study presents a practical and reliable approach for the prediction of PV power output using an intelligent-based technique namely Cuckoo Search Algorithm — Least Square Support Vector Machine (CS-LSSVM). Available historical output power data are analyzed and appropriate features are selected for the model. There are two inputs vectors to the model consists of solar irradiation and ambient temperature. Cuckoo Search Algorithm (CS) is hybrid with LS-SVM in order to optimize the RBF parameters for better prediction performance. The performance of CS-LSSVM is compared with those obtained from LS-SVM using cross-validation technique in terms of accuracy. In this paper, Mean Absolute Percentage Error (MAPE) is used to quantify the performance of the prediction. Besides that, evaluation also carried out by calculating the correlation of determination. The historical PV data is utilized to validate the workability of the proposed technique. The results showed that CS-LSSVM provides better performance in predicting photovoltaic system power output.