Abstract

In this paper we propose a different strategy to produce a 24-hours ahead PV power forecasts. The main idea consists in developing a set of simple univariate models based on Least square Support Vector Regression (LsSVR), each of the developed models will forecast the PV power of each 30 minutes of the next day. To compare LsSVR results and also to demonstrate the generic aspect of the proposed strategy, we use the Feed Forward Neural Network (FFNN) as benchmark model. We provide a comprehensive evaluation of the proposed approach using publicly available PV power database from Ausgrid utility.

Citation

A. Fentis, C. Lytridis, V. G. Kaburlasos, E. Vrochidou, T. Pachidis, E. Bahatti, M. Mestari, “A machine learning based approach for next-day photovoltaic power forecasting”, The Fourth International Conference on Intelligent Computing in Data Sciences (ICDS 2020), Fez, Morocco, 21-23 October 2020 (accepted).