K-Means Algorithm to Identify the Optimal Production Period of a Photovoltaic Array Installed at Any Point in a Very Large Area

Abstract

This work presents a technique to identify the optimal production periods of a PV generator installed at any geographical point located in a very large area. This method is based on artificial intelligence techniques of time series clustering. Irradiation, temperature and precipitation were determined as influencing parameters of the PV power. As a study area, the entire Beninese territory was chosen, where 418 geographical points were strategically selected. influencing parameters were downloaded from the NASA database from January 1st 2012 to December 31st 2021. Using the k-means clustering algorithm, irradiation, temperature and precipitation clusters were formed. We arrived at 19 irradiation clusters, 41 temperature clusters and 41 precipitation clusters. Using these clusters, 78 sub-zones were then formed within the study area by gathering points with the same irradiation, temperature and precipitation clusters. The clustering models were developed in Python 3 using the tslearn library. The Silhouette score was used to evaluate the quality of the clusters and a minimum of 0.99 was obtained within the clusters as Silhouette score. The optimal production periods depend not only on the potential of the location where it is installed, but also on the evolution over time of the parameters impacting its power output.