Clustering-based assessment of solar irradiation and temperature attributes for PV power generation site selection: A case of Indonesia’s Java-Bali region

Abstract

This study presents clustering-based assessments of solar attributes for locating potential solar photovoltaic (PV) power plant sites using k-means and density-based spatial clustering of applications with noise (DBSCAN) by examining the yearly average single-attribute and three-attribute clustering on a dataset of long-term hourly-based direct and diffuse irradiation, ambient temperature, and solar PV power output from 2005 to 2022. Three-attribute clustering enables stakeholders to better understand the characteristics of a cluster by collectively identifying three solar attributes and the magnitude of each attribute in an area or cluster. The presence of this information, which constitutes the clusters, suggests that these attributes have different effects on solar PV output power in different clusters. Although k-means is an effective method for investigating potential locations for PV power plant placements, DBSCAN offers users an alternative method for accomplishing a similar goal. In the case of three-attribute clustering of direct irradiation with k-means and DBSCAN, the 18-year mean value of clusters with the highest yearly average value is achieved at very similar values of 0.305 kW/m2 and 0.310 kW/m2, respectively. It turns out that only six years of direct irradiation had an annual mean value of less than 0.305 kW/m2. This finding implies that in the long run, the solar resources in terms of direct irradiation will typically surpass 0.3 kW/m2/MW installed capacity over all areas suitable for PV power plants. While focusing on the Java-Bali region, Indonesia, the findings, and methods appear to be of broader interest to policymakers, particularly in developing countries where solar PV is considered an option for sustainable energy generation.

Keywords

• clustering
• solar pv
• site selection
• solar attributes
• long-term