Data Mining Approach for the Optimal Locations of a Distributed Hybrid Renewable Energy System: A Case Study in Rural Western Australia

The adverse effects of coal and gas energy production with the subsequent rapid increase in energy consumption emphasize the importance for Australia to adopt more renewable energy sources to counteract these dismissive contributions to climate change. This work presents a data mining approach for optimally selecting the best locations for installing a distributed hybrid renewable energy generation system for rural regions in Western Australia. The K-Means and K-Medoids clustering algorithms were used to divide the constructed dataset into clusters. In total, 69 locations were selected for the overall dataset, proceeding with the filtering process. The returned cluster data were graphically rendered on a Western Australia map for the region. Using the Dunn index, the clustering algorithms were evaluated, such that K-Means (0.1458) performed to a higher degree compared to K-Medoids (0.0715), given the nature of our dataset. After passing the generated clusters to HOMER to generate the potential wind and solar energy output for each centroid, K-Medoids produced a set of locations that generated higher solar and wind energy on average. However, due to the reduced internal validation, K-Medoids might not be as valuable as K-Means, it does not cluster the data points very well, and within-cluster location energy requirements are not considered in our study.