Enhancing PV Forecasting Accuracy through Optimal Clustering Method Selection and Impact Assessment

This study investigates the impact of clustering on photovoltaic (PV) power prediction by utilizing two openly accessible datasets obtained from an Australian Microgrid. It evaluates the performance of five popular clustering algorithms—KMeans, DBScan, Gaussian Mixture Model (GMM), Spectral Clustering (SC), and Fuzzy C-Means (FCM)—by comparing their accuracy through metrics such as the Calinski-Harabasz (CH) Score, Davies-Bouldin (DB) Score, and Silhouette Score (SS). Additionally, the research examines two deep-learning models to assess how clustering influences forecasting accuracy. Performance metrics including Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), R-squared (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>) score, and Standard Deviation of Error (SDE) are used to compare the results of clustered versus non-clustered data. The findings suggest that clustering improves forecasting performance, with the most effective clustering method providing the best overall predictions, highlighting it as the most optimal approach.