Forecasting Short-Term Solar PV Using Hierarchical Clustering and Cascade Model

Abstract

With the large-scale deployment of solar PV installations, managing the efficiency of the generation system became essential. Generally, the power output is heavily influenced by solar irradiance and sky conditions which are consistently changing. Thus, the ability to accurately forecast the solar PV power is critical for optimizing the generation system, estimating revenue, sustaining profits, and ensuring the quality of service. In this paper, the authors propose a solar PV forecasting model using multiple blocks of GRUs and RNN in a cascade model combined with hierarchical clustering to improve the overall prediction accuracy of solar PV forecast. This proposed model is a combination of hierarchical clustering, the Pearson correlation coefficient for feature selection, and the cascade model with GRU layer from k-means clustering and hierarchical clustering. These results, which are evaluated using NRMSE, show that hierarchical clustering is more suitable for solar PV forecast than k-means clustering.