RES on Power Operation: K-Means Clustering Over Elbow Approach

In this paper, a new optimum number of operation modes is developed for low, medium, and high penetrations of renewable energy source (RES). The data-driven approach is improved considering the unsupervised learning approach of the K-Means classifier with the assistance of the elbow method to decide the number of optimum cluster of the unknown data mode of the power grid. On the other hand, a combined approach of raw data and a 2-dimension Principle Component Analysis (PCA-2) feature reduction approach is developed to improve the performance in terms of the operation mode switching frequency (OMSF), but the highdimensional variance (HDV) is degraded as the RES penetration increases. The simulation results show that 1- dimensional PCA (PCA-1) can maintain the power system operation regardless of the penetration level of RES and the theoretical equations of the power system can be maintained under any penetration level of RES. In simulation, the proposed approach of combining the elbow method results in three, five, and seven clusters for low, medium, and high penetration levels of RES, respectively. As expected, this results in OMSF exceptionally reduction. However, the HDV is degraded in some scenarios of RES penetration. On the other hand, the PCA-1 approach results in constant three clusters for all RES penetration levels and the OMSF and HSV indices are maintained at the lowest value. Also, the seasonal consistency (SC) index is outperformed other techniques at the highest value.