Analysis of Electricity Consumption Patterns a Hybrid Algorithm K-Means Clustering and Support Vector Machine

Abstract

Fluctuations in electricity consumption make it difficult to manage and control energy use efficiently. Therefore, an analytical method capable of accurately identifying electricity consumption patterns is needed. Previous studies have generally applied clustering and classification separately and relied mainly on historical data, limiting their ability to capture dynamic electricity consumption patterns and classify new observations. This study aims to analyze electricity consumption patterns using a machine learning approach based on K-Means Clustering and Support Vector Machine (SVM). The data used consist of historical data and 24-hour real-time data obtained from PT PLN (Persero) UP3 Kendari. The research stages include data preprocessing, feature engineering, standardization using Z-scores, the clustering process using K-Means, and classification using SVM in a hybrid approach. The novelty of this study lies in using K-Means cluster labels as target classes for SVM while combining historical and real-time data. The results show the formation of four clusters: Cluster 0 representing stable moderate consumption, Cluster 1 representing high consumption, Cluster 2 representing low consumption, and Cluster 3 representing fluctuating consumption. Evaluation yielded a Silhouette Score of 0.5095 and a Davies-Bouldin Index of 0.6868, indicating fairly good cluster quality. The SVM model achieved an accuracy of 99.23% with high precision, recall, and F1-score values. These results demonstrate that the hybrid approach is effective in improving analysis performance and producing a reliable model for classifying new electricity consumption data