Analyzing the Relationship Between Meteorological Parameters and Electric Energy Consumption Using Support Vector Machine and Cooling Degree Days Algorithm

• Published in: Journal of information systems and informatics (Palembang.Online) Vol. 6; no. 2; pp. 729 - 750
• Main Authors: Azizah, Nabila Wafiqotul, Puspaningrum, Eva Yulia, Mas Diyasa, I Gede Susrama Susrama
• Format: Journal Article
• Language: English
• Published: Informatics Department, Faculty of Computer Science Bina Darma University 13.06.2024
• Subjects: cdd; electricity; meteorological parameters; svm crisp-dm
• ISSN: 2656-5935; 2656-4882; 2656-4882
• DOI: 10.51519/journalisi.v6i2.719

Nowadays, electricity is increasing rapidly. This increase is caused by several factors, one of which is meteorological factors. Meteorological parameters have various types, but this research uses three types in the form of temperature, humidity, and wind speed. The selection of these three types is due to the fact that they have a very close relationship with human life. In line with that, this research uses datasets obtained from the official websites of BMKG (Meteorology, Climatology and Geophysics Agency) and PLN (State Electricity Company). On this occasion, researchers used several methods, namely Cross-Industry Standard Process for Data Mining (CRISP-DM), Cooling Degree Days (CDD), and Support Vector Machine (SVM).  The CRISP-DM method is useful for describing the data mining cycle so that the process can be more organized. The SVM algorithm is useful for predicting electricity consumption based on meteorological parameters in January to April 2024, while the CDD method is useful for knowing the correlation of meteorological parameters to electricity consumption in winter. In line with this, this research produces predictions of electricity consumption based on meteorological parameters in January 2024 to April 2024 with an average range of 20.9 Watts per day. In addition, trends and predictions during model evaluation obtained a precision value of 0.796, recall of 0.793, F1 score of 0.793, MAPE of 17.2%, RMSE of 0.41, MAE of 0.167 and accurate of 0.98. These values indicate that the performance of the accuracy model is very high.