Prediction of Electrical Energy Consumption of Educational Buildings in the Tropics Using Support Vector Regression (SVR)

The Rectorate Building of Hasanuddin University (Unhas) is located in a tropical region. Due to the activities inside this eight-floors, 3825.48 m2 building, it has a significant degree of energy use. Planning the electricity budget for a building necessitates estimating electrical energy consumption to aid in making energy management and conservation decisions. Enhancing the planning of energy performance is essential to conserve energy and mitigate environmental impacts, particularly in reducing CO2 emissions. The study used an SVR model to estimate electrical energy requirements in the Unhas Rectorate Building based on daily and hourly usage data, using 129 training and 39 test data points. SVR model in this study is using the Linear, RBF, and Polynomial Kernel. In this study the Kernel that produced the best result with value of $\varepsilon=0.01$ , and C = 1 was RBF kernel with RMSE value = 17169.47 and MAPE = 11.67% while Linear kernel had RMSE value = 39097.10 and MAPE = 38.00% and Polinomial kernel had RMSE value = 45320.47 and MAPE = 40.44%. Data on electrical energy usage patterns in this topic was collected from measurements and audits conducted at the Unhas Rectorate Building.