<![CDATA[Every building has certain electricity consumption patterns that depend on its usage. Building electricity budget planning requires a consumption forecast to determine the baseline electricity load and to support energy management decisions. In this study, an algorithm to model building electricity consumption was developed. The algorithm is based on the support vector regression (SVR) method. Data of electricity consumption from the past five years from a selected building object in ITB campus were used. The dataset unexpectedly exhibited a large number of anomalous points. Therefore, a tolerance limit of hourly average energy consumption was defined to obtain good quality training data. Various tolerance limits were investigated, that is 15% (Type 1), 30% (Type 2), and 0% (Type 0). The optimal model was selected based on the criteria of mean absolute percentage error (MAPE) < 20% and root mean square error (RMSE) < 10 kWh. Type 1 data was selected based on its performance compared to the other two. In a real implementation, the model yielded a MAPE value of 14.79% and an RMSE value of 7.48 kWh when predicting weekly electricity consumption. Therefore, the Type 1 data-based model could satisfactorily forecast building electricity consumption.]]>
Copyrights © 2021
