Forecasting Household Electricity Consumption in the Province of Aceh Using Combination Time Series Model

The forecasting accuracy of household electricity consumption is very important for decisions making and policies activities. One of the needs of electricity power regionally is determined by forecasting household electricity consumption accurately. There are many linear exploration models have been established and widely used in time series forecasting. In this paper, we proposed the combination (hybrid) linear and non-linear models for modeling the household electricity consumption. The accuracy performance shown that the combination (hybrid) models from the three categories of data sets outperform the other individual models. In this analysis, we developed the Multiplicative SARIMA, Subset ARIMA, Feedforward Neural Networks (FFNN), and Combination (hybrid) model. The Multiplicative SARIMA and the Subset ARIMA models belongs to the linear models classes based on a well-known Box-Jenkins methodology model building process. The FFNN belongs to non-linear model class and the combination (hybrid) model is a mixing of linear and non-linear models. From the three categories of data sets, the proposed models for consumption of 450VA-2200VA, consumption of 3500VA-5500VA, and consumption of above 6600VA are hybrid Subset ARIMA-FFNN, hybrid Multiplicative SARIMA-FFNN, and hybrid Subset ARIMA-FFNN, respectively. These practical results show that the combination (hybrid) models are superior in comparison to the other individual models.