Forecasting Annual Electricity Consumption in Indonesia: A GRNN-based Approach

Forecasting Annual Electricity Consumption in Indonesia: A GRNN-based Approach

Forecasting electricity consumption has played a critical role in planning a country's electrical energy system. This paperpresents a method of forecasting such annual consumption in Indonesia by the GRNN (generalized regression neural network)with k-fold cross-validation plus golden-section search and interpolation techniques. We proposed to resolve the best spreadparameter which could govern the performance of GRNNs. Cubic spline and polynomial regression were performed ascounterparts to be compared. Allowing for real data of more than two decades, numerical analysis indicates that the proposedmethod could outperform counterpart schemes, maintaining the least error of estimation of 1.47%. The well-trained model shallthen be utilized to forecast Indonesia’s annual electricity consumption for the next three years 2020 to 2022. Our model suggeststhat electricity consumption tends to increase steadily.