Estimation of Electrical Energy Needs Using Backpropagation Neural Network

The need for electrical energy continues to increase, and electricity providers need to estimate (forecast) to take the right policies in the future. In this study, a test analysis was carried out to determine the electrical energy needs of the North Sumatra region since 2018 using the backpropagation neural network method. After several repeated training trials, a network architecture was obtained best with six entries consisting of input data on the number of customers (household, industrial, business, social), total population and GRDP of North Sumatra, a hidden layer with three neurons, and one neuron in the output layer, namely the consumption of electrical energy in units of GWh years 2018. Based on forecasting results and actual data in 2018, the average obtained (% MAPE) was 1.037%, with the MSE value of training at 0.00405 and the MSE value of testing at 0.00545. The forecast result is very accurate, that is, it is smaller (<10%).