Implementation of Artificial Neural Network Method for Estimating Connected Power and Electric Energy Consumption

Abstract

Abstract—Electricity is vital for modern society’s welfare. Daily electricity usage depended on the customers’ type. Hence, there was a difference between the connected power with consumption. Therefore, there needed an estimation method for long-term connected power and energy consumption to improve the safety of energy management and operation plan for the generator. This research used the Artificial Neural Network method with a backpropagation algorithm model to estimate the connected power and electricity consumption. This method has the advantage of following past patterns after the training process. This research used data such as total population, Gross Regional Domestic Product, total customers, produced energy, remaining energy, distribution loss, total transformer, peak load, and load factor as the independent data. The energy consumption and connected power served as the dependent data. The data was taken from Srengat Network Service Unit, East Java, for ten years, which started in 2008. This research used literature study, information and data collection, information and data process, data estimation and analysis, and conclusion as the procedures. Based on the results, the best network structure was 9-9-2 with the 10-6 goal, 0.9 momentum value, and 0.15 learning rate to produce the smallest Mean Squared Error of 0.00442 in 2015, Mean Absolute Percentage Error of 7.88% for the connected power, and 11.27% on electricity consumption target.