Comparison Analysis of Forecasting Accuracy for Electricity Consumption Using Extreme Learning Machine and Backpropagation Algorithms

Accurate calculations are essential in forecasting. Therefore, it is important to select the appropriate forecasting method, which can be determined by testing the accuracy level of the forecasting results using the MAPE value. The purpose of this research is to analyze the comparison of the accuracy of electricity consumption forecasts using the Extreme Learning Machine and Backpropagation algorithms. The data used are monthly electricity consumption data as input pattern designs. The input data will be divided into two parts: training data and testing data, followed by simulations using the Extreme Learning Machine and Backpropagation algorithms to obtain the 2022 forecast results. The accuracy of each algorithm's forecast results will then be calculated using the MAPE value. The research results indicate that the Extreme Learning Machine algorithm has a MAPE value of 4.67%, where the error rate is higher than the Backpropagation algorithm, which has only 2.27%. From the algorithm testing stages, the most accurate forecasting result for electricity consumption was obtained using Backpropagation, as it has the lowest MAPE value, indicating the highest accuracy and lowest error rate.