PENERAPAN METODE ELMAN RECURRENT NEURAL NETWORK DAN PRINCIPAL COMPONENT ANALYSIS (PCA) UNTUK PERAMALAN KONSUMSI LISTRIK

Electricity consumption in Indonesia each year continues to increase in line with national economic growth. Therefore, forecasting electricity demand in Indonesia is needed in order to describe the condition of the electrical current and the future. This study aims to apply the method of Elman Recurrent Neural Network and Principal Component Analysis (PCA) to construct a system for electricity consumption forecasting applications. Forecasting techniques used in this study is ARIMA Box Jenkins method used to determine the lag-lag effect on forecasting and Principal Component Analysis (PCA) is used to simplify the observed variables by means shrinking (reducing) dimension. Elman Recurrent Neural Networks Neural networks are used to model complex relationships between inputs and outputs to discover data patterns. Factors to be input ANN is a factor of population, GDP growth, industrial growth and the demographic data that includes customer electricity consumption of household, industrial, business, social and public. The results showed that the application of methods of Principal Component Analysis (PCA) to determine the dominant factors affecting power consumption and ARIMA Box Jenkins model can already be used to determine the lag-lag input data. Elman-RNN method is used to simulate the network parameters are established then performed to obtain the training and validation of the value of Mean Square Error (MSE) network. Accuracy of forecasting was measured using Mean Absolute Percentange Error (MAPE) and the average value of MAPE forecast in samples with 5-year forecast period for forecasting total consumption amounted to 0.33% 1, 2 total consumption amounted to 0.64%, 1.21% of households, industry 2.62% , business 3.25%, 0.77% and public social 0.49%