Permodelan Jaringan Saraf Tiruan Menggunakan Metode Backpropagation Untuk Prediksi Beban Listrik Di Sumatera Bagian Tengah

Muhammad Mayandre Bethatian, Rahyul Amri

Abstract


Electricity loads thrive every time, where the amount of loads affect the availability and supply of electricity every day. The calculation of changes in electrical loads for every 30 minutes in 24 hours, electrical loads
could produce various electrical pattern developments at certain times in different days, to understand the changes of the pattern of electricity loads for the future. These pattern could be use for prediction or
forecasting method on the daily data of electrical loads using Artificial Neural Network (ANN) with the Backpropagation Method. The prediction is applied to daily electricity loads data for the SUMBAGTENG
region (Central Sumatra), data from 2013 are used for training, 2014 are used for model testing, and data from 2015 as comparison of ANN prediction results from 2014 data testing. Optimal model training is
obtained by trying each training function and changing various training parameters to get the lowest Mean Square Error (MSE) value. The results of various training ANN models showed that using the traincgp
training function at 200 hidden layers and learning rate is 0.01 obtaining a training MSE value is 10025,265. By applying the optimal ANN model on the 2014 electrical loads to predict the 2015 electrical
loads, the accuracy of error from the ANN model is obtained by the value of Mean Absolute Perscent Error (MAPE) is 5.42%.
Keywords: Electric loads, electric loads forecast, artificial neural network, Backpropagation.

Full Text:

PDF

Refbacks

  • There are currently no refbacks.