Short-Term Forecast for Turkey's Electricity Demand Dnn Vs Lstm

Abstract

This study aims to estimate Turkey's short-term electricity demand using artificial intelligence algorithms. Electrical systems are complex structures; therefore, many details must be considered for the prediction. Electricity demand forecasting depends on many conditions such as climate, calendar effect ( holidays, day of the week, etc.), demographic data, and economic data. Turkey is a relatively large and crowded country, whose population distribution is concentrated in some regions and climatic conditions, population-weighted meteorological data were used as independent variables. Predicting the future is challenging machine learning can help us understand how systems behave by identifying and analyzing patterns in data. Two advanced artificial neural network models were deployed in this study: a deep neural network (DNN) model, and stacked (deep) long short-term memory (LSTM) model. Their outputs provided estimates of hourly electricity consumption compared with the actual data. For this comparison, mean square error (MSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) metrics were used. It was observed that the DNN model predicted more accurately than the stacked LSTM model.