Short-term load forecasting without meteorological data using AI-based structures

Authors: İDİL IŞIKLI ESENER, TOLGA YÜKSEL, MEHMET KURBAN

Abstract: STLF is used in making decisions about economical power generation capacity, fuel purchasing, safety assessment, and power system planning in order to have economical power conditions. In this study, Turkey's 24-hour-ahead load forecasting without meteorological data is studied. ANN, wavelet transform and ANN, wavelet transform and RBF NN, and EMD and RBF NN structures are used in STLF procedures. Local holidays' historical load data are changed into data with normal day characteristics, and the estimation results of these days are not included in error computation. To obtain more accurate results, a regulation on forecasted loads is proposed. Regulated and unregulated forecasting error percentages are computed as daily average MAPE and maximum daily MAPE, and compared between the proposed structures. A simulation is performed for the years 2009--2010 via the user interface created using MATLAB GUI.

Keywords: Short-term load forecasting, artificial neural networks, radial basis function neural networks, wavelet transform, empirical mode decomposition

Full Text: PDF