Hybrid Forecasting Model Based on Nonlinear Auto-Regressive Exogenous Network, Fourier Transform, Self-organizing Map and Pattern Recognition Model for Hour Ahead Electricity Load Forecasting

Unlike other goods, electricity generated cannot be stored on an industrial scale. Adding to that, the supply and demand keeps on fluctuating in the market. If the supply and demand in the electricity market is mismatched, the change in speed of the generator causes change in system frequency. This may result in the addition or removal of either generation or connected load. The amount by which the electricity is lost in transmission and the loss due to congestion directly affects the price of electricity in the market. Therefore, it becomes vital to forecast the load of the electricity with high accuracy. A hybrid load forecasting model has been proposed in this paper that aims to reduce the overall forecasting error. Noise, the leading problem due to which load forecasting gets erroneous is addressed in this paper. The paper has also discussed the method to simplify the input waveform, for improving the performance of learning algorithm of the proposed load forecasting model.