Short-Term Load Forecasting Using Grey Models by Considering Demand Response

Recently, demand side management become strategic concept in smart grids. Most probably, demand response programs will fail in the absence of accurate load prediction. Hence in this thesis, short term load forecasting using proposed grey models has been studied. So far numerous methods with different accuracy have been proposed to model and forecast electric load in short-term. Most of them utilize large amounts of data and other parameters of the predictor variable such as temperature. This is one of the biggest disadvantages of this methods, because the information is not always available, or these methods require an explicit mathematical model to determine the relationship between load and several input. In this thesis, grey model and rolling grey model that can use low data and high accuracy to modelling and predicting time series, is improved using a method based on iterative. To increase the accuracy of proposed grey model, fourier residual correction grey model is employed. Four different scenarios for demand side management programs is considered. Then, the performance of the proposed models before and after the demand response have been compared with artificial neural network and neuro-fuzzy based methods by applying them on Iran and NewEngland networks in MATLAB software environment. Also, the sensitivity of the proposed models to the number of required data and prediction step size has been investigated. Simulation results show high performance and accuracy of the proposed methods in the modeling and load forecasting with using low number of data. Among these models, the error of fourier grey model is 0.8141% that showed highest accuracy, and error of rolling grey model toward grey model is greater. It was observed with increasing number of data and prediction step size error of models increases.