Implementation of System Identification Models to Predict Demand Response Potential in Smart Grids

Under the smart grid environment, utilizing Demand Response Resources (DRRs) can lead to more suitable management of the growing energy consumption as well as considerable reduction of system expenditures. Accordingly, precise forecasting of DRRs potential is contemplated as a crucial issue in power system studies. In this paper, different Linear Models of System Identification (LMSI) including ARX, ARMAX, OE, BJ and, 4SID are implemented to predict the potential of DRRs. Here, the most effective factors including power consumption, energy price, and ambient temperature are chosen as input data of the system identification models. These models are trained using the BOSTON dataset in 2015. The least squares estimator is utilized to adjust the coefficients of the models. Then, the validity of the trained models is assessed via RMSE, MAD and MAPE indicators. Finally, the accuracy of the developed models is compared with the Multi-Layer Perceptron (MLP) neural network. The obtained results show that the BJ model, as a LMSI, provides the best prediction performance.