Hybrid intelligent system for daily maximum temperature forecasting in smart grids

This paper proposes a hybrid intelligent system for temperature forecasting in smart grids. In recent years, the uncertainties increase due to the competitive power markets and the emergence of renewable energy such as PV and wind power generation. The prediction of one-step ahead daily maximum temperature plays a key role to deal with Demand Response (DR) and PV under smart grid environment as well as load forecasting and electricity price forecasting. In this paper, a hybrid intelligent system is proposed for daily maximum temperature forecasting. As a precondition technique for a predictor, Regression Tree (RT) of data mining is used to classify input data into clusters where GP (Gaussian Process) of the kernel machine is constructed to predict temperature precisely. GP has advantage to evaluate the upper and the lower bounds of the predicted value as well as the expected value of the predicted one. The proposed method is successfully applied to real data of the daily maximum temperature. A comparison is made between the proposed and the conventional intelligent methods such as MLP of artificial neural network (ANN) and GP.