Predictor-corrector method for weather forecast improvement using local measurements

Weather forecast is a crucial input for prediction of local building consumption and power production profiles in the building's microgrid. E.g., prediction of solar irradiance components and air temperature is used to predict photovoltaic array power production, while air temperature and humidity are often used to predict building consumption during the day. Due to the computation complexity of meteorological models, new prediction sequence becomes available every 6 h at best, and often comes with a nearly 4 h lag. In this paper we develop a linear and nonlinear corrector models to improve weather forecast by using local measurements only. The main motivation behind this approach is to correct prediction sequence by using local measurements as they become available, i.e. prediction sequence is refreshed every 1 h instead of every 6 h. The proposed approach is validated on historical air temperature prediction sequences and actual measurements during 6 months period.