Meteorological and weather forecast data-based prediction of electrical power delivery of a photovoltaic panel in a stochastic framework

The aim of this work is to yield a procedure for predicting expectation and covariance of power delivered by a generally tilted photovoltaic panel, based on available expectations and covariances of incident weather forecast data. The procedure consists of several steps: (i) neural-network-based identification of direct and diffuse solar irradiance models as static functions of available meteorological data; (ii) implementing the model of overall solar irradiance received on a tilted surface; (iii) determination and implementation of a thermal model of a photovoltaic panel, and (iv) implementing the unscented transformation to compute expectation and covariance of a stochastic variable after passing through a nonlinear mapping. Obtained simulation results present usefulness of the approach which is to be used within predictive control strategies of electrical storages in a microgrid with photovoltaic panels, that will optimize a combination of technical and economical criteria of the microgrid operation.