Forecasting a Photovoltaic Power Output with Ordinary Differential Equation Solutions Using the "Aladin" Model

Accurate forecasting of the renewable power generation is important for the system operation, utilization and integration in the electricity grid. The photovoltaic output power is primarily dependent on the solar radiation, which short-term local forecasts, available from the numerical model "Aladin", can enter power models, trained with corresponding real time-series of few last days, to predict the following day electricity production. Presented daily updated polynomial derivative models can describe fluctuant function relations between input solar irradiance time-series and the scalar output power, which conventional regression solutions usually fail. Differential polynomial network is a new neural network type, which can define and solve a selective form of the linear ordinary sum differential equation to model 1-variable function series. Partial sum relative fraction terms, produced in all layers nodes of the network backward structure, can substitute for the time derivatives at several time-points of data series.