An Efficient Power Dispatching Strategy to Cope with Renewable Energy Sources Uncertainties

This paper proposes the implementation of a receding horizon procedure to insert accurate and periodically updated Photovoltaic (PV) power production forecasts into an Energy Management System (EMS). This allows the EMS to compute the optimal scheduling of the dispatchable units of a Microgrid (MG) or polygenerative plant to minimize costs. The PV forecasts are obtained as outcomes of a Recurrent Neural Network (RNN). To allow reproducibility, the RNN is developed so that it requires input data available on open repositories and accessible with a timeliness consistent with the application. The RNN is used to input PV forecasts to the EMS. Two approaches to feed the EMS are compared: the first one hourly updates the PV forecasts and then launches a run of the EMS to dispatch the units for the next <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">${2 3}$</tex> hours. Then, only the control action related to the following hour are considered, and the other ones are discarded. The process is replicated hourly, giving origin to a receding horizon strategy. The second approach receives at midnight the PV forecasts for the next <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">${2 3}$</tex> hours and the EMS is run daily to dispatch the units. The effectiveness of the approaches is validated on historical data of a real MG comparing them with the ideal conditions, in which the PV forecast is exact.