Preventing Reverse Power Flow by Optimal Coordination of PV Feed-In: a Case Study in Southern Brazil
Abstract
System operators (SOs) in Brazil are experiencing growing times with reverse power flows at their sub-stations. These reverse power flows are due to increasing installations of photovoltaic (PV) systems, and may lead to technical complications. Therefore, SOs are looking for solutions to counter these reverse power flows. One approach is to curtail PV to zero feed-in for times with reverse power flows. This results in the loss of sustainable energy from PV for the Brazilian energy system. Alternatively, this paper presents an optimization model which limits the power flow for a feeder at the sub-station level, by coordinating a fair down-regulation of PV across the grid. PV estimate and prediction, based on machine learning, are used, to calculate the feeder’s pure load and to predict times of reverse power flows. This approach enables the use of excess PV by nodes without PV generation, increasing the consumption of sustainable energy in the system while preventing reverse power flows at the SOs sub-station. Results show that the ANN models produced satisfying results, and were deemed adequate for the purposes of this study. Up to 3.2 MW of reverse power flow occurred during the considered time periods, resulting in 1071 MWh over all analyzed periods. If all distributed PV were to be curtailed, 2817 MWh of renewable energy would not be utilized, which represents approximately 46.6% of total PV generation considered in the feeder. Through the dispatch optimization, the curtailment would only affect 1071 MWh of PV energy, representing 17.7% of total PV generation in the feeder.
Keywords
power system, reverse power flow, photovoltaic, optimization, machine learning