干旱区生态环境知识资源中心

When large-scale photovoltaic power plants (PV-PPs) operate under partially-shaded conditions, their power output can be extremely fluctuating. This situation may compromise the energy balance of the electricity grid, which in turn threatens its secure operation from a frequency control viewpoint. In this context, the development of control strategies to reduce the variability of the power generated by PV-PPs is a key issue towards reaching sustainable electric systems. With this purpose, this paper proposes a novel control strategy to reduce the negative effects that PV-PPs operating under partially-shaded conditions may cause on the frequency control of electricity grids. The control operates the PV-PP in deload mode, i.e. keeping power reserves. The deload level of the PV-PP is set dynamically during the day considering a 10-min forecast of solar generation. The forecast is performed with artificial neural networks, first predicting the day-type (sunny, cloudy, overcast) and then the solar power. The controller continuously monitors the condition of the PV-PP: when the plant is under non-uniform shaded conditions, it deploys the power reserves to smooth the PV power. The proposed control was applied to a Chilean case study focused on the Atacama Desert, testing different control rules for the deload level. The obtained results show that the implementation of the proposed control considerably improves the frequency performance of the electricity grid. Although operating in deload mode implies energy losses in the PV-PP, the use of a dynamic deload level minimizes these losses when compared to a constant deload level. Altogether, the dynamic simulations show that such a control can play a relevant role for frequency control in electrical power systems with high shares of photovoltaic power. Our findings give important insights to electricity regulators about the technical requirements that they should impose to large-scale PV-PPs in electric power systems dominated by renewables energies. (C) 2018 Elsevier Ltd. All rights reserved.