Objectives and background
Micro-grids delivering power in isolated areas (industrial mining sites, rural or insular inhabitations) are increasingly fuelled by hybrid PV-Diesel systems. Fuel transportation and consumption is costly, dangerous and pollutant. Maximizing the PV power integration while fulfilling the demand is critical for such micro-grid operators. The objectives of this work consist in assessing the benefits of cloud nowcasting to reduce the fuel consumption for different load profiles (residential and/or industrial).
Method
We used a nowcasting method based on images processing of a ground sky imager. The thermal-infrared whole sky imager Sky InsightTM designed by Reuniwatt, has unique advantages for an accurate very short-term forecasting of surface solar irradiance: its automatic distinction of cloud from clear sky is very robust whatever the position of the Sun in the sky; a non-dazzling effect from sunlight enables to forecast sudden PV production drop in the next few minutes through a clear distinction of the clouds surrounding the Sun; a capacity from a single observation point to retrieve and forecast a complete solar surface irradiance map.
Using 3 months of continuous Sky InsightTM observations in a maritime tropical area, we generated a 20-minute time horizon forecast every single minute of the daytime period. Then, we introduce these forecasts into a simulator of gensets consumption from a fictive hybrid PV-Diesel plant. The simulator is defined by the control rules of the hybrid system. A first rule sets the genset’s spinning reserve according to the forecast in order to compensate the worst forecasted drop. A second rule does not consider the PV output forecast: the spinning reserve is set as the current PV-output in order to compensate the potential loss of solar power. Gensets consumption is then simulated for each rule separately during these 3 months. Forecasts gain is assessed by comparing the fuel consumption generated by the first rule from the one of the second rule.
Several simulations are then processed considering whether the load profile is residential (correlated with daytime duration and then PV production), industrial (mostly uncorrelated with PV production) or mixed. Results are computed for several relevant configurations of PV and Gensets capacities (up to 30 MW in PV capacity; from 20 to 80% in peak power PV penetration).
Principal findings
Using solar irradiance forecasting is beneficial on several aspects. The blackout risk is easier to manage sudden drop from sharp cloud event occurs. Fuel consumption is reduced, even if this reduction is very sensitive to safety criterions applied for demanding load. A PV penetration rate higher than 30% can be safely managed. However, accuracy forecast using Sky InsightTM depends on cloud patterns. Difficult cases occurs when cloud appears spontaneously or if fast clouds move from outside the imager’s field of view at low solar elevation angle.
Conclusion
Using all-sky imagers forecast provides benefits for hybrid PV-Diesel micro-grids, in terms of fuel consumption and blackout risk management. Further studies are required to assess benefit of the shadow map forecasting over large size PV plants versus single or multiple punctual forecasts.