Regional Photovoltaic Power Curtailment with Forecasting and Unit Commitment Scheduling: A Study on the Kanto Region in Japan
Abstract
Photovoltaic, PV, power generation is reaching unprecedented levels of penetration in many power grids around the world. Such trend is forcing utilities and transmission system operators to rethink the traditional ways to schedule the operation of their power generators. In this study we investigate the impact of day-ahead forecasts of PV power on the unit commitment scheduling of a regional power system. To do that we assumed a scenario of 33 GW of PV installed in the Kanto region in Japan, and that PV power can be curtailed. Using day-ahead forecasts of PV power, we generated daily unit commitment scheduling for one year, and compared them with those generated with a forecast without error. Annually, the amount of power curtailed based on the day-ahead forecasts differed in 1% of the amount required when a perfect forecast was used. Nevertheless, day-ahead forecast errors caused an increase 1.7% of the generators fuel costs (a significant value for cost). Day-ahead forecast errors also caused a need of curtailment in months when the relation between power demand and PV power generation by itself would not require it, such as June and July. Finally, the detection of PV curtailment based on day-ahead forecasts had a positive predictive value (precision) of 73%, but a true positive rate (hit rate) of 32%. Such results are mainly explained by the poor performance of the forecasts on spring and summer in Kanto, indicating a specific period when the forecast error should be improved.