Short-term forecast of automatic frequency restoration reserve from a renewable energy based virtual power plant

This paper presents the initial findings on a new forecast approach for ancillary services delivered by aggregated renewable power plants. The increasing penetration of distributed variable generators challenges grid reliability. Wind and photovoltaic power plants are technically able to provide ancillary services, but their stochastic behavior currently impedes their integration into reserve mechanisms. A methodology is developed to forecast the flexibility that a wind-photovoltaic aggregate can provide. A bivariate Kernel Density Estimator forecasts the probability to provide reserve. The methodology is tested on a case study where volumes of automatic Frequency Restoration Reserve (aFRR) are forecasted on a day-ahead horizon. It is found that the wind-photovoltaic aggregate can dedicate a limited share of its forecast production to aFRR. The frequency of insufficient reserve capacity is assessed, by comparing the capacities offered with the measured production.