We have been working with TERNA since the Call for Growth – Next Energy project, to develop a tool in support of the analyses for the reserve procurement and balancing services.
MMP approach started from the analysis of the statistical distributions of the historical forecasting errors of the PV and wind generation and load demand at market-node level, by applying Bayesian Mixture Models.
Recently, we extended this approach with the adoption of Extreme Value Analysis, a method generally used to predict return times of extreme weather events. This way, the tails of highly skewed residual load forecast error distributions are correctly caught in most of the various yearly and daily conditions. Moreover, robust inference on the occurrence probability of extreme forecast errors (beyond 99% of historical data) is made available despite the limited depth of the available datasets, to support TERNA in meeting reserve operating requirements.
