Very short-term scenario-based probabilistic forecasting of PV park power production

• DOI: 10.1049/icp.2023.2811
• Publication Date: 2023-12-14T20:09:01Z
• Authors: J. Munkhammar; O. Lindberg; M. Koubar
• Keywords: Probabilistic Forecasting

Grid-connected photovoltaic (PV) parks are increasing in number and size. For local optimal battery control, electricity market participation generally for delivering ancillary services to the grid from PV parks, it is important be able forecast park power generation. This study investigates short-term probabilistic forecasts scenario-based on clear-sky index photovoltaics with two Markov-chain mixture distribution (MCM) models, Persistence Ensemble (PeEn) Climatology. The models were trained on, used forecast, a 5 minute resolution data set of generation years Vasakronan AB's Uppsala, Sweden. shows that MCM outperform PeEn Climatology five ahead terms continuous ranked probability score point MAE. It also concluded outperforms Climatology, which despite lack accuracy has highest similarity result output. In scenario-forecasting, where compared outputs all have similar CDF goodness-of-fit. autocorrelation, superior. Based results, model, regardless setting, recommended as advanced benchmark very production forecasts.