An optimized probabilistic forecasting approach for hybridized wind power plants
02 HYB24-60
Presented by: António Couto
Hybrid power plants (HPPs), incorporating diverse energy sources sharing the same electrical substation, are currently an interesting concept to fulfil European decarbonisation objectives, due to their multiple benefits such as sharing the same substation. Historically, power forecast research has predominantly focused on individual analysis of wind or solar photovoltaic (PV) power, neglecting the potential benefits of their combination. This study addresses this gap by investigating the synergistic effects of integrating wind and PV technologies within (utility-scale) HPPs from a forecasting perspective. Other gaps in existing literature comprise whether the integration of these technologies within HPPs can lead to reduced forecast errors and increased profitability compared to the traditional approach of forecasting and bidding for individual technologies in electricity markets. Being a relatively recent topic of research, the forecasts for this type of power plants, namely, the optimal forecasting approach and if the forecasts should be conducted separately or for pre-aggregated time series, are still underexplored in the existing literature.
This work presents a probabilistic power forecast approach applied to HPPs. The forecasting methodology utilizes a sequential forward feature selection algorithm, employing two distinct objective functions and an artificial neural network approach. Then, probabilistic power forecasts are obtained using a quantile spline regression technique. The approach supports the identification of the i) optimal quantile and ii) the exogenous features (e.g., meteorological input features from numerical weather prediction – NWP models) to increase the profitability of each HPP in an electricity market environment. The methodology is applied to three case studies in Portugal assuming the hybridization of existing wind power plants. The results are evaluated using technical and economic metrics, such as the profitability of HPP in the day-ahead and balancing markets of the Iberian electricity market and the root mean square error.
As expected, hybridization increases the remuneration of wind power producers compared to existing wind plants, regardless of complementarity levels. However, the increase in remuneration is superior in the case study with the highest generation complementarity. Exogenous parameters identified for each case study differ highlighting the need to select the input data carefully and have tailored-made power forecast models. Finally, the optimal forecast based on quantiles proves to be crucial for increasing the remuneration compared to the traditional deterministic approach based on the expected value.
This method was developed within the scope of the EU-funded H2020 project TradeRES - Tools for the Design and Modelling of New Markets and Negotiation Mechanisms for a ~100% Renewable European Power System.
This work has received funding from the EU Horizon 2020 research and innovation program under TradeRES project (grant agreement No 864276
This work presents a probabilistic power forecast approach applied to HPPs. The forecasting methodology utilizes a sequential forward feature selection algorithm, employing two distinct objective functions and an artificial neural network approach. Then, probabilistic power forecasts are obtained using a quantile spline regression technique. The approach supports the identification of the i) optimal quantile and ii) the exogenous features (e.g., meteorological input features from numerical weather prediction – NWP models) to increase the profitability of each HPP in an electricity market environment. The methodology is applied to three case studies in Portugal assuming the hybridization of existing wind power plants. The results are evaluated using technical and economic metrics, such as the profitability of HPP in the day-ahead and balancing markets of the Iberian electricity market and the root mean square error.
As expected, hybridization increases the remuneration of wind power producers compared to existing wind plants, regardless of complementarity levels. However, the increase in remuneration is superior in the case study with the highest generation complementarity. Exogenous parameters identified for each case study differ highlighting the need to select the input data carefully and have tailored-made power forecast models. Finally, the optimal forecast based on quantiles proves to be crucial for increasing the remuneration compared to the traditional deterministic approach based on the expected value.
This method was developed within the scope of the EU-funded H2020 project TradeRES - Tools for the Design and Modelling of New Markets and Negotiation Mechanisms for a ~100% Renewable European Power System.
This work has received funding from the EU Horizon 2020 research and innovation program under TradeRES project (grant agreement No 864276