Probabilistic Wind Power Forecasting with Tree-Based Machine Learning and Weather Ensembles

TL;DR

A comparative analysis across three state-of-the-art probabilistic prediction methods-conformalised quantile regression, natural gradient boosting and conditional diffusion models-all of which can be combined with tree-based machine learning finds the conditional diffusion model is found to yield the best overall probabilistic and point estimate of wind power generation.

Abstract

Accurate production forecasts are essential to continue facilitating the integration of renewable energy sources into the power grid. This paper illustrates how to obtain probabilistic day-ahead forecasts of wind power generation via gradient boosting trees using an ensemble of weather forecasts. To this end, we perform a comparative analysis across three state-of-the-art probabilistic prediction methods-conformalised quantile regression, natural gradient boosting and conditional diffusion models-all of which can be combined with tree-based machine learning. The methods are validated using four years of data for all wind farms present within the Belgian offshore zone. Additionally, the point forecasts are benchmarked against deterministic engineering methods, using either the power curve or an advanced approach incorporating a calibrated analytical wake model. The experimental results show that the machine learning methods improve the mean absolute error by up to 53% and 33% compared to the power curve and the calibrated wake model. Considering the three probabilistic prediction methods, the conditional diffusion model is found to yield the best overall probabilistic and point estimate of wind power generation. Moreover, the findings suggest that the use of an ensemble of weather forecasts can improve point forecast accuracy by up to 23%.

Figures (6)

• Figure 1: Location of the turbines in the Belgian offshore zone and the grid points from the DWD ICON-EU model (with a horizontal grid spacing of $6.5$ km).
• Figure 2: High-level overview of the probabilistic forecasting methodology.
• Figure 3: Visual example of the conformal prediction procedure with conformalized quantile regression from (\ref{['eq:cp_cqr']}), with the empirical quantile of the conformal scores $\hat{s}>0$. The dark blue area indicates the confidence interval predicted by the quantile regression model and the light blue area shows how the confidence interval is adapted after conformal calibration. The difference between the light blue and the dark blue area is determined by the empricial quantile of the conformal scores $\hat{s}$.
• Figure 4: The wind farms present in the Belgian offshore zone (in color) and the surrounding wind turbines from Borssele wind farm zone (NL) (in grey). In total, the Belgian offshore zone comprises more than 2 GW of installed capacity divided over 9 different wind farms. The first project, C-Power, was built in 2009 and the last projects, Mermaid and Northwester 2, were finished in 2020. The Borssele wind farm zone is operational since 2021.
• Figure 5: Illustrative example of the different day-ahead forecasts (including 50% and 80% confidence intervals) where the wind speed forecast was around the cut-in wind speed during the day.