Evaluating the Prediction of Wind Power Ramping Events in the Belgian Offshore Zone

TL;DR

This paper addresses the challenge of predicting short-term wind power ramps in the Belgian Offshore Zone (BOZ) and evaluates multiple forecast streams including ALARO-4km with and without Wind Farm Parameterization (WFP), power curves, and machine-learning based power forecasts. To make ramp predictions actionable for operators, the authors introduce a buffered ramp verification framework with time and power tolerances and a Ramp Alignment Score (RAS) to quantify temporal misalignment, defined as $RAS = |t_{pred}-t_{obs}|/T$. Their results show that WFP reduces false alarms and improves ramp timing relative to pure power-curve forecasts, while neural networks and XGBoost reduce power bias and overall MAE but do not consistently lower ramp MAE. The study also reveals that severe precipitation serves as a strong signal for large, predictable ramps, whereas moderate to light precipitation yields more uncertain ramps, underscoring the need for operationally relevant verification and suggesting future work on probabilistic ramp forecasting and Transformer-based sequence models.

Abstract

Evaluations are presented for the prediction of wind power ramping events in the Belgian Offshore Zone. Two models from the Royal Meteorological Institute of Belgium are verified: the operational ALARO-4km and its version with Wind Farm Parameterization (WFP). Power predictions are produced using power curves and machine learning (ML). As standard metrics such as MAE are insufficient for evaluating ramps, the proposed framework incorporates time and power buffers, enabling a flexible assessment that tolerates minor errors. Results indicate that WFP models enhance ramping prediction skill, while ML provides more balanced forecasts by reducing both misses and false alarms. A Ramp Alignment Score is also introduced to quantify temporal errors by forecast lead time, confirming that WFP models yield smaller average timing errors. Moreover, the framework reveals that severe precipitation is a strong indicator of large, predictable ramps, whereas lighter precipitation is associated with greater forecast errors.

Figures (17)

• Figure 1: The location of BOZ wind farms and installed capacity of each wind farm.
• Figure 2: PC model prediction bias (top) and MAE (bottom) of BOZ aggregated wind power predictions against power production in percentage of installed capacity, verified over the years 2022 and 2023.
• Figure 3: ML model bias (top) and MAE (bottom) of BOZ power predictions in percentage of installed capacity, verified over the year 2023.
• Figure 4: 15-minute ramping MAE (top) and 1-hour ramping MAE (bottom) of BOZ power ramping in percentage to the total capacity. The verification period is year 2023.
• Figure 5: PSD of wind power time series of observation and model predictions. The vertical line at 12 day$^{-1}$ represents the high-frequency ramping events.