A Direct Classification Approach for Reliable Wind Ramp Event Forecasting under Severe Class Imbalance

Alejandro Morales-Hernández; Fabrizio De Caroa; Gian Marco Paldino; Pascal Tribel; Alfredo Vaccaro; Gianluca Bontempi

cs.LGMar 20, 2026

A Direct Classification Approach for Reliable Wind Ramp Event Forecasting under Severe Class Imbalance

Authors: Alejandro Morales-Hernández, Fabrizio De Caroa, Gian Marco Paldino, Pascal Tribel, Alfredo Vaccaro, Gianluca Bontempi

Organizations: Machine Learning Group, Universite Libre de Bruxelles, Brussels, 1050, Belgium · University of Sannio, Benevento, 82100, Italy · WEL Research Institute, Wavre, 1300, Belgium

Abstract

Decision support systems are essential for maintaining grid stability in low-carbon power systems, such as wind power plants, by providing real-time alerts to control room operators regarding potential events, including Wind Power Ramp Events (WPREs). These early warnings enable the timely initiation of more detailed system stability assessments and preventive actions. However, forecasting these events is challenging due to the inherent class imbalance in WPRE datasets, where ramp events are less frequent (typically less than 15% of observed events) compared to normal conditions. Ignoring this characteristic undermines the performance of conventional machine learning models, which often favor the majority class. This paper introduces a novel methodology for WPRE forecasting as a multivariate time series classification task and proposes a data preprocessing strategy that extracts features from recent power observations and masks unavailable ramp information, making it integrable with traditional real-time ramp identification tools. Particularly, the proposed methodology combines majority-class undersampling and ensemble learning to enhance wind ramp event forecasting under class imbalance. Numerical simulations conducted on a real-world dataset demonstrate the superiority of our approach, achieving over 85% accuracy and 88% weighted F1 score, outperforming benchmark classifiers.

Figures & tables

\mathcal{F}_{i}=sgn\left(\sum_{j=1}^{S}\alpha_{i,j}h_{i,j}(x)-\theta_{i}\right)

Algorithm 2 EasyEnsemble

Figure 1: Sample of the wind power time series adopted in this study. Five different events are highlighted with different colors to indicate the five different classes considered in the forecasting task.

Figure 2: Probability of ramp occurrence in the dataset

Reducing the number of classes (from 5 to 3 classes)
Lag	Shannon entropy (%)	Accuracy	Balanced accuracy	Kappa score	Weighted F1 score
4 (1 hour)	41.95	0.907	0.915	0.679	0.92
8 (2 hours)	41.955	0.875	0.611	0.49	0.882
12 (3 hours)	41.959	0.853	0.699	0.5	0.874
Using the original class definition (5 classes)
Lag	Shannon entropy (%)	Accuracy	Balanced accuracy	Kappa score	Weighted F1 score

Table 1: Performance of EasyEnsemble with different lag values and number of classes. The best results are highlighted for each metric

Panel of Figure 3 — Figure 3: Confusion matrix of the classification results using EasyEnsemble with $l=4$ and (a) 3 classes and (b) 5 classes. The values were normalized over the true conditions (i.e. rows)

Reducing the number of classes (from 5 to 3 classes)
Algorithms	Acc.	Balanced accuracy	Kappa score	Weighted F1 score	Training time (s)	Test time (s)
Our proposal	0.907	0.915	0.679	0.917	2710.68	19.4
BB	0.849	0.679	0.473	0.867	0.16	$<0.01$
BRF	0.819	0.727	0.437	0.846	0.51	$<0.01$
RF	0.777	0.724	0.385	0.816	1.21	$<0.2$
RB	0.877	0.727	0.54	0.887	0.65	$<0.01$

Table 2: Results for forecasting the ramp event’s occurrence using

l=4

(1 hour). The best results are highlighted for each metric

Figure 4: Feature relevance in EasyEnsemble computed through the Gini Importance or Mean Decrease in Impurity (MDI)

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Panel of Figure 1 — Figure 1: Data pre-processing using $l=4$ and four events ( $C=4$ ). (a) The original bivariate time series $\{(x^{(t)},r^{(t)})\}_{t=1}^{T}$ , with rows as variables and columns as timestamps. (b) Selection of training instances and targets by applying a moving window on the time series (with a stride of 5). (c) Each instance is flattened, and (d) the information regarding the ramp type is masked depending on its availability, and additional features are extracted from the generated power window.

Figure 2: Real-time processing of the bivariate time series for ramp event forecasting

Hyperparameter	Type	Domain	Optimal value
Hyperparameter	Type	Domain	3 classes	5 classes
Max depth	Int.	[1, 10]	7	4
Estimators ( $\mathcal{L},S$ )	Int.	[1, 200]	199	90
Criterion	Cat.	[‘log_loss’, ‘gini’, ‘entropy’]	‘entropy’	‘entropy’
Min samples to split	Cont.	(0, 1]	0.159	0.327
Min samples in a leaf	Cont.	(0, 1)	0.03	0.059

Appendix

Table 1: Hyperparameter configuration of EasyEnsemble

Wind Power Forecasting

Wind power forecasting predicts turbine and farm output from operational data, weather forecasts, and remote observations to support reliable energy planning.

See all 24 papers

Hybrid Semantic Context-Enhanced Ensemble Learning for Wind Power Ramp-Event Forecasting and Uncertainty-Aware Evaluation
Aug 29, 2026Momina Liaqat Ali, Muhammad Abid, Muhammad Abdullah +1Wind Power Forecasting Time Series Forecasting
Department of Industrial and Systems Engineering, University of Tennessee, Knoxville Knoxville, TN 37996, USA. · Department of Mechanical and Aerospace Engineering, University of Tennessee, Knoxville Knoxville, TN 37996, USA. · Department of Mechanical Engineering, Pakistan Institute of Engineering and Applied Sciences, Islamabad, Pakistan. +1
WPBench: A Comprehensive Benchmark for Wind Power Forecasting
Sep 21, 2026Yuhan Zhu, Jilin Hu, Xinying Cai +7Multivariate Time Series Forecasting Benchmark Design
School of Data Science & Engineering, East China Normal University, Shanghai, China · College of Computer Science and Technology, Zhejiang University, Hangzhou, China · Hangzhou Hikvision Digital Technology Co., Ltd., Hangzhou, China
Tyan-WP: A Wind Power Foundation Model for Ultra-Short-Term Probabilistic Forecasting
Jun 7, 2026Jiahui Huang, Ao Luo, Lei Liu +6Zero-Shot Time Series Forecasting Time Series Forecasting
School of Information Science and Technology, University of Science and Technology of China, Hefei, 230022, China · China Electric Power Research Institute, Beijing, 100192, China

Aug 29, 2026J/K move · Enter open · S save

Hybrid Semantic Context-Enhanced Ensemble Learning for Wind Power Ramp-Event Forecasting and Uncertainty-Aware Evaluation

Momina Liaqat Ali, Muhammad Abid, Muhammad Abdullah, Aneela Zameer

Department of Industrial and Systems Engineering, University of Tennessee, Knoxville Knoxville, TN 37996, USA. · Department of Mechanical and Aerospace Engineering, University of Tennessee, Knoxville Knoxville, TN 37996, USA. · Department of Mechanical Engineering, Pakistan Institute of Engineering and Applied Sciences, Islamabad, Pakistan.+1

Wind Power Forecasting Time Series Forecasting Ensemble Learning

Open paper AskPDF arXiv

Wind power ramp events which are sudden, large swings in turbine output over short windows are difficult to estimate, and standard models often miss them. Hybrid forecasting approach is built which augments semantic context to ramp-event forecast. Rather than applying an extensive language model directly to predict turbine operating data, we have implemented a pipeline where turbine operating data is converted to simplified text, which is then converted to dense embeddings to be used as inputs for ensemble models incorporated with other features. Testing runs are performed at multiple intervals within the SDWPF dataset, including 10-minute, 30-minute, and 60- minute horizons, with ramp events constituting the highest change in future power output. We check robustness against autoregressive, LSTM, and GRU baselines plus several ensemble configurations, using Diebold-Mariano tests and bootstrap confidence intervals, and we vary the ramp threshold, compress the embeddings with PCA, and validate externally on Kaggle SCADA and NREL data with uncertainty-aware scoring. The semantic-context features produce negligible yet statistically significant gains over the baselines in multiple paired ensemble runs, most clearly at the 30- and 60-minute horizons where these gains hold across different ramp-threshold definitions, and PCA compression helps in some longer-horizon cases. The best context- augmented ensembles rank near the top overall, though the GRU model still posts the lowest ramp-event RMSE at 30 and 60 minutes. External tests confirm the error reduction generalizes across datasets, but the size of the gain depends on both model and dataset. Prediction intervals cover most test cases well but weaken during ramp events, pointing to a localized shift in the data distribution.

Abstract

Wind Power Forecasting

Explore similar work

Hybrid Semantic Context-Enhanced Ensemble Learning for Wind Power Ramp-Event Forecasting and Uncertainty-Aware Evaluation

WPBench: A Comprehensive Benchmark for Wind Power Forecasting

Tyan-WP: A Wind Power Foundation Model for Ultra-Short-Term Probabilistic Forecasting