The paper "Towards Resilient Energy Forecasting: A Robust Optimization Approach" by Stratigakos et al. addresses the challenge of missing features in energy forecasting models, particularly in industrial applications where data availability is uncertain. The authors propose a robust optimization approach to design energy forecasting models that are resilient against missing features at test time. They formulate a robust regression model that minimizes the worst-case loss when a subset of features is missing, considering both point and probabilistic forecasting. Three solution methods are developed for the robust formulation, all leading to Linear Programming (LP) problems: vertex enumeration, deterministic reformulation, and affinely adjustable reformulation. The methods are evaluated in four prevalent energy forecasting applications—electricity price, load, wind production, and solar production—using extensive empirical validation. The results show that the proposed robust optimization approach outperforms imputation-based models and exhibits similar performance to retraining without missing features, while maintaining computational practicality. The paper contributes to the field by introducing resilient energy forecasting and benchmarking against missing feature scenarios, which are common in industrial applications.The paper "Towards Resilient Energy Forecasting: A Robust Optimization Approach" by Stratigakos et al. addresses the challenge of missing features in energy forecasting models, particularly in industrial applications where data availability is uncertain. The authors propose a robust optimization approach to design energy forecasting models that are resilient against missing features at test time. They formulate a robust regression model that minimizes the worst-case loss when a subset of features is missing, considering both point and probabilistic forecasting. Three solution methods are developed for the robust formulation, all leading to Linear Programming (LP) problems: vertex enumeration, deterministic reformulation, and affinely adjustable reformulation. The methods are evaluated in four prevalent energy forecasting applications—electricity price, load, wind production, and solar production—using extensive empirical validation. The results show that the proposed robust optimization approach outperforms imputation-based models and exhibits similar performance to retraining without missing features, while maintaining computational practicality. The paper contributes to the field by introducing resilient energy forecasting and benchmarking against missing feature scenarios, which are common in industrial applications.