AutoPQ: Automating quantile estimation from point forecasts in the context of sustainability

Optimizing smart grid operations relies on critical decision-making informed by uncertainty quantification, making probabilistic forecasting a vital tool. However, designing such forecasting models presents three key challenges: achieving accurate and unbiased uncertainty quantification, reducing the workload for data scientists during the design process, and minimizing the environmental impact of model training. In order to address these challenges, we introduce AutoPQ , a novel method that fully automates and optimizes probabilistic forecasting for smart grid applications. AutoPQ enhances forecast uncertainty quantification by generating high-quality quantile forecasts from an existing point forecast by using a conditional Invertible Neural Network (cINN). Furthermore, AutoPQ automates the selection of the optimal point forecasting method and fine-tunes hyperparameters, ensuring the best-possible model and configuration for each application. For flexible adaptation to various performance needs and available computing power, AutoPQ comes with a default and an advanced configuration, making it suitable for a wide range of smart grid applications. We demonstrate that AutoPQ surpasses state-of-the-art probabilistic forecasting methods while effectively limiting computational effort and hence environmental impact. Additionally and in the context of sustainability, AutoPQ provides full transparency regarding the electricity consumption required for performance improvements. • AutoPQ uses a conditional Invertible Neural Network (cINN) to convert point forecasts into quantile forecasts. • AutoPQ automates model selection and hyperparameter optimization for improving probabilistic performance. • It offers two configurations making it suitable for a wide range of smart grid applications: AutoPQ-default for general systems and AutoPQ-advanced for HPC systems. • AutoPQ shows a 15.1 % improvement over direct probabilistic benchmarks and a 9.1 % improvement over point forecast-based probabilistic benchmarks. • The ablation study shows the effectiveness of applied methods in reducing AutoPQ ’s electricity consumption.