Navigating the Safety-Fidelity Trade-off: Massive-Variate Time Series Forecasting for Power Systems via Probabilistic Scenarios
Researchers have introduced PowerPhase, a new benchmark for probabilistic time series forecasting in power systems, designed to evaluate models on a much larger scale than existing benchmarks. PowerPhase includes six transmission grids with up to 36,964 jointly forecasted channels and incorporates constraint-aware metrics alongside standard ones to assess both distributional accuracy and operational constraints. A proposed model, PowerForge, a scenario-based quantile forecaster, demonstrated superior performance across all grids by effectively balancing these safety and fidelity trade-offs. AI
IMPACT Introduces a large-scale benchmark and a novel model for power system forecasting, potentially improving grid stability and efficiency.