Tóm tắt xin tài trợ

Uncertainty in weather, demand, storage, and grid conditions drives renewable-energy curtailment, fossil-fuel peaker use, and disproportionate pollution exposure in communities near conventional power plants. Existing forecasting and optimization tools operate in isolation, rely on centralized data collection, and fail to convert predictions into reliable real-time decisions. This two-year project will develop the first integrated, health-aware artificial-intelligence framework that combines probabilistic forecasting, privacy-preserving federated learning, and grid optimization to coordinate distributed renewable-energy resources. Unlike systems that optimize only cost and reliability, the framework will incorporate population vulnerability and estimated pollution exposure directly into dispatch decisions, allowing operators to assess technical performance, emissions, equity, and health impacts. We will build multimodal models that integrate weather observations, power-generation data, electricity demand, storage status, and public environmental-health data to generate calibrated forecasts of renewable supply and grid stress. We will design a federated reinforcement-learning method to coordinate solar, wind, batteries, and flexible loads across multiple sites without sharing sensitive raw data. Large-scale grid simulations and pilot datasets will benchmark the framework against standard forecasting and dispatch methods using accuracy, emissions, reliability, equity, and estimated health benefits. The project will improve renewable integration, reduce curtailment and fossil-fuel backup generation, and deliver transparent decision-support tools for utilities, municipalities, and community-energy organizations. By linking energy management to public-health outcomes, the framework will reduce pollution-related respiratory and cardiovascular risks in overburdened communities. Open-source software, reproducible datasets, and practitioner guidance will accelerate adopti