This technology proposes an optimization scheme based on deep reinforcement learning, targeting HVAC systems that account for 40% of household energy consumption in smart grids. It adopts a cloud-d parameter + local neural network online learning architecture, eliminating reliance on building thermodynamic models or prior information, and coordinating energy storage and loads in real-time to reduce energy costs while ensuring user thermal comfort. The scheme adapts to environmental changes and has broad application prospects.

Technology provider:Nanjing University of Posts and Telecommunications