Enhancing the smart readiness of buildings: Combining Collective intelligence and Reinforcement learning in Building Energy Management

Abstract

This research introduces a novel Energy Management approach, named CIRLEM, aiming to enhance the smartness of buildings by focusing on technical systems operations, environmental variations, and occupants' needs. Deployed in a simulated environment using Building Performance Simulation and Python integration, the study evaluates CIRLEM's performance under future extreme cold weather scenarios, employing a set of representative climate data. The pilot case, two building blocks in Sweden, undergoes assessment for energy demand, peak power, and thermal comfort. Results indicate that CIRLEM, particularly when driven by demand and price signals, effectively reduces energy demand and costs, demonstrating strong adaptability to extreme weather conditions. Thermal comfort is maintained regarding the temperature limits and variations. Ongoing developments attempt to refine the reward function and signal generation for thermal comfort enhancement and real-world implementation.