Building Thermal Model Calibration for Energy-Comfort Optimization

The patent creates a digital twin (copy) of the building that replicates its thermal behavior and energy consumption patterns. This digital model is calibrated using measured data from sensors and then used for optimization simulations, avoiding the need for complex physical measurements while maintaining accuracy. The digital copy allows virtual testing of different energy management strategies without interfering with the actual building operation.

The patent introduces a calibration phase as an intermediary step between data collection and optimization. During this phase, the digital model is adjusted to match actual building behavior using measured data. This intermediary calibration process ensures the digital twin accurately represents the physical building without requiring direct complex measurements during the optimization phase, thus reducing overall system complexity.

The patent transforms the statistical model into a physics-based thermal model by changing the underlying parameters from purely statistical correlations to physically meaningful parameters (thermal conductivity, heat capacity, insulation properties). This allows the model to maintain computational efficiency while improving prediction accuracy, as the physics-based parameters provide causal explanations for energy consumption patterns rather than just statistical associations.

The patent creates a simplified digital representation that copies only the essential thermal behavior of the building without replicating all physical details. This simplified model captures the dominant thermal dynamics sufficient for energy optimization while maintaining fast calculation speeds. The model includes key thermal zones and heat transfer paths necessary for accurate prediction without the computational burden of detailed building physics simulations.

The patent implements a calibrated digital twin that uses only the essential data and model components necessary for reliable energy optimization, rather than requiring complete and precise data on all building systems. The calibration process adjusts the model parameters to match actual building behavior, compensating for incomplete or imprecise input data. This partial approach focuses on the critical thermal zones and heat transfer paths that dominate energy consumption, achieving reliable optimization without the burden of comprehensive data collection.

The system performs self-calibration by automatically adjusting the digital model parameters to match measured building behavior during an initial phase. This self-service calibration reduces the need for manual data collection and model parameter specification by building experts. The system autonomously identifies the relationship between control actions and thermal responses, making the implementation process more accessible to building operators without specialized expertise.

The patent segments the building into discrete thermal zones and represents the optimization problem as a set of independent or weakly coupled sub-problems. Each thermal zone can be optimized separately or in small groups, reducing the overall computational burden. The NSGA-II algorithm is applied to these segmented zones rather than treating the entire building as a single complex system, significantly reducing calculation time while maintaining optimization effectiveness.

The optimization is performed periodically rather than continuously, using the calibrated digital twin to evaluate multiple scenarios at discrete time points. The system uses the thermal inertia of buildings to maintain comfort between optimization cycles, allowing computationally intensive multi-objective optimization to be performed at manageable intervals rather than in real-time, thus reducing total computational resources required.