Ensemble Calibration of lumped parameter retrofit building models using Particle Swarm Optimization

• Published in: Energy and buildings Vol. 155; pp. 513 - 532
• Main Authors: Andrade-Cabrera, Carlos, Burke, Daniel, Turner, William J.N., Finn, Donal P.
• Format: Journal Article
• Language: English
• Published: Lausanne Elsevier B.V 15.11.2017; Elsevier BV
• Subjects: Building model calibration; Building retrofit; Buildings; Calibration; Computer simulation; Electricity; Electricity generation; Energy conservation; Energy modeling; Functions (mathematics); Heating; Heating load; Integrated analysis; Lumped parameter models; Mathematical models; Optimization; Particle Swarm Optimization; Residential buildings; Retrofitting; Building retrofit; Integrated analysis; Lumped parameter models; Building model calibration; Particle Swarm Optimization
• ISSN: 0378-7788; 1872-6178
• DOI: 10.1016/j.enbuild.2017.09.035

•A methodology to calibrate a number of building energy models is proposed.•Each model represents a combination of internal, external and ceiling insulation.•The building models are modelled as parametric functions of insulation thickness.•All building models are accurately calibrated (Mean Average Error below 0.5°C). Simulation-based building retrofit analysis tools and electricity grid expansion planning tools are not readily compatible. Their integration is required for the combined study of building retrofit measures and electrified heating technologies using low-carbon electricity generation. The direct coupling of these modelling frameworks requires the explicit mathematical representation of Energy Conservation Measures (ECMs) in building-to-grid energy system models. The current paper introduces an automated calibration methodology which describes retrofitted buildings as parametric functions of ECMs. The buildings are represented using a lumped parameter modelling framework. A baseline model, representative of the building prior to retrofit, and the retrofit functions are calibrated using Particle Swarm Optimization. Synthetic temperature and heating load time-series data were generated using an EnergyPlus semi-detached house archetype model. The model is representative of this residential building category in Ireland. It is shown that the proposed methodology calibrates retrofitted building models to an acceptable level of accuracy (MAE below 0.5°C). The methodologies introduced in the current paper are capable of generating lumped parameter building models with similar dynamics for different ECMs for any archetype building energy model. The identified building retrofit models have the potential to be integrated with electricity grid models in a computationally-efficient manner.