An innovative probabilistic approach to enhance confidence levels in energy performance projections of heterogeneous building stocks

Abstract

The EU Green deal stipulates that effective policy measures are required to significantly increase building renovation, which is one key area for achieving Europe’s decarbonisation target for 2050. The established cost-optimal method of the 2010 Energy Performance of Buildings Directive (EPBD) provides a harmonised framework for EU Member States (MS) to define Energy Performance (EP) benchmarks and energy efficiency measures that will best drive buildings to Nearly Zero Energy Building (NZEB) status. However, despite the positive push, literature has identified large EP gaps in the EPBD software and benchmark divergences between MS that highlight the limitations of this tool in devising successful policy measures. These shortcomings potentially stem from ’non-calibrated and deterministic’ Reference Buildings (RBs) characterised using only single and non-calibrated parameter values and which do not take into consideration the building parameters’ uncertainties and building stock diversities. This could result in a significant divergence between the cost optimal calculations and the real financially feasible determinants, especially for heterogeneous building stocks. This would ultimately lead to ineffective energy efficiency policy measures and a gradual loss of confidence in the methodology’s outcomes among prospective investors and energy consumers. This thesis has focused on proposing solutions to these limitations, through an innovative EPBD cost-optimal approach that integrates ’probabilistic Bayesian calibrated RBs’ into the current EPBD methodology. RB uncertain parameters are defined as prior distributions, and metered consumption data is utilised to calibrate the RBs model and reduce the uncertainties to narrower posterior distributions. The resulting calibrated RBs and the cost-optimal plots are then employed in an objective approach to define NZEB EP benchmarks according to four distinct levels of EP ambition. Ultimately, a probabilistic risk analysis that is propagated from the posterior parameter distributions is used to quantify the robust financial risk to reach each ambition level. The approach was optimised for heterogeneous building stocks via an innovative methodology to define RBs and by developing the ’reference zone’ concept. This concept replaces full-space models with reduced space energy models to improve the computational efficiency of calibration. The RB definition methodology was applied to a 5-star hotel building stock, followed by the validation of the proposed EPBD cost-optimal method using a derived hotel RB. For this RB, the ’reference zone’ approach successfully calibrated the model in compliance with ASHRAE [1, 2] Coefficient of the Variation of the Root Mean Square Error (CVRMSE) and Normalised Mean Bias Error (NMBE) metrics for monthly data and replicated the monthly electricity energy end-uses of the full model with a 4000 % improvement in simulation run time. A comparison of the current EPBD cost-optimal approach with the innovative approach for the hotel RB demonstrated that non-calibrated RBs can provide a large EP gap exceeding 30 %, which have resulted in a highly unrealistic financial feasibility and misleading EP improvement projections. Furthermore, a probabilistic risk analysis considering parameter uncertainty and diversity successfully uncovered the full associated financial risk associated with each EP ambition level and the required financial support to establish realistic benchmarks to trigger renovation. Therefore, this research provides tangible findings and insight for the eventual upgrading of the current EPBD cost-optimal approach to the proposed one to increase the chances for devising robust policy measures to meet the 2050 carbon-neutrality goals.