Data-driven model-based flow measurement uncertainty quantification for building central cooling systems using a probabilistic approach

Uncertainties inevitably exist in measurements and may lead to biases in making management and control decisions, and thus affect the energy performance of building central cooling systems. Water flow meters are essential for the monitoring and operational control of building central cooling systems, but they often suffer from significant measurement uncertainties due to site constraints and unfavorable working environment. An effective method to quantify the flow measurement uncertainties is urgently needed. This study proposes a data-driven model-based flow measurement uncertainty quantification strategy using Bayesian inference and Markov chain Monte Carlo sampling methods. The proposed strategy is tested and validated systematically on an air-cooled chiller. Four case studies with different levels of flow measurement uncertainties are conducted. The test results show that both systematic and random uncertainties of flow measurements are quantified accurately by this strategy. The 95% Bayesian credible intervals of systematic and random uncertainties contain their pre-set (actual) values, and their posterior means (estimated values) are very close to their pre-set values. The relative errors in quantifying flow measurement uncertainties are within 10%. The performance of the proposed method is quite satisfactory. This study provides a cost-effective and promising alternative for on-site flow meter calibration in engineering practice.

Automated summary

Uncertainties in measurements can affect the energy performance of building central cooling systems. This study proposes a data-driven model-based strategy to quantify flow measurement uncertainties using Bayesian inference and Monte Carlo sampling methods. The strategy is tested on an air-cooled chiller and accurately quantifies both systematic and random uncertainties, with relative errors within 10%.