Uncertainty assessment of the hydraulics properties surrounding a standing column well with a thermal response test

Abstract

The standing column well (SCW) is known for being a highly efficient ground heat exchanger as it relies on both conduction and advection heat transfer processes. Therefore, the interpretation of a thermal response test (TRT) is strongly influenced both by the hydraulic and thermal properties surrounding the SCW. In this study, it is shown that a TRT can allow identifying the thermal and hydraulic properties around a SCW. The analysis is conducted in a Bayesian framework allowing an accurate and robust identification of the hydraulic properties and their uncertainties. A closed-form expression of the likelihood is used to consider the autocorrelation of the residuals between observed and simulated temperatures. A coupled numerical model is used to generate a training database for an artificial neural network. Then, the latter serves as an emulator of the SCW's short-term g-function given various input parameters. A case study is presented based on a 100-hour TRT performed on a SCW built at a demonstration site located in the city of Mirabel, Canada. For the specific site studied, hydraulic properties were identified with an uncertainty of less than 30 % at a two-sigma level. Such important results lead to more appropriate and efficient design of SCWs.