Comparing transient and steady-state methods for the thermal conductivity characterization of a borehole heat exchanger field in Bergen, Norway

A comparative study was carried out aiming at characterizing the thermal conductivity of rocks sampled in a borehole heat exchanger field. Twenty-three samples were analysed with four different methods based on both steady-state and transient approaches: transient divided bar (TDB), transient line source (TLS), optical scanning (OS), and guarded hot plate (GHP). Moreover, mineral composition (from XRD analyses), P-wave velocity, and density were investigated to assess the petro-physical heterogeneity and to investigate possible causes of divergence between the methods. The results of thermal conductivity showed that TLS systematically underestimates thermal conductivity on rock samples by 10–30% compared to the other devices. The differences between TDB and OS, and GHP and OS are smaller (about 6% and 10%, respectively). The average deviation between TDB and GHP, for which the specimen preparation and the measurement procedure were similar, is about 10%. In general, the differences are ascribable to sample preparation, heterogeneity and anisotropy of the rocks, and contact thermal resistance, rather than the intrinsic accuracy of the device. In case of good-quality and homogeneous samples, uncertainty can be as low as 5%, but, due to the above-mentioned factors, usually uncertainty is as large as 10%. Opposite relationships between thermal conductivity and P-wave velocity were observed when analysing parallel and perpendicular to the main rock foliation. Perpendicular conductivity values grow with increasing perpendicular sonic velocity, while parallel values exhibit an inverse trend. Thermal conductivity also appears to be inversely correlated to density. In quartz-rich samples, high thermal conductivity and low density were observed. In samples with calcite or other likely dense mineral phases, we noticed that lower thermal conductivity corresponds to higher density. The presence of micas is likely to mask major differences between silicate and carbonate samples.