Polynomial chaos-based uncertainty quantification of the performance of a closed loop deep geothermal borehole

Geothermal energy has the potential to become a key technology in the transition away from fossil fuels. Deep borehole heat exchangers (DBHEs) are closed loop geothermal systems, and one benefit of a closed loop is that existing wells can be repurposed, reducing development costs. Although modelling of DBHEs has advanced in recent years, the effect of uncertainty in geological properties has not been widely explored, particularly when the borehole penetrates diverse rock strata. This paper uses polynomial chaos expansions to quantify the effect of thermogeological uncertainty on the performance of a closed loop deep geothermal borehole. The focus is on the Science Central borehole in Newcastle upon Tyne, UK, which was drilled 1820 m through a heterogenous sedimentary basin and is a candidate for repurposing as a DBHE. A recent semi-analytical model of a coaxial DBHE is extended to account for variable heat loads and combined with an energy demand model for a neighbouring building. The results show that the DBHE could support a 20-year constant heat load of between 132 (P90) and 154 kWth (P10) and over 90% of the variability in this long-term output is governed by rock thermal conductivity. Investigation of a year-long variable heat load revealed that while deeper formations generally control heat transfer, shallower formations grow in importance during times of lower heat demand and cooling. Using mean geological properties could be unconservative as a deterministic model with a minimum predicted temperature of 7.1 °C over one year had a non-negligible failure probability of order 1 0 − 5 . • The paper quantifies the effect of geological uncertainty on coaxial DBHE output. • A semi-analytical DBHE model is extended for constant and variable heat loads. • Polynomial chaos expansions are used for efficient uncertainty propagation. • Geological heterogeneity is shown to be a crucial factor in DBHE performance. • The use of mean properties may result in non-negligible failure probabilities.