Feasibility assessment of LT-ATES system: accounting for groundwater natural flow and uncertainty in rock attributes

A variety of actions have been recently taken, and pledges have been made worldwide to alleviate the catastrophic consequences of the global warming. In this context, underground thermal energy storage has stepped forward as a potential solution to support the pledges by supplying heating and cooling demand – activities that currently depend heavily on fossil fuels – which consequently contributes to the reduction of the greenhouse gases emissions. In this study, we will numerically explore the feasibility of installing a Low Temperature Aquifer Thermal Energy Storage (LT-ATES) system in the Bicocca zone of the Milan Metropolitan area in Italy. The aquifer in the studied area is characterized by a shallow depth which makes it convenient for low temperature geothermal applications. A main challenge addressed in this work is the presence of a relatively effective natural flow. In addition, the heterogeneity of the subsurface rock properties and the scarcity of the related observation data imposed the problem of the uncertainty, which will be addressed here by using geostatistical techniques: i.e., the Ordinary Kriging (OK) method for estimation of geological properties at unsampled locations, and the Monte Carlo (MC) technique for uncertainty assessments. A homogeneous field model was also characterized and used as a basis for comparison purposes. We investigated four different scenarios of five-spot multiple-well placement and arrangement: (i) single doublet wells with central cold well (S1), (ii) single doublet wells with central hot well (S2), (iii) double doublet wells with central cold well (S3), and (iv) double doublet wells with central hot wells (S4). The optimal well placement distances were evaluated taking into account the hydraulic gradient to avoid the negative thermal interference between warm water and cold water. Every simulation scenario was run over a period of 5 years with a seasonal alternating of injection/production cycle: i.e., (i) warm water is injected in summer over a period of 4 months extending from May until September, while cold water is extracted for cooling purposes during, and (ii) a period of cold water injection and hot water extraction for heating purposes during the winter season from November until March. Those two seasons are separated by resting periods (of two months) where all wells are switched off. Performance evaluation of the four different scenarios was based on the calculation of the energy recovery considering the homogeneous characterization of the field: S3 outperformed the rest of the arrangements for heat recovery (85%) and S4 showed better performance regarding the cold water extraction (70%). However, S1 was the scenario of our choice considering two points: (i) the study area is characterized by higher heating degree days (HDD), consequently higher number of warm wells is favorable, and (ii) doubling the number of wells would contribute to only 7% increase in the heat recovery. An assessment of the uncertainty associated with our analysis is reported for the selected simulation scenario, S1. The results show some discrepancies in the simulation responses between the median values of MC simulation ensembles and the homogeneous field simulation responses. Indeed, the latter falls within the confidence range of 5%-95% of the MC simulation results. Overall, a reduction in the recovery factor for both warm and cold wells was observed ranging between 5% and 32% for all scenarios when employing a stochastic model instead of the homogeneous deterministic approach. Finally, a preliminary surface facility model of the ATES network — integrated with a heat pump — was developed to assess the feasibility of extracting stored thermal energy for real-world heating applications and to evaluate well performance. Based on the median hot well responses from the Monte Carlo simulations, the system is capable of continuously supplying heat to 28 office buildings during the winter season. In comparison, the lowest-performing realization supports 31% fewer buildings, while the highest-performing realization supports 57% more, highlighting the significant impact of subsurface heterogeneity on system performance.