Revolutionizing Drilling Through Natural Fractures: Leveraging Causal Artificial Intelligence (AI) and Real-Time Feed Zone Monitoring

Abstract Geothermal operations face drilling challenges like stuck-pipe incidents, posing significant project risks. In this paper, we introduce a real-time monitoring system for early detection of natural fractures that correspond to permeable feed — subsurface areas where geothermal fluids enter the wellbore from surrounding rock formations. This system is integrated with the Stuck-Pipe Risk Advisor (SPRA), a novel tool we created that enhances drilling efficiency through causal artificial intelligence (AI) and semantic web technologies, aimed at revolutionizing geothermal drilling practices. Our approach integrates the real-time feed zone monitoring system with SPRA, to identify key risk factors for stuck-pipe incidents. The time-series analysis of drilling parameters like rate of penetration (ROP) and standpipe pressure (SPP) predicts the occurrence of feed zones, which triggers the mechanism to run the explainable decision support model and understand the causal relationships among the different formation and well properties in the proximity of feed zones. The system's use of semantic web technologies enhances transparency and aids in the comprehension of the predictive outputs, while a flow-diagram interface provides real-time visualizations of risk factors and potential interventions. Field application of this integrated workflow between the real-time monitoring system and SPRA, has demonstrated accurate prediction on the onset of feed zones, a key component in decreasing stuck pipe incidents. Observations confirm that proactive time series analysis of ROP and SPP across seven wells in three fields showed a 97% accuracy in early detection of feed zones that triggered the mechanism to predict risks of stuck pipe from the SPRA. Actionable insights based on a root-cause analysis diagram for drilling operators, provide real-time visualizations of risk factors, potential mitigations, and interactive explanations, empowering the wellsite team to reduce the frequency of stuck-pipe incidents, enhancing drilling efficiency, and contributing to the overall safety and success of geothermal drilling projects. The semantic annotations add further details to the causal links and risk factors involved, enhancing operational decision-making, and enabling continuous improvement for safe drilling practices. This integrated approach has been validated and endorsed by domain experts (drilling engineers and geologists), showing significant advancements in risk management and drilling efficiency. We present a pioneering neurosymbolic integration of causal AI with time-series analysis (neural) and semantic web technologies (symbolic) in geothermal drilling operations, marking a first in the industry. The SPRA's capability to predictively assess feed zones and preempt stuck pipe using real-time data analytics offers significant advancements in operational safety and efficiency. This innovative methodology not only improves hazard prevention but also provides engineers with actionable and understandable insights that enhance both immediate and long-term drilling operations success.