Drilling boreholes through hydrogen sulfide (H2S) bearing formations such as Umm Er Radhuma and Tayarat Formations poses a critical challenge for the oil and gas industry in southern Iraq. In this era of increased concern for personal safety and environmental factors, the industry needs additional tools and methods for handling this deadly and corrosive gas. This paper describes how actual formation field data is entered into unsupervised learning software to train a H2S monitoring simulator to improve the drilling operators’ ability to detect H2S, analyse and then adjust mud pH to neutralize free H2S present in drilling fluids.
The H2S concentration measurements for five drilled wells in the Umm Er Radhuma and Tayarat fields in the South of Iraq used in a previous study are extended to include data from an additional 12 wells with 22 H2S events. This paper describes a real-time H2S simulator which consists of a computer interface kit implementing MATLAB code. The kit is provided with three LEDs (green, yellow, and red) representing different alert levels. The system monitors changes in H2S levels and triggers alarms depending on the levels. The setup of the alert levels for the simulator are adopted from both Health Safety Environment (HSE) policy and unsupervised vector quantization via Fuzzy Adaptive Resonance Theory (ART) neural network category ranges.
Results show that using Fuzzy ART to train the H2S simulator alert levels based on formation data leads to a smarter and more sophisticated alert system compared to the typical HSE set of alerts. Specifically, the Fuzzy ART derived alerts enable earlier detection of H2S events, faster response to changing H2S levels, and it computes workers’ exposure over time to H2S, to prevent excessive accumulation of H2S in the respiratory system.
The system demonstrates a smarter and more robust method for reducing risks to drilling personnel, rig equipment, and the environment while drilling in areas with H2S hazards.