Artificial Intelligence and Machine Learning Used as an Enabler for Dynamic Risk Management

Abstract Applying big data, data science, business process automation (BPA) and domain expertise to operational and project risk in the upstream O&G space, will create a paradigm when applied to wellbore construction. A scalable, dynamic risk ecosystem enables seamless integration of risk into all aspects of the well construction process, a cornerstone of this approach is interoperability at a system level. A lot of risk management is subjective, risk registers and mitigations are generated by workshopping with SME's. Risk scoring is often performed in a similar manner. The financial sector now applies data science techniques, for example in the fields of auditing and compliance. The aim of this paper is to discuss how these new techniques are integrated into a well-established existing risk management processes. The starting point for digitizing the process is data. A ‘Risk Ecosystem’ was developed with a risk engine at its core linked to curated input data and providing outputs through dedicated GUI's and direct links into offline planning and real-time operational software. The risk management process has become dynamic and integrated through the value chain. A major change is the integration with the real-time decision making process. Risk assessments combine analysis by hazard, risk, probability and barriers. When applying risk dynamically to well construction operations the barriers can be data, analysis, procedural and/or physical. The interplay of these barriers is the mitigation. In the risk engine barriers can be managed by the probability of effectiveness and the uncertainty of the input and output data. Artificial intelligence (AI) and machine learning (ML) can now enable risk and mitigation identification, supporting probability and uncertainty to be analysed in real-time to provide dynamic solutions. Data connectivity enables dynamic risk management where the risk updates when new data is available or an analysis is improved. The probability and uncertainty of analysis supports risk ranking. Dynamic risk management in planning leads through to risk management during operations, as risk changes and updates the source data can be revisited efficiently as it is connected through hazard, risk or mitigation. AI is triggered, by a change in the data, and the improved risk information is distributed through a revised mitigation scenario. The integration of ML recommends mitigations trained to global experience and continues to improve as SME's qualify mitigations. Exploration of the interaction of data management, AI, ML and personnel for risk management is important, achieving the correct balance reduces risk, making risk management more efficient. We propose that the improvement in data science for hazard, risk and mitigation identification supports the drive towards a risk management standard of a minimum risk analysis for the industry. In time this will lead to operational efficiency and consistency of performance.