KBC (a Yokogawa company) has just produced a 53-page ‘Value chain optimization manifesto’ that encourages operators to ‘digitalize with purpose’. The manifesto is based on KBC/Yokogawa’s experience of working on the world’s largest organizations in energy and chemicals. The Manifesto spans upstream oil and gas, LNG, refining and petrochemicals.
The Manifesto advocates a holistic approach to optimization that considers complete value chains such as that extending from the reservoir, through the gathering system, top-side processing constraints and on to market demand. Process improvement dates back to (at least) 1911 when Fredrick Taylor, wrote ‘In the past the man has been first; in the future the system must be first’. This led to operations research into logistics and the use of mechanization to improve labor-intensive processes. Computers and ERP systems came to the fore in the second half of the 20th Century, integrating siloed databases. Today, supply chain management is an ‘overarching operations management activity that dictates operations delivery performance’.
The Manifesto drills down into topics such as asset optimization through ‘molecule management’, a concept that spans reservoir engineering through refinery feedstock selection and yield optimization. 20th Century linear programming has given way to more accurate non-linear approaches, able to incorporate system-wide dynamics, although linear models ‘remain useful’.
The Manifesto’s goal is operational autonomy of the asset, achieved through increased plant automation and an evolution of the business model. ‘Achieving autonomous operations involves empowering the plant to run, learn, adapt and thrive in tomorrow’s environment; whatever that might be’. Enter the first principle-based Integrated Asset Model (IAM), operationalized with real-time data to create a digital twin around which ‘holistic, cross-functional convergence in understanding and action across organization silos can occur’. The IAM digital twin constitutes the heartbeat of the plant which drives all other applications.
Artificial intelligence observes asset behavior patterns and attempts to correlate these with positive or negative outcomes such as energy savings, yield improvements or machine failures. AI models are simple to use, fast to execute and do not require deep chemical, mechanical or electrical engineering knowledge. While some claim that AI provides similar capabilities to the simulators of the digital twin, their successes are limited to simple problems where a correlation-based model is good enough to represent reality. The main reason for the failure of AI is ‘too many false positives’. AI can succeed at the machine level, but fails at the machine-plus-processes level. Better results have been obtained using first principles-based approaches in tandem with AI. ‘Cognitive’ AI can be trained to interpret the results of the rigorous asset simulation model, ‘homing in on what the problem or opportunity might be and presenting the engineer with a triaged, reduced set of options to explore further’.
Plant knowledge management that connects information from a variety of sources is needed to provide context and enable interpretation and understanding. Achieving meaningful levels of knowledge is a challenge for current methods. Semantic web technologies such as knowledge graphs show promise and allow for entity pairing. For example, between assets (refinery, oil field, chemical plant), operating conditions, feed type, product slate, shift, etc. utilizing data from a variety of underlying sources.
Industry is still largely dependent human-based decision making and associated tacit knowledge of experienced operators. However, demographics are changing as top-flight, experienced engineers approach retirement and fewer petroleum engineers are graduating. The net loss of tacit knowledge needs to be managed and the ‘rules of thumb’ need to be codified, validated and institutionalized.
© Oil IT Journal - all rights reserved.