Refining India March 2026 Issue

Input feed

Constraints Operating parameters – temperature, pressure etc. Equipment duties

Optimiser

Individual feedstock rate Individual feed composition Any recycle rate Recycle composition

Iteration i th

Objective function

Plant model output of products/ inputs to AI objective function

Operating parameters Operating parameters – temperature, pressure etc. No. of furnaces online at any point

Prices and costs

User inputs

Optimal output

Market prices of products Cost of inputs

Operating data, LIMS data

Optimal operating conditions and optimised objective Desired product mix Operating parameters – temperature,

APC/MPC

Predictive model

pressure, CV % open etc. Input feed mix volumes Fuel gas consumption

Oine until optimiser

Minimisation of error

Figure 4 A case of live RTO example

validated with real-time and laboratory inputs. It uses predictive plant models and an AI-based optimiser to run iterative scenarios, searching for optimal conditions under varying constraints such as temperature, pressure, and equipment limits. • Input handling: Operational data from distributed control systems (DCS), laboratory information management systems (LIMS), and manual user entries are validated and fed into the plant model. • Hybrid modelling: The plant model uses both first-principles physics and AI-driven data science to simulate product outputs under different operating scenarios. • AI optimisation loop: The AI optimiser evaluates the plant model’s outputs and seeks configurations that maximise profits or yields, considering current market prices and costs, while satisfying constraints. • Automatic setpoint recommendation: After finding optimal targets (for example, product mix, temperatures, pressures, feed rates), the system sends these recommendations to advanced process control (APC/MPC) systems for real-time implementation, continuously minimising error between target and plant operations. AI-driven RTO delivers key business outcomes:

• Profit maximisation: Fine-tuned process control unlocks more value from existing assets without major capital expenditure. • Yield improvement: Optimised product mix and process parameters enable consistently higher yields from varied feedstock. • Energy savings: Reduced waste and smarter energy usage lower the operational cost and carbon footprint. • Bottleneck identification: Continuous optimisation highlights process bottlenecks, supporting debottlenecking strategies and cost- effective expansion decisions. Figure 5 reflects not just the value realised in terms of economic benefits for a $3 billion petrochemical major in India. This use case is also directly relevant to the value of AI adoption in the petrochemical context, demonstrating how advanced optimisation blends engineering expertise and AI to drive measurable operational and financial benefits. Current state of AI adoption in global oil and gas industry While this case of RTO implementation

exemplifies what is possible, such comprehensive, enterprise-scale AI

deployments remain the exception rather than the norm across the global oil and gas industry.

Refining India