A Computational Intelligence Framework for Real-Time Optimisation of Gas Processing Plant Operations in the Niger Delta

Real-time process optimisation in gas processing plants offers substantial value through improved NGL recovery, reduced fuel gas consumption, and enhanced product specification compliance, yet the application of computational intelligence methods to Niger Delta plant conditions presents distinctive challenges not addressed by existing frameworks. This study develops and validates a computational intelligence framework for real-time optimisation of gas processing plant operations in the Niger Delta, combining artificial neural network (ANN) process models, model predictive control (MPC), and multi-objective evolutionary optimisation algorithms. The framework development is grounded in an 18-month operational data collection programme at a gas processing facility in Bayelsa State, encompassing 12,000 hours of process sensor readings across the inlet separation, dehydration, NGL extraction, and recompression sections. An ensemble of ANN process models is trained and validated for each major plant section, achieving prediction accuracy within 2 percent of measured process responses across the operational envelope. The MPC framework is implemented using a receding horizon optimisation approach, with the NSGA-III multi-objective evolutionary algorithm used to simultaneously optimise three competing objectives: NGL liquid recovery maximisation, fuel gas consumption minimisation, and product specification compliance constraint satisfaction. The study's original contributions include: a Nigeria-specific process disturbance taxonomy for Niger Delta gas plant feedstock variability, a novel ANN ensemble uncertainty propagation method for process constraint confidence intervals, and a computational load management scheme enabling real-time execution of the framework on standard industrial control hardware. Simulation results demonstrate potential annual improvements in NGL recovery yield of 6 to 9 percent and fuel gas savings of 11 percent compared to conventional operator-controlled baseline. Recommendations include a phased field deployment programme and collaborative industry-university research agreement for continuous framework improvement. Keywords: computational intelligence, gas processing optimisation, ANN, model predictive control, Niger Delta.