This study investigates the optimization of a co-generation system involving multiple steam boilers and turbines, aiming to minimize CO2 emissions and energy consumption while maintaining reliable energy delivery. A hybrid Genetic Algorithm (GA) and Sequential Quadratic Programming (SQP) method is implemented within a Multi-Objective Mixed-Integer Nonlinear Programming (MOO-MINLP) framework. The approach effectively captures the nonlinear behavior of efficiency and operational constraints. The results show a reduction of up to 10 % in CO2 emissions and over 35 % in energy savings compared to GA-only approaches. Maximizing biomass usage at Extreme Point A achieves the lowest emissions (554.29 kg) and an energy cost of 4253.69 GJ, while minimizing energy consumption at Extreme Point C leads to 3532.67 GJ but higher emissions (708.86 tons). This study demonstrates the hybrid GA-SQP method's potential to optimize both CO2 emissions and energy consumption, offering decision-makers a balanced approach between cost and environmental impact. The results underscore the significance of fuel allocation, especially biomass, in reducing emissions despite lower efficiency, presenting a cost-effective and sustainable solution for co-generation system optimization.

Load optimisation within the cogeneration system is crucial in enhancing energy efficiency. Instead of constructing the mathematical optimisation model or applying the commercial utility optimisation software with a licensing fee, this study proposes a holistic P-graph method to model and optimise the cogeneration system using the free and user-friendly software, P-graph Studio. To consider actual performance of unit operations, novel slope-constant element is introduced in the P-graph structure to adapt the variable output-input ratios in the form of linear performance model with non-zero constant. This overcomes the functionality of the conventional P-graph structure that only considers fixed output-input ratio. A case study of an industrial cogeneration system is optimised using the proposed P-graph method, resulting in 1.24 % reduction of operating cost and CO2 emission: equivalent to savings of RM 12,822,300/year and 4,300 tonnes CO2 emission/year. Two operating strategies are proposed to revise the optimal operating method by modifying the P-graph superstructure to ensure adequacy of the utility margin in meeting the potential maximum utility demand. The operating cost saving of 0.53 % is achieved after revision to meet both operational efficiency and reliability of the cogeneration system which results in savings of RM 5,454,900/year and 1,800 tonnes CO2 emission/year.

This paper examines Real Time Optimization (RTO) for an industrial cogeneration plant featuring a tightly coupled multi boiler turbine network, in which fluctuating steam and power demands and fuel price volatility necessitate continual economic re optimization while preserving closed loop stability. Three evolutionary optimizers are Differential Evolution (DE), Hybrid Differential Evolution (HDE), and Adaptive Differential Evolution (ADE) deployed as the supervisory RTO layer above the regulatory controllers, with Model Predictive Control (MPC) regulating boiler pressure (Control Variable 1, CV1) and drum level CV2 and PI or PI loops regulating turbine power. A deterministic, repeatable stress test is introduced through sequential step changes in high pressure steam demand, medium pressure steam demand, power demand, and natural gas price, enabling systematic evaluation of transient adaptability and robustness. Over five boilers and the turbine network, multi run mean and deviation results show that ADE delivers the most consistent overall behavior, yielding smoother operating trajectories, improved tracking, and lower energy usage. Specifically, the total integrated energy consumption is approximately 895 MWh with ADE, compared to 926 MWh with DE and 1259 MWh with HDE, equivalent to reductions of about 3 percent versus DE and 29 percent versus HDE. Control performance improves in parallel the mean boiler pressure (Integral Square Error) ISE CV1 drops by roughly 68 percent relative to DE and 71 percent relative to HDE, while turbine regulation shows substantial enhancement with turbine ISE reduced by about 98 percent compared with DE. Overall, the results demonstrate that adaptive evolutionary optimization strengthens coordination between the RTO and control layers, providing a robust and energy efficient strategy for real time cogeneration operation under dynamic demand and price disturbances.