To optimize gas flaring in Kirkuk refinery in various seasons via artificial intelligence techniques

Unavoidable flaring in downstream oil industry causes pollutant emission in large amounts which is potentially harmful to nearby cities or farms. Hence one must manage exhaust toxic gases to raise enough in atmosphere or redirect from such places. Since Kirkuk refinery in north Iraq is next-door to agricultural farms on west yet to residential areas on east optimizing its layout for flare stacks is something acute. In this work we wrote codes in MATLAB software to simulate incomplete rather than complete oxidation as well as pollutant generation reactions. Then we made use of FLEUENT software to simulate pollutant propagation in Kirkuk oil purifier complex yet also farther to city as well as farms with respect to seasonal air currents on lowest troposphere layer. Finally, we set neural network approach to train on simulation data thereafter to unify outcomes to turn into a fast technique for layout optimization. Results show that optimization process efficiency relies on air current velocities as well as its direction. At intermediate air flow rates optimum layout includes only a selective portion of existent flare stacks. Outcomes also illustrate that heuristic techniques that have stronger local search such as particle swarm or artificial immune system can improve flare layout in seasons with intermediate air currents here summer plus early months in autumn while approaches with weak local search like Monte Carlo are more appropriate in winter for which we have no or low air flows in Kirkuk governorate.