Data Driven Modelling and Optimization of Efficiency and NOx Emission Rate of Fired Heaters in an Oil Refinery

Fired heaters are energy intensive units in petroleum refineries. For the economical operations,energy conservation is of prime importance that bring focus to combustion in fired heaters asa major energy consumption process. Any improvement in the heater efficiency can result inlarge reductions in energy consumption, emissions, and associated costs. The aim of thisresearch is to develop a practical computational tool that can predict and optimize the thermalefficiency and NOx emission rate of fired heaters, based on artificial Intelligence application.An Artificial neural networks (ANNs) and genetic algorithms (GAs) model has been developedand used based historical operational data to predict and optimize efficiency and emissionsfrom a fired-heater unit in petroleum refinery., Thermal efficiency was predicted and optimizedas a function of three operating parameters: percentage excess oxygen, heater draft and theenthalpy change (energy transferred) to the process fluid . The model accuracy was found tobe acceptable on the data collected from the five-year operation of a fired heater. Theeffectiveness of the optimization method was demonstrated through trial runs practically fordifferent seasons. In the second part of the study, extensive operational data with emissionswere collected to predict the thermal efficiency and NOx emissions using the earlier adoptedconcept of ANN. The network configured using four operating parameters: excess Percentageoxygen, heater draft, the enthalpy change of the process fluid and fuel gas to predict two outputs(efficiency and NOx emission rate). Multi Objective Genetic Algorithm (MOGA) was used tooptimize both outputs. The models accuracy were found to be acceptable in terms of RMSEand showed good prediction of the outputs (efficiency and emission)