Modeling and Optimization of NOX Emission in a Coal-fired Power Plant using Advanced Machine Learning Methods

A new methodology combining the advanced extreme learning machine (ELM) and harmony search (HS) was proposed to model and optimize the operational parameters of the boiler for the control of NOX emissions in a 700 MW pulverized coal-fired power plant. About five days’ worth of real data were obtained from supervisory information system (SIS) of the power plant to build the ELM NOX model. HS was employed to optimize the operational parameters of the boiler to minimize NOX emissions based on the prediction of NOX by ELM. Compared with the widely used learning method such as ANN and SVR, ELM performed better both in accuracy and computing time for the modeling of NOX emission. The proposed comprehensive methodology can provide desired and feasible optimal solutions within one second, which is acceptable for the online optimization.