On data-driven soft sensor of NOx emission in power station boiler

To predict the NO_x emission level precisely in power station boiler, a data-driven method is presented to solve the problem of the absence of precision model. In this method, the operating data is utilized sufficiently to establish the model based on statistical learning theory. Firstly, the data on field is cleaned to avoid noise and abnormal value. To find the optimal model parameters, genetic algorithm is used for model optimization. So a modeling method is presented to describe the NO_x emission level during varying load. The simulation is implemented on a 300MW coal-fired unit with several different working loads, and compared to the way based on neural network, the results show that the model can predict the NO_x emission more precisely, which provides the foundation for further operating optimization.