Convolutional neural network approach for reduction of nitrogen oxides emissions from pulverized coal-fired boiler in a power plant for sustainable environment

•The proposed model reduced the NOx emissions from a coal-fired power plant by 50.9%.•The proposed model may help power plants achieve environmental sustainability goals.•1d-CNN outperformed with least testing RMSE compared to ANN and LSTM.•1d-CNN was found computationally inexpensive compared to ANN and LSTM.•TLSO reached the global minimum relatively quickly. Coal-fired power plants are the main electric power source across many countries and cause major air pollution problems such as acid rain, smog, ozone depletion, and global warming. According to the best of the authors' knowledge, this is by far the first study that proposed 1-dimensional Convolutional Neural Network (1d-CNN) in combination with teaching learning self-study optimization (TLSO) algorithm for NOx emissions reduction by optimizing process input variables in a pulverized coal-fired power plant. The proposed model reduced the NOx emissions by 50.9%. In addition, the reduction experiment resulted in the early convergence superiority of the TSLO (130 s, 30th iteration) compared to genetic algorithm and Bayesian optimization. Based on the result, it is evident that combination of computationally inexpensive 1d-CNN and relatively fast converging TLSO could help process engineers reduce NOx emissions, which could ultimately contribute towards the goal of a sustainable environment. [Display omitted]