Artificial Intelligence-Based Emission Reduction Strategy for Limestone Forced Oxidation Flue Gas Desulfurization System

Abstract

The emissions from coal power plants have serious implication on the environment protection, and there is an increasing effort around the globe to control these emissions by the flue gas cleaning technologies. This research was carried out on the limestone forced oxidation (LSFO) flue gas desulfurization (FGD) system installed at the 2*660 MW supercritical coal-fired power plant. Nine input variables of the FGD system: pH, inlet sulfur dioxide (SO2), inlet temperature, inlet nitrogen oxide (NOx), inlet O2, oxidation air, absorber slurry density, inlet humidity, and inlet dust were used for the development of effective neural network process models for a comprehensive emission analysis constituting outlet SO2, outlet Hg, outlet NOx, and outlet dust emissions from the LSFO FGD system. Monte Carlo experiments were conducted on the artificial neural network process models to investigate the relationships between the input control variables and output variables. Accordingly, optimum operating ranges of all input control variables were recommended. Operating the LSFO FGD system under optimum conditions, nearly 35% and 24% reduction in SO2 emissions are possible at inlet SO2 values of 1500 mg/m3 and 1800 mg/m3, respectively, as compared to general operating conditions. Similarly, nearly 42% and 28% reduction in Hg emissions are possible at inlet SO2 values of 1500 mg/m3 and 1800 mg/m3, respectively, as compared to general operating conditions. The findings are useful for minimizing the emissions from coal power plants and the development of optimum operating strategies for the LSFO FGD system.

Optimization of a 660 MWe Supercritical Power Plant Performance—A Case of Industry 4.0 in the Data-Driven Operational Management Part 1. Thermal Efficiency

“…It has been extensively reported in the literature that mining causal relationships out of such data is beyond the capability of any type of multi-variate regression technique. AI-based data analytic techniques perform significantly better for modeling such scenarios [39,58]. The sensors' location for measuring the power plant's different operating parameters is shown with numbers in Figure 2.…”

“…The algorithm is computationally time-consuming to achieve the optimum results under the influence of many decision variables and the big volume of the process's operation data [35]. However, AI process modeling techniques like ANN and LSSVM have proved to be computationally inexpensive and efficient to model the complex process and find the optimal solution of the objective function [38][39][40].…”

Optimization of a 660 MWe Supercritical Power Plant Performance—A Case of Industry 4.0 in the Data-Driven Operational Management. Part 2. Power Generation

“…SOFM is an unsupervised learning machine that maps the underlying possible statistical features of the high-dimensional input space data on the nodes of a two-dimensional square lattice. The square lattice's length is equal to the input space dimensions [2,36]. SOFM has an excellent ability to distribute the input space data on the nodes in homogenous groups and is used in many real-life applications [37][38][39].…”

“…The feedforward backpropagation network algorithm is used in this work. Gradient descent with momentum is employed as a training function, tangent hyperbolic and purelin is employed as transfer functions at the hidden and output layer of MLP, respectively [2,44]. ANN training is carried out until one of the two stopping criteria is met, i.e., either a 0.0000001 change in convergence error or a maximum number of epochs is reached [2,49].…”

“…Over the past few decades, researchers and energy experts are closely collaborating in a wide range of research activities to mitigate the harmful impacts of power generation to ensure sustainable power generation and environmental protection. [2][3][4][5][6][7][8][9][10][11][12][13][14][15][16].…”