Ligang Zheng

Prediction of NOx Concentration from Coal Combustion Using LS-SVR

Nitrogen oxide (NOx) is one of main pollutants emitted from coal fired power plants and is a significant pollutant source in the environment. Therefore, the monitoring or prediction of NOx emissions is an indispensable process in coal-fired power plant so as to control NOx emissions. In this paper, NOx emissions modeling for real-time operation and control of a 300MWe coal-fired power generation plant is studied. A least square support vector regression (LS-SVR) model was proposed to establish a non-linear model between the parameters of the boiler and the NOx emissions. The results show that the LS-SVR model predicted NOx emissions with good accuracy. LS-SVR model is much more accurate than the GRNN model previously reported by the authors. LS-SVR model will be a good alternative to a neural network based model which is commonly used to implement the predictive emission monitoring system (PEMS).

Support Vector Regression and Ant Colony Optimization for Combustion Performance of Boilers

Support vector regression (SVR) is a powerful tool for modeling the inputs and output(s) of complex and nonlinear systems. However, the control parameters are critical to the performance of SVR and also difficult to be selected. For actual applications in most cases, self-modeling of studied systems without any manual operation was needed. In this work, ant colony (ACO) optimization was developed to search the optimal control parameters so as to achieve this purpose. ACO is a meta-heuristic optimization algorithm for solving both discrete and continuous optimization problems. As a case study to demonstrate the applicability of the proposed method, SVR model is constructed for correlating historic data comprising values of operating and output variables of a boiler. Parameters selection was performed with the help of ACO. Next, model inputs describing process operating variables are also optimized using ACO with a view to maximize the combustion efficiency of the boiler. The results showed that the proposed approach, by comparing with neural network model, was an efficient way to model boiler in automation style with good predictive accuracy. ACO and SVR provide a useful tool for maximizing the combustion efficiency of boiler. Also, the method can be easily extended to other applications.

Use of differential evolution in low NO x combustion optimization of a coal-fired boiler

The present work focuses on low NOx emissions combustion modification of a 300MW dual-furnaces coal-fired utility boiler through a combination of support vector regression (SVR) and a novel and modern differential evolution optimization technique (DE). SVR, used as a more versatile type of regression tool, was employed to build a complex model between NOx emissions and operating conditions by using available experimental results in a case boiler. The trained SVR model performed well in predicting the NOx emissions with an average relative error of less than 1.14% compared with the experimental results in the case boiler. The optimal ten inputs (namely operating conditions to be optimized by operators of the boiler) of NOx emissions characteristics model were regulated by DE so that low NOx emissions were achieved, given that the boiler load is determined. Two cases were optimized in this work to check the possibility of reducing NOx emissions by DE under high and low boiler load. The time response of DE was typical of 20 sec, at the same time with the better quality of optimized results. Remarkable good results were obtained when DE was used to optimize NOx emissions of this boiler, supporting its applicability for the development of an advanced on-line and real-time low NOx emissions combustion optimization software package in modern power plants.

Effect of blockage ratios on the characteristics of methane/air explosion suppressed by BC powder

To investigate the effect of blockage ratios on the explosion suppression by powder suppressant, an experimental study was performed to suppress the methane-air explosion in a 5L duct with different blockage ratios and various concentrations of BC dry powder. The results indicate that flames experienced both the spherical and finger-shaped stages. Furthermore, the smoothness of flame front initially decreased and then increased. Flame propagation velocities were higher with larger blockage ratios except for φ = 1. The maximum peak overpressure (MPP) with the blockage ratio was slightly increased till φ reached 0.7 then surged sharply. The MPP decreased as the powder concentration increased. The maximum drop rate in the MPP being 34.8%-59.9%, depending on powder concentrations, occurred at the blockage ratio between 0.4 and 0.6. The result is ascribed to the competition between the suppression augmentation by the higher venting-generated turbulence and the suppression attenuation by the shorter residence time of the particle. However, the drop rate was relatively less promoted by increasing the concentration from 80 g/m3 to 240 g/m3. The inhibitor at higher concentration was less effective. An inhibition mechanism is explained by analogy to droplet group combustion, in which the decomposition regime of NaHCO3 differs at different concentrations.

Prediction of nitrogen oxides from coal combustion by using response surface methodology

In this paper, NOx emission prediction was studied. A simple model based on response surface methodology (RSM) was first put forward. Response surface models are multivariate polynomial models. Four RSM models were tried. The predicted NOx emission was compared with the measured ones. The RSM model with quadratic terms showed the best agreement with the measurement, and had the mean relative error of 1.6719%. The frequency of those samples whose relative error is less than 5% was 96.8610%. The RSM model was simpler than the non-analytic models such as generalized regression neural network and support vector regression. The present study will be an alternative to developing predictive emissions monitoring systems (PEMS).

Prediction of Explosion Limits of Multi-Component Gas Mixture Using LS-SVR

In safety engineering, lower and upper explosion limits are the important indices to evaluate the safety of multi-component explosive gas mixture such as hydrogen and methane. There is a nonlinear dependence of explosion limits on the composition (components and theirs concentration) of multi-component explosive gas mixture. Therefore, a least square support vector regression (LS-SVR) model was proposed to establish a non-linear model between the composition of the explosive mixture and the explosion limits. The results show that the LS-SVR model predicted explosion limits with good accuracy. The selection of input variables for the LS-SVR showed significant effect on the predictive accuracy.

Reducing NO x emission from a coal-fired boiler based on regression and optimization

NOx emission from coal combustion poses terrible threat to the surrounding environment. In order to mitigate NOx emission for coal combustion of a coal-fired boiler, a nonlinear regression model based on support vector regression (SVR) was employed to build a relationship between NOx emissions and operating parameters of the case boiler. Then, an optimization tool based on simulated annealing (SA) was utilized to regulating the combustion parameters of the case boiler aiming to achieve low NOx emission. The six levels of secondary air velocities and four levels of primary air velocities were chosen as design variables. Remarkable good results were obtained when SA was used to optimize NOx emissions of this boiler, supporting its applicability for the development of an advanced on-line and real-time low NOx emissions combustion optimization software package in modern power plants.

Optimization of NO x emission from coal combustion process using pattern search

This study presents a new approach based on a pattern search algorithm to solve combustion optimization problem (i.e. achieving expected goals by optimizing the parameters of the combustion process). In the optimization problem, the objective function was implicitly expressed by a surrogate model, which was illustrated by a support vector machine. Ten inputs for this “black box” surrogate model were chosen as design variables of the pattern search algorithm. Then, the proposed method was applied to a coal combustion process, aiming at reducing the nitrogen oxides (NOx). The outcome is very encouraging and suggests that PS methods may be very efficient when solving low NOx combustion optimization problem.