An Intelligent Emissions Controller for Fuel Lean Gas Reburn in Coal-Fired Power Plants

ABSTRACT

The application of artificial intelligence techniques for performance optimization of the fuel lean gas reburn (FLGR) system is investigated. A multilayer, feedforward artificial neural network is applied to model static nonlinear relationships between the distribution of injected natural gas into the upper region of the furnace of a coal-fired boiler and the corresponding oxides of nitrogen (NO_x) emissions exiting the furnace. Based on this model, optimal distributions of injected gas are determined such that the largest NO_x reduction is achieved for each value of total injected gas. This optimization is accomplished through the development of a new optimization method based on neural networks. This new optimal control algorithm, which can be used as an alternative generic tool for solving multidimensional nonlinear constrained optimization problems, is described and its results are successfully validated against an off-the-shelf tool for solving mathematical programming problems. Encouraging results obtained using plant data from one of Commonwealth Edison's coal-fired electric power plants demonstrate the feasibility of the overall approach.

Preliminary results show that the use of this intelligent controller will also enable the determination of the most cost-effective operating conditions of the FLGR system by considering, along with the optimal distribution of the injected gas, the cost differential between natural gas and coal and the open-market price of NO_x emission credits. Further study, however, is necessary, including the construction of a more comprehensive database, needed to develop high-fidelity process models and to add carbon monoxide (CO) emissions to the model of the gas reburn system.     

Identification and Control of NOx Emissions Using Neural Networks

ABSTRACT

We investigate the application of two classes of artificial neural networks for the identification and control of discrete-time nonlinear dynamical systems. A fully connected recurrent network is used for process identification, and a multilayer feedforward network is used for process control. The two neural networks are arranged in series for closed-loop control of oxides of nitrogen (NO_x) emissions of a simplified representation of a dynamical system. Plant data from one of Commonwealth Edison's coal-fired power plants are used for testing the approach, with initial results indicating that the method is feasible. However, further work is required to determine whether the method remains feasible as the number of state variables and control variables are increased.     

Nitrogen oxides emission control options for coal-fired electric utility boilers

Recent regulations have required reductions in emissions of nitrogen oxides (NOx) from electric utility boilers. To comply with these regulatory requirements, it is increasingly important to implement state-of-the-art NOx control technologies on coal-fired utility boilers. This paper reviews NOx control options for these boilers. It discusses the established commercial primary and secondary control technologies and examines what is being done to use them more effectively. Furthermore, the paper discusses recent developments in NOx controls. The popular primary control technologies in use in the United States are low-NOx burners and overfire air. Data reflect that average NOx reductions for specific primary controls have ranged from 35% to 63% from 1995 emissions levels. The secondary NOx control technologies applied on U.S. coal-fired utility boilers include reburning, selective noncatalytic reduction (SNCR), and selective catalytic reduction (SCR). Thirty-six U.S. coal-fired utility boilers have installed SNCR, and reported NOx reductions achieved at these applications ranged from 15% to 66%. Recently, SCR has been installed at >150 U.S. coal-fired utility boilers. Data on the performance of 20 SCR systems operating in the United States with low-NOx emissions reflect that in 2003, these units achieved NOx emission rates between 0.04 and 0.07 lb/10(6) Btu.     

电厂NO_x控制政策与技术：美国的经验及对我国的启示

我国NOx排放的快速增长导致其环境和生态影响日益加剧,已经引起全球关注,但是现行的NOx污染控制技术和政策远远落后于实际需求。为减少酸雨、臭氧和颗粒物带来的环境问题,美国采取排放控制和排污交易等方式削减了燃煤电厂的NOx排放。系统分析了美国NOx污染控制法规、标准、规划和控制技术应用情况,总结了其NOx控制的成功经验,进一步提出了符合我国国情的NOx控制对策和技术选择。     

Evaluation of Tire-Derived Fuel for Use in Nitrogen Oxide Reduction by Reburning

ABSTRACT

Tire-derived fuel (TDF) was tested in a small-scale (44 kW or 150,000 Btu/hr) combustor to determine its feasibility as a fuel for use in reburning for control of nitrogen oxide (NO). TDF was gravity-fed into upward flowing combustion gases from a primary natural gas flame doped with ammonia to simulate a high NO combustion process. Emissions of NO, oxygen, carbon dioxide, carbon monoxide, and particulate matter were measured. The tests varied the nominal primary NO level from 600 to 1,200 ppm and the primary stoichiometry from 1.1 to 1.2, and used both natural gas and TDF as reburn fuels. The reburn injection rate was varied to achieve 8–20% of the total heat input from the reburn fuel. NO emissions reductions ranged between 20 and 63% when using TDF, depending upon the rate of TDF injection, primary NO, and primary stoichiometry. NO emission reductions when using natural gas as the reburn fuel were consistently higher than those when using TDF. While additional work remains to optimize the process and evaluate costs, TDF has been shown to have the potential to be a technically viable reburning fuel.     

Modeling and optimization of wet flue gas desulfurization system based on a hybrid modeling method

Sulfur dioxide (SO₂) is one of the main air pollutants from many industries. Most coal-fired power plants in China use wet flue gas desulfurization (WFGD) as the main method for SO₂ removal. Presently, the operating of WFGD lacks accurate modeling method to predict outlet concentration, let alone optimization method. As a result, operating parameters and running status of WFGD are adjusted based on the experience of the experts, which brings about the possibility of material waste and excessive emissions. In this paper, a novel WFGD model combining a mathematical model and an artificial neural network (ANN) was developed to forecast SO₂ emissions. Operation data from a 1000-MW coal-fired unit was collected and divided into two separated sets for model training and validation. The hybrid model consisting a mechanism model and a 9-input ANN had the best performance on both training and validation sets in terms of RMSE (root mean square error) and MRE (mean relative error) and was chosen as the model used in optimization. A comprehensive cost model of WFGD was also constructed to estimate real-time operation cost. Based on the hybrid WFGD model and cost model, a particle swarm optimization (PSO)-based solver was designed to derive the cost-effective set points under different operation conditions. The optimization results demonstrated that the optimized operating parameters could effectively keep the SO₂ emissions within the standard, whereas the SO₂ emissions was decreased by 30.79% with less than 2% increase of total operating cost.

Implications: Sulfur dioxide (SO₂) is one of the main pollutants generated during coal combustion in power plants, and wet flue gas desulfurization (WFGD) is the main facility for SO₂ removal. A hybrid model combining SO₂ removal mathematical model with data-driven model achieves more accurate prediction of outlet concentration. Particle swarm optimization with a penalty function efficiently solves the optimization problem of WFGD subject to operation cost under multiple operation conditions. The proposed model and optimization method is able to direct the optimized operation of WFGD with enhanced emission and economic performance.     

Reburning for Cyclone Boiler N0x Control

Recent pilot-scale testing cosponsored by the Electric Power Research Institute (EPRI) and the Gas Research Institute (GRI) indicates that reburning can reduce NO_x emissions by 40-60 percent in cyclone boilers when pulverized coal (PC), oil, or natural gas is used as the reburn fuel. The pilot tests, performed at the Babcock and Wilcox (B&W) Alliance Research Center using a six million Btu/h cyclone-fired pilot-scale furnace, were designed to confirm and expand upon the conclusions of an earlier B&W feasibility study. That study predicted reburning could reduce NO_x by 50 percent in most cyclone boilers now in operation.     

Comparing lifetime emissions of natural gas and conventional fuel vehicles: an application of the generalized ANCOVA model

New regulations and incentives are encouraging the use of clean, alternative fuel vehicles (AFVs) in urban areas. These vehicles are seen as one option for reducing air pollution from mobile sources. However, because of the limited number of AFVs on the road, little is known about actual lifetime emissions characteristics of in-use AFVs. This study describes the use of a generalized analysis of covariance model to evaluate and compare the emissions from natural gas vehicles with emissions from reformulated gasoline vehicles. The model describes fleet-wide emissions deterioration, while also accounting for individual vehicle variability within the fleet. This ability to measure individual vehicle variability can then be used to provide realistic bounds for the emissions deterioration in individual vehicles and the fleet as a whole. In order to illustrate the use of the model, the carbon monoxide, oxides of nitrogen (NOx), non-methane hydrocarbon (NMHC), and carbon dioxide emissions characteristics of a fleet of dedicated natural gas Dodge Ram vans and a fleet of dedicated reformulated gasoline Dodge Ram vans operating in the U.S. government fleet are explored. The analysis demonstrates the utility of the statistical method and suggests a potential for natural gas Dodge Ram vans to be generally cleaner than their conventional fuel counterparts. However, in the case of NOx and NHMCs, the analysis also suggests that these emissions benefits might be reduced over the vehicle lifetime due to higher emissions deterioration rates for natural gas vehicles. As this paper is aimed at illustrating the analysis of the covariance model, the results reported herein should be considered within the context of a more comprehensive study of these data before general conclusions are possible. Generalization of these findings to other vehicle models and alternative fuel technologies is not justified without further study.     

Effect of NOx control processes on mercury speciation in utility flue gas

The speciation of Hg in coal-fired flue gas can be important in determining the ultimate Hg emissions as well as potential control options for the utility. The effects of NOx control processes, such as selective catalytic reduction (SCR) and selective non-catalytic reduction (SNCR), on Hg speciation are not well understood but may impact emissions of Hg. EPRI has investigated the reactions of Hg in flue gas at conditions expected for some NOx control processes. This paper describes the methodology used to investigate these reactions in actual flue gas at several power plants. Results have indicated that some commercial SCR catalysts are capable of oxidizing elemental Hg in flue gas obtained from the inlets of SCR or air heater units. Results are affected by various flue gas and operating parameters. The effect of flue gas composition, including the presence of NH3, has been evaluated. The influence of NH3 on fly ash Hg reactions also is being investigated.     

The potential near-source ozone impacts of upstream oil and gas industry emissions

Increased drilling in urban areas overlying shale formations and its potential impact on human health through decreased air quality make it important to estimate the contribution of oil and gas activities to photochemical smog. Flares and compressor engines used in natural gas operations, for example, are large sources not only of NOx but also offormaldehyde, a hazardous air pollutant and powerful ozone precursor We used a neighborhood scale (200 m horizontal resolution) three-dimensional (3D) air dispersion model with an appropriate chemical mechanism to simulate ozone formation in the vicinity ofa hypothetical natural gas processing facility, based on accepted estimates of both regular and nonroutine emissions. The model predicts that, under average midday conditions in June, regular emissions mostly associated with compressor engines may increase ambient ozone in the Barnett Shale by more than 3 ppb beginning at about 2 km downwind of the facility, assuming there are no other major sources of ozone precursors. Flare volumes of 100,000 cubic meters per hour ofnatural gas over a period of 2 hr can also add over 3 ppb to peak 1-hr ozone somewhatfurther (>8 km) downwind, once dilution overcomes ozone titration and inhibition by large flare emissions of NOx. The additional peak ozone from the hypothetical flare can briefly exceed 10 ppb about 16 km downwind. The enhancements of ambient ozone predicted by the model are significant, given that ozone control strategy widths are of the order of a few parts per billion. Degrading the horizontal resolution of the model to 1 km spuriously enhances the simulated ozone increases by reducing the effectiveness of ozone inhibition and titration due to artificial plume dilution.     

Idle emissions from heavy-duty diesel and natural gas vehicles at high altitude

Idle emissions of total hydrocarbon (THC), CO, NOx, and particulate matter (PM) were measured from 24 heavy-duty diesel-fueled (12 trucks and 12 buses) and 4 heavy-duty compressed natural gas (CNG)-fueled vehicles. The volatile organic fraction (VOF) of PM and aldehyde emissions were also measured for many of the diesel vehicles. Experiments were conducted at 1609 m above sea level using a full exhaust flow dilution tunnel method identical to that used for heavy-duty engine Federal Test Procedure (FTP) testing. Diesel trucks averaged 0.170 g/min THC, 1.183 g/min CO, 1.416 g/min NOx, and 0.030 g/min PM. Diesel buses averaged 0.137 g/min THC, 1.326 g/min CO, 2.015 g/min NOx, and 0.048 g/min PM. Results are compared to idle emission factors from the MOBILE5 and PART5 inventory models. The models significantly (45-75%) overestimate emissions of THC and CO in comparison with results measured from the fleet of vehicles examined in this study. Measured NOx emissions were significantly higher (30-100%) than model predictions. For the pre-1999 (pre-consent decree) truck engines examined in this study, idle NOx emissions increased with model year with a linear fit (r2 = 0.6). PART5 nationwide fleet average emissions are within 1 order of magnitude of emissions for the group of vehicles tested in this study. Aldehyde emissions for bus idling averaged 6 mg/min. The VOF averaged 19% of total PM for buses and 49% for trucks. CNG vehicle idle emissions averaged 1.435 g/min for THC, 1.119 g/min for CO, 0.267 g/min for NOx, and 0.003 g/min for PM. The g/min PM emissions are only a small fraction of g/min PM emissions during vehicle driving. However, idle emissions of NOx, CO, and THC are significant in comparison with driving emissions.     

Forecasting Daily Source Air Quality Using Multivariate Statistical Analysis and Radial Basis Function Networks

Abstract

It is vital to forecast gas and particle matter concentrations and emission rates (GPCER) from livestock production facilities to assess the impact of airborne pollutants on human health, ecological environment, and global warming. Modeling source air quality is a complex process because of abundant nonlinear interactions between GPCER and other factors. The objective of this study was to introduce statistical methods and radial basis function (RBF) neural network to predict daily source air quality in Iowa swine deep-pit finishing buildings. The results show that four variables (outdoor and indoor temperature, animal units, and ventilation rates) were identified as relative important model inputs using statistical methods. It can be further demonstrated that only two factors, the environment factor and the animal factor, were capable of explaining more than 94% of the total variability after performing principal component analysis. The introduction of fewer uncorrelated variables to the neural network would result in the reduction of the model structure complexity, minimize computation cost, and eliminate model overfitting problems. The obtained results of RBF network prediction were in good agreement with the actual measurements, with values of the correlation coefficient between 0.741 and 0.995 and very low values of systemic performance indexes for all the models. The good results indicated the RBF network could be trained to model these highly nonlinear relationships. Thus, the RBF neural network technology combined with multivariate statistical methods is a promising tool for air pollutant emissions modeling.     

Effectiveness in the use of natural gas for the reduction of atmospheric emissions: case study--industrial sector in the metropolitan region of Santiago,Chile

The purpose of this investigation was to quantify the potential of natural gas to reduce emissions from stationary combustion sources by analyzing the case study of the metropolitan region of Santiago, Chile. For such purposes, referential base scenarios have been defined that represent with and without natural gas settings. The method to be applied is an emission estimate based on emission factors. The results for this case study reveal that stationary combustion sources that replaced their fuel reduced particulate matter (PM) emissions by 61%, sulfur oxides (SOx) by 91%, nitrogen oxides (NOx) by 40%, and volatile organic compounds (VOC) by 10%. Carbon monoxide (CO) emissions were reduced by 1%. As a result of this emission reduction, in addition to reductions caused by other factors, such as a shift to cleaner fuels other than natural gas, technological improvements, and sources which are not operative, emission reduction goals set forth by the environmental authorities were broadly exceeded.     

Technology innovations and experience curves for nitrogen oxides control technologies

This paper reviews the regulatory history for nitrogen oxides (NOx) pollutant emissions from stationary sources, primarily in coal-fired power plants. Nitrogen dioxide (NO2) is one of the six criteria pollutants regulated by the 1970 Clean Air Act where National Ambient Air Quality Standards were established to protect public health and welfare. We use patent data to show that in the cases of Japan, Germany, and the United States, innovations in NOx control technologies did not occur until stringent government regulations were in place, thus "forcing" innovation. We also demonstrate that reductions in the capital and operation and maintenance (O&M) costs of new generations of high-efficiency NOx control technologies, selective catalytic reduction (SCR), are consistently associated with the increasing adoption of the control technology: the so-called learning-by-doing phenomena. The results show that as cumulative world coal-fired SCR capacity doubles, capital costs decline to approximately 86% and O&M costs to 58% of their original values. The observed changes in SCR technology reflect the impact of technological advance as well as other factors, such as market competition and economies of scale.     

PAH and soot emissions from burning components of medical waste: examination/surgical gloves and cotton pads

This is a laboratory investigation on the emissions from batch combustion of representative infectious ("red bag") medical waste components, such as medical examination latex gloves and sterile cotton pads. Plastics and cloth account for the majority of the red bag wastes by mass and, certainly, by volume. An electrically heated, horizontal muffle furnace was used for batch combustion of small quantities of shredded fuels (0.5-1.5 g) at a gas temperature of approximately 1000 degrees C. The residence time of the post-combustion gases in the furnace was approximately 1 s. At the exit of the furnace, the following emissions were measured: CO, CO2, NOx, particulates and polynuclear aromatic compounds (PACs). The first three gaseous emissions were measured with continuous gas analyzers. Soot and PAC emissions were simultaneously measured by passing the furnace effluent through a filter (to collect condensed-phase PACs) and a bed of XAD-4 adsorbent (to capture gaseous-phase PACs). Analysis involved soxhlet extraction, followed by gas chromatography-mass spectrometry (GC-MS). Results were contrasted with previously measured emissions from batch combustion of pulverized coal and tire-derived fuel (TDF) under similar conditions. Results showed that the particulate soot) and cumulative PAC emissions from batch combustion of latex gloves were more than an order of magnitude higher than those from cotton pads. The following values are indicative of the relative trends (but not necessarily absolute values) in emission yields: 26% of the mass of the latex was converted to soot, 11% of which was condensed PAC. Only 2% of the mass of cotton pads was converted to soot, and only 3% of the weight of that soot was condensed PAC. The PAC yields from latex were comparable to those from TDF. The PAC yields from cotton were higher than those from coal. A notable exception to this trend was that the three-ring gas-phase PAC yields from cotton were more significant than those from latex. Emission yields of CO and CO2 from batch combustion of cotton were, respectively, comparable and higher than those from latex, despite the fact that the carbon content of cotton was half that of latex. This is indicative of the more effective combustion of cotton. Nearly all of the mass of carbon of cotton gasified to CO and CO2 while only small fractions of the carbon in latex were converted to CO2 and CO (20% and 10%, respectively). Yields of NOx from batch combustions of latex and cotton accounted for 15% and 12%, respectively, of the mass of fuel nitrogen indicating that more fuel nitrogen was converted to NOx in the former case, possibly due to higher flame temperatures. No SO2 emissions were detected, indicating that during the fuel-rich combustion of latex, its sulfur content was converted to other compounds (such as H2S) or remained in the soot.     

Nitric oxide formation in an iron oxide pellet rotary kiln furnace

A one-dimensional numerical model was developed to simulate the effects of heat and mass transfer on the formation of oxides of nitrogen (NOx) in a rotary kiln furnace for iron oxide pellet induration. The modeled kiln has a length-to-diameter ratio of approximately seven. The principal mechanism of heat transfer is radiation from the flame, which was described by the net radiation method. The well known Zeldovich mechanism was used to predict thermal NOx generation. Temperature fluctuations in the vicinity of the flame were estimated with a clipped Gaussian probability density function. The thermal energy and mass balance model equations were solved numerically. The model is capable of predicting temperature profiles and NOx production rates in agreement with observed plant performance. The model was used to explore the effects of process changes on the total NOx formation in the kiln. It was concluded that the gas temperature as well as the partial pressure of oxygen in the process gases controls the rate of NOx formation. Lowering the temperature of the kiln gases by increasing the secondary air flow rates requires simultaneously decreasing the pellet production rate in order to maintain the pellet temperatures needed for blast furnace conditions.     

An alternative to additional SO3 injection for fly ash conditioning

Small concentrations, approximately 2-10 parts per million (ppm), of injected sulfur trioxide (SO3) have improved particulate collection efficiencies of electrostatic precipitators burning lower-sulfur coal. However, the addition of extra SO3 not only incurs costs but also presents negative environmental effects. This work explored a method that could be applied to existing coal-fired power plants to convert the sulfur dioxide (SO2) already present in the flue gas to sufficient levels of SO3 for fly ash conditioning as an alternative to adding SO3 by burning elemental sulfur. During this research, a pre-mixed natural gas flame was used to promote the conversion of SO2 to SO3 in a drop-tube furnace with average non-flame, free stream gas temperatures of 450 and 1000 K. SO3 concentrations measured by wet chemistry and confirmed using elemental balances of other sulfur species measured by gas chromatography revealed that as much as 7% of SO2 was homogeneously transformed to SO3. The results also showed that at low temperatures, the rate at which SO3 is converted back to SO2 decreased, thus extending the time period during which SO3 concentrations would be sufficient for ash conditioning. An additional benefit of this technique is speculated to result from increased flue gas humidity.     

Investigation of selective catalytic reduction impact on mercury speciation under simulated NOx emission control conditions

Selective catalytic reduction (SCR) technology increasingly is being applied for controlling emissions of nitrogen oxides (NOx) from coal-fired boilers. Some recent field and pilot studies suggest that the operation of SCR could affect the chemical form of mercury (Hg) in coal combustion flue gases. The speciation of Hg is an important factor influencing the control and environmental fate of Hg emissions from coal combustion. The vanadium and titanium oxides, used commonly in the vanadia-titania SCR catalyst for catalytic NOx reduction, promote the formation of oxidized mercury (Hg2+). The work reported in this paper focuses on the impact of SCR on elemental mercury (Hg0) oxidation. Bench-scale experiments were conducted to investigate Hg0 oxidation in the presence of simulated coal combustion flue gases and under SCR reaction conditions. Flue gas mixtures with different concentrations of hydrogen chloride (HCl) and sulfur dioxide (SO2) for simulating the combustion of bituminous coals and subbituminous coals were tested in these experiments. The effects of HCl and SO2 in the flue gases on Hg0 oxidation under SCR reaction conditions were studied. It was observed that HCl is the most critical flue gas component that causes conversion of Hg0 to Hg2+ under SCR reaction conditions. The importance of HCl for Hg0 oxidation found in the present study provides the scientific basis for the apparent coal-type dependence observed for Hg0 oxidation occurring across the SCR reactors in the field.     

Economics of an integrated approach to control SO2, NOX,HCl, and particulate emissions from power plants

An integrated approach for the simultaneous reduction of major combustion-generated pollutants from power plants is presented along with a simplified economic analysis. With this technology, the synergistic effects of high-temperature sorbent/coal or sorbent/natural gas injection and high-temperature flue gas filtration are exploited. Calcium-based (or Na-based, etc.) sorbents are sprayed in the post-flame zone of a furnace, where they react with S- and Cl-containing gases to form stable salts of Ca (or Na, etc.). The partially reacted sorbent is then collected in a high-temperature ceramic filter, which is placed downstream of the sorbent injection point, where it further reacts for a prolonged period of time. With this technique, both the likelihood of contact and the length of time of contact between the solid sorbent particles and the gaseous pollutants increase, because reaction takes place both in the furnace upstream of the filter and inside the filter itself. Hence, the sorbent utilization increases significantly. Several pollutants, such as SO2, H2S, HCl, and particulate (soot, ash, and tar), may be partially removed from the effluent. The organic content of the sorbents (or blends) also pyrolyzes and reduces NOx. Unburned carbon in the ash may be completely oxidized in the filter. The filter is cleaned periodically with aerodynamic regeneration (back pulsing) without interrupting furnace operation. The effectiveness of this technique has been shown in laboratory-scale experiments using either rather costly carboxylic salts of Ca or low- to moderate-cost blends of limestone, lime, or sodium bicarbonate with coal fines. Injection occurred in the furnace at 1150 degrees C, while the filter was maintained at 600 degrees C. Results showed that 65 or 40% SO2 removal was obtained with calcium formate or a limestone/coal blend, respectively, at an entering calcium-to-sulfur molar ratio of 2. A sodium bicarbonate/coal blend resulted in 78% SO2 removal at a sodium-to-sulfur molar ratio of 2. HCl removal efficiencies have been shown to be higher than those for SO2. NOx reductions of 40% have been observed with a fuel (coal)-to-air equivalence ratio, phi, around 2. The filter has been shown to be 97-99% efficient in removing PM2.5 particulates. Calculations herein show that this integrated sorbent/filter method is cost-effective, in comparison with current technologies, on both capital cost ($/kW) and levelized cost ($/ton pollutant removed) bases, if a limestone/coal mixture is used as the sorbent for fossil fuel plants. Capital costs for the filter/sorbent combination are estimated to be in the range of $61-$105/kW for a new plant. Because current technologies are designed for removing one pollutant at a time, both their cost and space requirements are higher than those of this integrated technique. At the minimum projected removal efficiencies for HCl/SO2/NOx of about 40%, the levelized costs are projected to be $203-$261/ton of combined pollutant SO2/HCl/NOx and particulates removed from coal-fired power plants.     

A feasibility study on the predictive emission monitoring system applied to the Hsinta power plant of Taiwan Power Company

The continuous emission monitoring system (CEMS) can monitor flue gas emissions continuously and instantaneously. However, it has the disadvantages of enormous cost, easily producing errors in sampling periods of bad weather, lagging response in variable ambient environments, and missing data in daily zero and span tests and maintenance. The concept of a predictive emission monitoring system (PEMS) is to use the operating parameters of combustion equipment through thermodynamic or statistical methods to construct a mathematic model that can predict emissions by a computer program. The goal of this study is to set up a PEMS in a gas-fired combined cycle power generation unit at the Hsinta station of Taiwan Power Co. The emissions to be monitored include nitrogen oxides (NOx) and oxygen (O2) in flue gas. The major variables of the predictive model were determined based on the combustion theory. The data of these variables then were analyzed to establish a regression model. From the regression results, the influences of these variables are discussed and the predicted values are compared with the CEMS data for accuracy. In addition, according to the cost information, the capital and operation and maintenance costs for a PEMS can be much lower than those for a CEMS.