Time-Variability of NOx Emissions from Portland Cement Kilns

Abstract

Due to the presence of autocorrelation between sequentially measured nitrogen oxide (NOx) concentrations in stack gas from portland cement kilns, the determination of the average emission rates and the uncertainty of the average has been improperly calculated by the industry and regulatory agencies. Documentation of permit compliance, establishment of permit levels, and the development and testing of control techniques for reducing NOx emissions at specific cement plants requires accurate and precise statistical estimates of parameters such as means, standard deviations, and variances. Usual statistical formulas such as for the variance of the sample mean only apply if sequential measurements of NOx emissions are independent. Significant autocorrelation of NOx emission measurements revealed that NOx concentration values measured by continuous emission monitors are not independent but can be represented by an autoregressive, moving average (ARMA) time series. Three orders of time-variability of NOx emission rates were determined from examination of continuous emission measurements from several cement kilns. Long-term variations are related to changes in process control, feed composition, fuel composition, and types of product produced and are reflected by nonperiodic drift in the time series. Intermediate-term variations result from changes in process control and operator technique. Very short-term variations, which are not normally visible due to the averaging process applied to data collection, are related to the dynamics of the flame. Autocorrelation effects within the emission rates extend over a period of 8−10 h. The high levels of autocorrelation observed between sequential readings result in underestimating the variance of the average emission rate unless autocorrelation effects are considered. Although thousands of NOx measurements may be recorded, autocorrelation of sequential measurements implies that many of them are redundant.