The Maturation of a Technology: Predictive Emissions Monitoring
By Paul S. Reinermann, III, Pavilion Technologies, Inc. |
Air pollution is monitored by many different regulatory programs, some of which require that emissions be measured continuously. This task is often accomplished by directly measuring air pollutants with continuous emissions monitoring systems (CEMS). In recent years, innovative technology has enabled a new approach that predicts emissions from process variables. Industry welcomes predictive emission monitoring systems (PEMS) because they reduce emissions-monitoring costs. Regulatory agencies are accepting these new systems, as proof of their accuracy and reliability is demonstrated.
The onset of PEMS was in 1992, when a chemical engineer used neural-networks (Figure 1), a pattern-learning technology, to accurately predict nitrogen oxides (NOx) emissions from a natural-gas-fired boiler. This model worked like no other first-principles model or regression model previously built — it was accurate over the entire operating range of the boiler, whereas the other models were accurate only in specific operations. Within a year, the neural-network-based NOx model evolved into the first PEMS that was approved by the U.S. Environmental Protection Agency (EPA) as an alternative to direct CEMS measurements. This first PEMS…