Predictive Emissions Monitoring System (PEMS) for a Paper Mill Power Boiler, 1997 Environmental Conference Proceedings

A Predictive Emissions Monitoring System (PEMS) was installed on a power boiler at Temple-Inland’s mill in Evadale, TX in October 1995. This is one of the first PEMS to be installed at a pulp and paper facility. The Temple-Inland PEMS underwent a relative accuracy test audit (RATA) after the initial installation, and has been recertified three times since. The most recent RATA results demonstrated that the accuracy of the predicted values for NO,, CO, and O2 continue to be well within acceptable limits. This paper will present the results of the RATA tests, and will provide a summary of the lessons that were learned during the development and certification of the power boiler PEMS.

The core technology in PEMS are run-time sensor validation and prediction models developed using a database neural network software system. A key feature of the PEMS is the incorporation of sensor validation. Each process sensor signal is first first screened for sensor drift or sensor failure by passing the sensor signal through a neural network sensor validation model. Validated process data are then passed to a run-time prediction model that predicts the emissions of nitrogen oxides (NO,), carbon monoxide (CO) and excess oxygen (0,). These predicted values are then reported to the state regulatory agency. The advantage of a PEMS is that it relies on process sensors that are already available as part of the unit’s operation. This eliminates the expense of purchasing and maintaining continuous emissions analyzers.