Validation of prediction models for NO_x-monitoring
Öberg, T.
Stiftelsen för värmeteknisk forskning, Stockholm, Sweden.
Report 586, October 1996. ISSN 0282-3722. (In Swedish)

Abstract
Predictive emission monitoring can be accepted for determining the environmental fees for NO_x. Requirements for validity are assumed to be set comparable to those for measurement instruments. This report describes how prediction models should be constructed to fulfil these requirements.

A substantial number of plants in the U.S. are already using prediction models to measure the emissions of nitrogen oxides. Suggested validation requirements are comparable to those for measurement instruments. The model shall be established on the basis of comparisons with a reference method during different operating condi-tions. The performance verification test is comparable, with specified requirements for precision and accuracy.

Detailed quality requirements are not yet specified in Sweden. The following items must however be considered:

• Input data for the model building.
• Model verification - recalibration.
• Performance and definition of calibration range.
• Model domain check and signal validation.

An evaluation of results from two completed projects shows that predictive emission monitoring for nitrogen oxides is technically feasible.

The report is concluded with a recommendation on how to "translate" the present performance requirements of the Swedish Environmental Protection Agency regarding measurement instruments. The most important performance requirements for predic-tion models at comparison tests are that the systematic and the standard deviations are not above 10 % of the calibration range and at the most 10 ppm dry gas.