PEMS: Predictive Emission Monitoring Systems, is a misnomer of sorts;
this is not monitoring, it replaces monitoring by modeling.
PEMS provides continuous emission estimates, same as a CEMS (Continuous Emission Monitoring System, or stack monitoring.
PEMS, however, does not use sensors to determine the flue gas composition from direct measurement:
PEMS uses dynamic process data to predict (model) the emissions.
AirWare uses a hybrid PEMS: the algorithms or models used
range from more complex (tier 3) emission factors to higher order Polynomials,
cumbustion models based on reaction kinetics, stoichiometry and thremodynamics.
MLT (multiple linear regression) to Case Based Reasoning (CBR),
The PEMS model is continuously fed with activity or process data.
These typically include fuel quality or composistion (could be more or less static), fuel flow,
air flow, process temperature and oxygen levels.
The emission model, real-time acquisition of process data, continuous self-check,
warnings, alerts, alarms and messages are driven by the Real-time Hybrid Expert System RTXPS
for real-time control.
To setup, calibrate and tune a PEMS, an inital set of observations of both the
process data and the corresponding emissions is needed to estimate the coefficient of the various models.
A dedicated test environment with synthetic or monitoring data facilitates that process,
including the determination of the calibration data requirements based on the convergence
of estimates and a residual within performance bounds.
Validation is defined by Performance Specification such as EPA
Performance Specification 16 (part of 40 CFR 60, Federal register Vol.74, 56 (2009),
that stipulates the nature and number of calibration runs and target accuracy for certification.
Predictive Emission Monitoring is generally less expensive and can
be as accurate as CEMS (Continuous Emission Monitoring Systems).
Apart from the cost of testing to obtain the correlation, Predictive Emission Monitoring utilizes
equipment and operational information otherwise necessary for the operation
of a power station. This leads to an added potential benefit of being able
to correlate operational parameters directly to emission rates.
Once a correlation has been established (the model calibrated from an inital set of stack measurements),
Predictive Emission Monitoring does not require regular system maintenance. In addition, it cvan be used
to "monitor" the process sensors, detected anomalies and drift.
Predictive Emission Monitoring is easily adaptable to existing facilities without
expensive and sensitive Continuous Emission Monitoring Systems.
Existing facilities without an installed Continuous Emission Monitoring Systems can face extremely high
retrofit costs when compared with initial installation costs of Continuous Emission Monitoring Systems
at the time of construction a new facility. It also minimizes downtime as no additional
equipment is required and increases the ability of a facility to comply with tighter monitoring requirements.
Another benefit is its ability to detect anomalies in the operation as well as to better understand
correlations between operating conditions and emission rates. This would assist in fine tuning operations
to maximize power output while minimizing emissions.
To obtain a sufficiently sized data set to develop a
Predictive Emission Monitoring, emission and plant operational data have to be collected
from a number of tests over a wide range of operating conditions.
This process can be expensive and time consuming.
Predictive Emission Monitoring is not a substitute for Continuous Emission Monitoring Systems
in cases when Continuous Emission Monitoring Systems is mandated by the applicable regulations.
See also: AirWare emission data and modeling examples