UNITED
STATES
ENVIRONMENTAL
PROTECTION
AGENCY
WASHINGTON,
D.
C.
20460
DRAFT
OFFICE
OF
AIR
AND
RADIATION
3/
23/
05
Mr.
Phillip
Polyak
Designated
Representative
Dearborn
Industrial
Generation
P.
O.
Box
126
Dearborn,
MI
48121­
0126
Re:
Additional
Quality
Assurance
Requirements
for
Conditionally­
Approved
Predictive
Emissions
Monitoring
System
at
Dearborn
Industrial
Generation
(
Facility
ID
(
ORISPL)
55088),
Unit
GTP1
Dear
Mr.
Polyak:

This
response
finalizes
approval
of
the
October
24,
2002
petition
submitted
under
§
75.66(
d)
and
40
CFR
Part
75,
Subpart
E.
This
petition
requested
approval
of
a
NO
x
predictive
emission
monitoring
system
(
PEMS)
at
Unit
GTP1
at
Dearborn
Industrial
Generation
(
DIG)
in
Dearborn,
Michigan.
Section
75.20(
f)
requires
EPA
to
publish
the
initial
certification
request
of
an
alternative
monitoring
system
in
the
Federal
Register
and,
following
a
public
comment
period
of
60
days,
to
issue
a
notice
of
approval
or
disapproval.
The
public
comment
period
closed
on
_______
with
no
comments
received.

There
is
an
Attachment
to
this
response
that
is
designed
to
provide
easier
implementation
of
the
final
requirements
for
Unit
GTP1.
The
Attachment
for
Unit
GTP1
consolidates
additional
quality
assurance
requirements
from
EPA's
field
study,
and
EPA's
September
2,
2003
conditional
approval.

The
final
requirements
for
Unit
GTP1
include
a
more
stringent
PEMS
sensor
alarm
system,
and
monthly,
three­
run
relative
accuracy
audits
to
be
performed
on
the
PEMS
during
the
ozone
season.
EPA
also
requires
an
initial
certification
and
recertification
NO
x
relative
accuracy
test
audit
(
RATA)
to
be
performed
at
three
operating
levels
and
an
F­
test,
correlation
analysis
and
t­
test
to
be
performed
using
the
RATA
data.
Because
DIG
has
already
complied
with
the
initial
certification
requirements
in
the
original
September
2,
2003
petition
response,
it
does
not
need
to
perform
another
initial
certification
unless
a
new
NO
x
PEMS
monitoring
system
is
installed.
Ongoing
NO
x
RATAs
are
to
be
performed
at
just
the
normal
operating
level
without
statistical
tests.
With
the
inclusion
of
the
additional
QA
measures
provided
in
this
letter,
EPA
fully
approves
the
use
of
the
subject
PEMS
on
Unit
GTP1
at
DIG.

Background
In
paragraph
(
e)
of
EPA's
September
2,
2003
conditional
PEMS
approval
for
Unit
GTP1
at
DIG,
EPA
reserved
the
right
to
require
the
owner
or
operator
of
GTP1,
i.
e.,
DIG,
to
use
portable
NO
x
and
diluent
gas
(
CO
2
or
O
2)
analyzers
(
or,
as
an
option,
to
use
EPA
reference
methods)
to
2
1
"
Evaluation
of
Portable
Analyzers
for
Use
in
Quality
Assuring
Predictive
Emission
Monitoring
Systems
for
NOx",
The
Cadmus
Group,
Inc.,
September
8,
2004
perform
periodic
assessments
of
the
accuracy
of
the
PEMS,
if
and
when
EPA
determined
that
portable
NO
x
analyzers
can
provide
adequate
PEMS
accuracy
checks.
EPA
stated
that
it
would
provide
DIG
with
the
necessary
performance
specifications,
sampling
frequency,
methodology,
and
reporting
guidance,
should
this
become
a
requirement.
EPA
also
stated
that
over
the
next
few
months,
it
would
test
several
portable
electrochemical
and
chemiluminescent
NO
x
analyzers
at
combustion
turbine
sites
to
determine
how
well
these
analyzers
work.
Finally,
EPA
indicated
that
if
periodic,
direct
checks
of
PEMS
accuracy
with
portable
analyzers
should
become
a
requirement,
it
would
be
implemented
in
such
a
way
that
the
unit
would
be
tested
at
different
operating
levels
from
check­
to­
check.

Since
issuing
the
September
2,
2003
conditional
PEMS
approval,
EPA
has
completed
a
field
study
of
portable
NO
x
monitors,
analyzed
the
results,
and
performed
a
cost
assessment1.
For
the
two,
low
NO
x
emitting,
natural
gas­
fired
combustion
turbines
tested,
the
accuracy
of
the
portable
analyzers
was
found
to
be
comparable
to
that
of
a
certified
Part
75
CEMS
and
to
EPA
Reference
Method
7E.
The
cost
assessment
suggests
that
the
cost
to
perform
monthly
relative
accuracy
audits
on
a
PEMS
is
reasonable.

In
paragraph
(
j)
of
EPA's
September
2,
2003
conditional
PEMS
approval
for
Unit
GTP1,
EPA
reserved
the
right
to
require
new
statistical
procedures
or
to
change
the
ones
currently
required.
Since
issuing
the
September
2,
2003
conditional
PEMS
approval,
EPA
has
completed
a
field
study
of
a
hybrid
neural
network
based
PEMS
at
two
gas­
fired
combustion
turbines.
The
study
suggested
that
application
of
the
Part
75,
Subpart
E
statistics
to
a
smaller
data
set
when
coupled
with
a
threelevel
RATA
to
evaluate
the
PEMS
predictions
across
the
PEMS
"
operating
envelope"
is
a
good
measure
of
PEMS
performance.

Justification
for
Additional
Quality
Assurance
(
QA)
Testing
of
PEMS
A
NO
x
PEMS
is
a
piece
of
software
that
provides
an
indirect
determination
of
NO
x
emissions.
It
can
provide
an
accurate
indication
of
NO
x
levels
if
it
is
properly
developed,
trained,
and
qualityassured
Normally,
a
PEMS
is
trained
over
a
one
week
(
or
longer)
time
period
and
over
a
wide
range
of
source
operating
conditions.
However,
even
the
best
training
regimen
cannot
include
all
possible
operating
conditions,
e.
g.,
upsets,
sticky
valves,
or
other
unforeseen
events,
that
can
affect
emissions,
but
are
not
reflected
in
the
PEMS
output.

One
safeguard
against
this
is
to
implement
a
PEMS'
algorithm
that
identifies
potentially
failed
sensors,
or
PEMS
input
parameters
outside
of
the
expected
parameter
values,
by
comparing
each
sensor
to
several
other
sensors
and
determining
expected
sensor
values
based
on
the
historical
sensor
relationships
developed
during
PEMS
training.
When
such
sensor
values
are
identified,
an
alarm
is
provided
and
the
PEMS
is
out
of
control
and
the
maximum
potential
NO
x
emission
rate
must
be
reported
until
the
sensor
is
fixed
or
the
PEMS
is
retrained.
Reporting
standard
missing
data
values
or
allowing
a
substitute
sensor
value
calculated
by
the
PEMS
is
not
an
adequate
solution
because
the
PEMS
cannot
determine
whether
the
abnormal
input
parameter
value
is
caused
by
a
failed
sensor
or
by
some
new
region
of
operation
not
represented
in
the
PEMS
training
data
that
results
in
unusually
3
2
EPA
notes
that
these
cost
estimates
assume
that
in­
house
personnel
will
perform
the
monthly
RAAs,
using
portable
analyzers.
The
cost
of
the
RAAs
will
be
considerably
higher
if
outside
testing
contractors
are
used.

3
"
Evaluation
and
Field
Testing
of
Nitrogen
Oxide
(
NOx)
Predictive
Emission
Monitoring
Systems
(
PEMS)
for
Gas­
fired
Combustion
Turbines
­
Synthesis
Report",
The
Cadmus
Group,
Inc.,
December
29,
2004
.
high
NO
x
levels.

An
even
better
safeguard
against
unforeseen
events
that
can
affect
emissions
not
reflected
in
the
PEMS
output
is
to
periodically
compare
the
PEMS
output
to
a
quality
assured,
direct
measurement
of
stack
emissions,
e.
g.,
by
performing
a
RATA.
However,
RATAs
are
costly
and
are
generally
performed
only
once
or
twice
a
year.
Therefore,
other,
less­
expensive
accuracy
checks
should
be
done
in­
between
the
RATAs,
to
provide
ongoing
assurance
of
data
quality.
For
continuous
emission
monitoring
systems
(
CEMS),
the
RATAs
are
supplemented
by
daily
calibration
error
checks
and
quarterly
linearity
checks,
which
use
calibration
gases.
However,
these
tests
cannot
be
done
on
a
PEMS,
because
calibration
gas
cannot
be
injected
into
a
PEMS.
Therefore,
some
other
type
of
periodic
accuracy
check
suitable
for
a
PEMS
is
needed
to
supplement
the
RATAs,
in
order
to
adequately
quality
assure
the
PEMS
data
for
use
in
a
cap
and
trade
program.
EPA
believes
that
monthly
3­
run
relative
accuracy
audits
(
RAAs)
will
provide
the
necessary
additional
QA
for
a
PEMS.
The
results
of
the
aforementioned
EPA
study
have
shown
that
for
gas­
fired
combustion
turbines
with
NO
x
concentration
levels
of
3
ppm
and
higher,
portable
NO
x
and
O
2
analyzers,
or
Method
7E
are
suitable
for
these
accuracy
tests.

In
addition
to
examining
the
feasibility
of
using
portable
analyzers
for
periodic
accuracy
checks
of
PEMS,
the
EPA
study
assessed
the
potential
cost
associated
with
a
monthly
RAA
requirement.
EPA
estimates
that
performing
the
additional
monthly
RAAs
during
the
ozone
season
would
bring
the
total
annual
cost
of
operating,
maintaining
and
quality­
assuring
a
PEMS
(
such
as
the
PEMS
on
Unit
GTP1
)
to
approximately
$
24,750.
This
cost
includes
$
3,000
annualized
equipment
cost
for
a
portable
analyzer
plus
$
6,750
operation
and
maintenance
(
O&
M)
costs
associated
with
QA
testing
(
including
an
annual
9­
run
RATA
performed
by
an
outside
test
contractor
and
monthly
3­
run
RAAs
for
the
5
month
ozone
season
performed
by
in­
house
staff
using
a
portable
analyzer),
and
$
15,000
PEMS
O&
M.
This
represents
an
increase
of
about
$
5,750
above
the
cost
without
the
monthly
RAAs.

EPA
believes
that
the
cost
of
the
additional
RAAs
is
reasonable.
According
to
EPA's
CEM
Cost
Model,
the
next
least
costly
option
for
GTP1
(
if
it
loses
its
status
as
a
peaking
unit)
to
comply
with
subpart
H
of
Part
75
would
be
NO
x­
diluent
CEMS.
The
total
annual
cost
of
operating
and
maintaining
a
CEMS
is
estimated
at
$
62,700.
This
cost
includes
$
15,000
annualized
equipment
cost
plus
$
47,700
O&
M
costs
(
including
an
annual
RATA).
Thus,
even
with
the
additional
RAA
requirement,
the
estimated
annual
cost
of
operating
and
maintaining
a
PEMS
for
Unit
GTP1
would
be
less
than
half
the
cost
associated
with
CEMS2.

EPA's
field
test
data
analysis3
of
the
first
720
hours
of
matched
PEMS
and
CEMS
data
for
Chesterfield
Unit
8
and
the
830
matched
data
pairs
for
Cane
Island
Unit
2,
revealed
that
most
30­
point
subsets
of
paired
data
passed
the
same
combination
of
statistical
tests
that
all
the
data
in
each
data
set
passed.
The
field
test
data
also
illustrated
the
importance
of
testing
the
PEMS
over
the
range
of
unit
operation
because
of
the
strong
correlation
of
NO
x
emissions
to
certain
unit
operating
4
parameters.
Based
on
this
evaluation,
abbreviated
Subpart
E
statistical
tests,
in
conjunction
with
an
initial
three
load
RATA,
and
ongoing
QA
tests,
can
provide
a
reliable
indication
of
PEMS
performance.

This
final
approval
is
appealable
under
Part
78
of
the
Acid
Rain
regulations.
If
there
are
any
further
questions
or
concerns
about
this
matter,
please
contact
John
Schakenbach
of
my
staff
at
202­
343­
9158
or
at
(
schakenbach.
john@
epa.
gov).

Sincerely,

Sam
Napolitano,
Director
Clean
Air
Markets
Division
cc:
John
Schakenbach,
EPA,
CAMD
Louis
Nichols,
EPA,
CAMD
Constantine
Blathras,
EPA
Region
5
Karen
Kajiya­
Mills,
MI
DEQ
Attachment
