Uploaded by ajenkins

Quantifying The Impact Of Weather Measurement Errors On Line Ratings

advertisement
NERC Alert Task Force:
Guidelines for Determining Conductor Temperatures During
Measurement of Sag Along Overhead Transmission Lines
Quantifying The Impact Of Weather
Measurement Errors On Line Ratings
Alastair Jenkins
GeoDigital International Corp
February 7, 2012
Management of Existing Overhead Lines WG Meeting
Tampa, Florida
Quantifying
TheTest
Impact
of Weather Measurement Error
Field
Methodology
On Transmission Line Ratings
Background:
•
•
•
•
•
Various authors have questioned accuracy of the IEEE methodology for NERC
LiDAR surveys due to possible errors in weather data and a resultant implied
conductor temperature uncertainty.
Historical industry practice has been IEEE 738 with the deployment of remote
Ground weather sensors every few (typ. 15‐25 Km co‐located with GPS base
stations)
Industry introduced in 2009 an improved methodology, based upon the IEEE 738
standard & an Airborne Metrological Data Probe (ADP) providing better
metrological data than achievable using remote ground based weather sensors.
The methodology currently deployed by three independent LiDAR Suppliers has
been formally evaluated by EPRI as an effective methodology for collecting
weather observations during LiDAR surveys (results will be available from EPRI)
To allow wider ADP use GeoDigital is placing in the open literature the vertical
profile correction methodology and an error estimation methodology developed
to statistically predict residual errors in Conductor Temperature based upon
weather conditions at the time of acquisition
Air Data Probe - introduction
APPLICATIONS:
Without
going into the technical details, the ADP is a highly accurate weather station –
Technology is
• it Originally
developed
in support
of NASA
research
is just mobile
and above
the wires
rather
thanprojects,
below them
deployed to
•
ADP
technology
is
in
active
use
today
at
world
leading
organizations.
Accuracy:
support real
• In use at three leading LiDAR providers for Transmission Line Rating
time spraying to
– Wind direction +/‐ 20 Degrees
predict aerosol
– Wind speed +/‐ 0.5 m/s (1.6 ft/s)
and chemical
drift
– Ambient temperature +/‐ 1C (1.8F)
Air Data Probe – Application and Methodology
ADP measures temp and wind speed 150 – 300’ above wires. This
must be corrected to appropriate line height
A measured Vertical Profile and “wind industry” methodology is
used to adjust data to line height ( not all vendors currently do this)
Typical wind vertical profile is exponential near the ground, due to
surface roughness, which highlights the inaccuracy of typical 6’ high
ground weather stations.
If ground weather stations are used it is recommended they be
elevated to 30’ or higher.
Aerial Data Probe
Ground Base Stations
30 ‐ 50 ft
150 – 300 ft
Vertical Wind Profile - Modeling
Surface roughness and
Log wind speed law
Vh – wind speed at ground base station
VH – wind speed at target height
h – height of the ground station sensor
H – target height
z0 – surface roughness measure
r – ratio between wind speed at target and base height
Sheltering
z0
High above ground, above
water
Open area, short grass
Cropland
Sheltered, tall trees
0.005
0.01
0.1
1
7.6
6.9
4.6
2.3
Wind speed
ratio, r
Power‐law parameter
equivalent
1.27
1.30
1.54
3.32
1/7
1/6
1/4
1/1
6.0
5.3
3.0
0.7
Recent research into wind turbines
Has generated a wealth of analysis
Tools and statistical data related to
vertical wind profiles
"Wind Speed vs Height" correction function
6
m/s
log(x/z0)
H=10m
h=2m
4
2
0
0
12
24
36
48
60
power 1/1.4
72
84
96
108
power
1/7
AGL, m
120
132
144
156
Conductor Temperature Validation & Uncertainty
Estimation Study
Study Objectives:
• Validate accuracy of conductor temperature calculations and resulting sag
modeling using LiDAR survey methodology and Airborne Metrological Probe
measurements.
• Compare accuracy of ADP to local ( in span) & more remote ground weather
stations
• Determine and quantify residual uncertainty in Conductor temperature
• Determine and Quantify residual error in terms of Conductor Sag
• Determine limits of Technology and publish Best practices to allow wider use
Field Tests:
– Conducted in USA & Internationally during 2011
– Methodology tested on over 2,700 miles of lines with over 100 ground
weather stations deployed as control spaced every 10 mi
– Three lines were additionally flown multiple times as part of an EPRI test
program with multiple in span weather and line temp sensors as control
– Statistical conclusions are derived from an analysis performed on over
16,000 miles of transmission lines flown over a 3 year period
Meteorological & Line Load conditions – Data Base
Temporal extent: 2007-11-23…2011-12-04
Geographical extent:
See map
linear totals:
With weather parameters only:
31000mi, 1745 circuits, 164000 spans
With calculable conductor temperature:
16000mi, 624 circuits, 62000spans
With Ground reference stations:
over 2,700 mi (analysis ongoing)
Meteorological conditions – The Data Base in practice
Survey Conditions.
Temperature by Season
Solar Radiation Distribution
winter
spring
summer
autumn
‐4
2
8
14
20
C
26
32
38
44
0
150
300
450
600
750
900
1050 1200 1350
W/m2
Survey Conditions By Season.
Wind Speed, Normalized
winter
spring
summer
autumn
0
1.2
2.4
3.6
4.8
m/s
6
7.2
8.4
The Data sets shows the
normalized distribution of
temperature wind and solar
load experienced ‐ data covers
all 4 seasons.
No attempt was made to
constrain operations to specific
weather conditions other than
as required for safe rotor craft
operations
Meteorological conditions EPRI Trials vs
Entire test Data Base
EPRI FIELD TRIAL DATA ‐ 3 days 3 lines
•
•
•
•
•
Ambient air temp:
Wind speed:
Load:
Current density:
Conductor temp above ambient
[9...20] °C
[0.8...4.5] m/s
[42…282] amps
[0.1‐0.9] amps/kcmil
[2.3..13.1] °C
Distribution of Conductor Over
Ambient Air Temperature Excess
For total Data Base of 16,000 mi
069kV
138kV
161kV
230kV
345kV
0
1
2
3
4
5
6
7
C0
8
9
10 11 12 13
500kV
Given the line‐load, span
location, cable type, survey
time and weather observations,
the individual conductor
temperature for each span was
calculated
Weather data was available on
a span by span basis thru use of
an Air Data probe
supplemented by ground
weather stations
The Data set demonstrates that
in practice while line
temperature varied
significantly the delta above
ambient was typically relatively
small consistent with operation
under low to moderate loads
Conductor Temperature Validation & Uncertainty Estimation
Methodology
•
•
•
•
•
•
Surveys: Flew multiple areas with different terrain, sheltering, conductor
types, ages, and load conditions
Acquisition: Test lines flown multiple times under different weather and load
conditions –weather data acquired using local, remote and Air data Probe
sensors
Production: Observations input into PLS‐CADD’s IEEE‐738 conductor
temperature calculator – dynamic model used when significant load changes
Analysis: Sag modeled under different load conditions and compared to
calculations from other flights of the same spans with different conditions
Temp Validation: Conductor Temperature Validated in specific tests by
conductor mounted sensors (EPRI and OLTM sensors were used during site
specific trials)
Sag Validation: Conductor Sag validated by comparison of LiDAR Data to
predicted position of the conductor with load and weather conditions during
the re‐flights
Conductor SAG Validation & Uncertainty Estimation
•
•
•
•
•
PLS‐CADD model is built using the first flight and observed sag
Temperature data from ADP/IEEE modeling is input into PLS‐CADD
Difference between modeled and observed sag is compared
Accuracy of fit directly correlates to ultimate rating accuracy
For scale reference the sag difference shown is 4ft on this 600ft span
Flight 1 – fit model to
observed sag and Temp
Flight 2 – sag model
from Flight1 to
temperature from Flight
2 & compare fit
Flight 3 – sag model
from Flight 1 to
temperature from Flight
3 & compare fit
Conductor SAG Validation & Uncertainty Estimation
• Sag errors observed on multiple spans were very similar
• Differences of less than 5cm (2 inches) observed – these are
not significant as up to 2cm is likely due to laser range noise
• Result provides high confidence in the subsequent rating
accuracy (Sag error will be similar i.e. a few cm at max sag)
GeoDigital IEEE Sag Discrepancy, cm
Line
Beaver‐
Line
1
Henrietta
Perry‐
Line
2
Ashtabula
Falconer‐
Line
3
Warren
Comparison RMS Avg
Min Max Max(abs)
Flight1 to 2
3.4
‐2.5
‐6.0
0.0
6.0
Flight1 to 3
1.2
0.3
‐1.0
2.0
2.0
Flight1 to 2
3.9
2.9
0.0
7.0
7.0
Flight1 to 3
3.0
0.6
‐3.0
7.0
7.0
Flight1 to 2
4.5
4.5
4.0
5.0
5.0
Flight1 to 3
1.4
1.0
0.0
2.0
2.0
Note: Data is in “Centimeters”
Max As
Observed Sag
Change
7.0
Sag 16.0
error
7.0 2 inches
Less than
RMS12.0
12.0
18.0
Conductor Temperature Validation – Using Line
Thermocouple Sensors
Plot compares predicted
value with Line mounted
temperature sensor
Sensor accuracy +/‐ 1deg C
Residual Error Estimation (confidence interval)
• We have published methodology
and a residual error estimator
(confidence interval calculator)
• Inputs to IEEE‐738 are the
following parameters:
– Cable properties (diameter, material,
emissivity*, resistance and others)
– Line load in amps
– Date, time and location, used to estimate
solar radiation if not measured
– Meteo parameters:
•
•
•
•
•
Wind speed
Wind direction, relative to span direction
Ambient air temperature
The ADP is designed to allow efficient
collection of the meteo‐parameters
required above, and in the immediate
vicinity ,of the transmission line.
All measurements will have a residual
error and error distribution
* Note: Emissivity estimates as applied in IEEE‐738 are not a major contributor to
temperature uncertainty calculations.
Uncertainty estimation
method description
Knowing the likely measurement error and error statistics (normal or
other statistical distribution ) it is possible to calculate using the IEEE
methodology not only the conductor temperature but also a
confidence interval.
The method has been implemented in a web application that performs
a Monte‐Carlo analysis scheme:
To calculate conductor temperature confidence interval:
‐ Generate N (~1000) IEEE input variables drawing from respective
distributions
‐ Perform deterministic temperature calculations
‐ Obtain lower and higher percentiles at the required confidence
levels
The reliability of the derived interval depends on the accuracy of the
assumptions made on input variables distribution types and
parameters and can be used for all source data – not just an ADP.
Residual Error Estimation
• We have created a
web calculator to
model conductor
temperature
accuracy
• URL available upon
request and
feedback
welcome!
www.geodigital.com/calculator
Predicted intervals are valid
Assumed stdevs of the inputs :
• Temperature = 1 °F
• Wind Speed = 2.1 ft/s
•
•
•
Wind Direction = 30°
Emissivity/absorptivity = 0.15
Solar Radiation = 70 W/m2
We’ve calculated T0 and expected confidence interval [lower,
upper] at 67% and 95% confidence levels.
62% and 92% of GT measurements actually within those intervals
‐ meaning that assumption of data inputs was only slightly
optimistic.
Best Practice Deliverable
IT is now possible to provide Utilities with not only a calculated Conductor temperature
But also a confidence interval specified for any desired level of certainty.
Note : Span by span weather observations
and conductor temperatures with confidence intervals
Comparison of various meteo sensor performances
Mean and Range of Cond Temp Error computed using
various meteo data sources
12.0
Temperature Error, °C
10.0
8.0
6.0
4.0
2.0
0.0
‐2.0
‐4.0
Mean 2.2
RMSE 4.9 Deg. C Mean 0.8
RMSE 3.6 Deg. C Mean 1.3
RMSE 3.5 Deg. C Mean ‐0.2
RMSE 1.8 Deg. C Mean 0.11
RMSE 1.3 Deg. C
All public
Only airports User‐installed at
5‐10mi
ADP
EPRI In Span
‐6.0
‐8.0
‐10.0
‐12.0
ADP error is significantly smaller than current “Industry Practice ”
and similar to in span elevated sensors at a fraction of the cost &
should be considered as a “Best Practice”
Practicality of Method and notes
• Analysis accounts for all (most) of the considerations
expressed previously at IEEE panel presentations
• We concur with the EPRI notes (Cigre TB‐299)
recommendation related to restricting wind direction
to 20‐70 deg.
• Confidence interval calculation is simple and rapid to
implement ‐ (eliminates uncertainty)
• Provides immediate indication of spans that may (due
to extreme circumstances) require special attention
due to higher temperature uncertainty
• Calculator may be extended to predict sag uncertainty
if span length is input as a parameter
Line Re-rating Workflow
Conductor Temperature Uncertainty
Conductor Temperature Uncertainty
Distribution by Wind Speed
±0.50m/s
±0.75m/s
±1.00m/s
0
1.2
2.4
3.6
4.8
6
7.2
C0
67%
95%
Temperature Uncertainty due to Wind speed uncertainty, ±x m/s @ 95%
0.25
0.5
0.75
1
1.25
1.5
1.75
0.3
0.7
1.1
1.5
1.9
2.4
2.9
0.8
1.7
2.5
3.5
4.6
5.9
7.5
Real world data demonstrates conclusively that
conductor Temperature uncertainty during LiDAR
surveys can be constrained
2
3.6
9.7
Given the individually observed
load and weather conditions it
is possible to use the model to
predict (for the entire data set)
the probability of a given
Temperature uncertainty due to
changes in wind speed or
direction
A line temp accuracy of less
than +/‐ 5 Deg C 95%
confidence is eminently
achievable this results in a few
Cm of possible Max sag error
Of more importance however is
that for any line flown at any
specific time the actual
uncertainty can be calculated
and provided to the engineer
for use in the ratings analysis.
Conclusions
•
•
•
•
•
While small errors in effective wind speed can create significant changes in
the final sag of a line when operating at close to maximum load, similar
errors during a LiDAR surveys (with more typical operating loads) will not
significantly change the initial or final sag.
Use of an Air Data Probe to determine local wind speed and direction rather
than multiple weather stations has been shown to provide improved
accuracy and accurate Line temperature calculations at low incremental cost
Line temperature uncertainty can and should be assessed by the LiDAR
survey company using a statistical analysis and sections re flown or re‐
evaluated on the rare occasions when the line load and average wind speed
is very unfavourable resulting in possible errors greater than acceptable
ADP when operated and calibrated appropriately has demonstrated ability to
match Line mounted sensor temperatures within +/‐ 2 Deg C RMSE or less
this is comparable to in span line height sensors and significantly better than
errors claimed by other new technologies
Line sag errors during LiDAR surveys are relatively small – not large, Industry
needs to focus on improving predictions at max sag not during survey
Download