Skip Maine state header navigation
BANGOR
HYDRO-ELECTRIC COMPANY
PROFILING
METHODOLOGY
FEBRUARY
1, 2000
This
report details the methodology Bangor Hydro-Electric Company will employ to
estimate the hourly loads of Competitive and Standard Offer Providers supplying
energy within its territory after retail competition begins on March 1, 2000.
BACKGROUND
In
order for ISO-NE to settle the electricity markets in New England each day,
ISO-NE must be able to assign load responsibility to the various
suppliers. Local Distribution Utilities
have been deemed to be in the best position to determine the hourly loads of
Suppliers operating within their territories.
For
customers with loads of sufficient magnitude to justify the expense of daily
polling of their meters, actual loads are able to be determined. In order for other customers to participate
in retail choice, their hourly loads must be estimated via Load Profiling.
CHOICE
OF PROFILING METHODOLOGY
Bangor
Hydro-Electric, Central Maine Power, Maine Public Service, and the MPUC agreed
that the utilities would use the same load profiling methodology for the three
profile groups defined in Chapter 321.
Each utility will generate static profiles that represent a typical
weekday or typical weekend day for each month of the year for each of the three
profile groups. Holidays will be
modeled using a weekend profile.
The typical day profiling approach is both straightforward and predictable. Use of this method helps ease the transition into the competitive market for both Bangor Hydro and suppliers. However, other methods of profiling are available. Although the typical day profiling methodology will be used for the first year of retail competition, other methodology types will be investigated. The MPUC requires that the Maine utilities file a report by June of 2000 that will contain the results of this investigation. Any recommended changes in profiling methodology or profile groups would become effective March 1, 2001.
BUILDING
PROFILES
The
three profiles will be built using a software tool called Load Vision (ICF
Consulting Group, Inc). Load Vision is
a load estimation, settlement, and imbalance pricing system. It calculates estimated hourly loads for a
given population of customers and their energy suppliers. Load Vision estimates the loads of each
customer individually, either by applying profiles that it creates or by
accepting actual loads from interval metered customers.
Interval data from participants in the three load research studies will be processed within MV-90, a meter data translation software system. Each non-normal interval has an associated status code. When the interval data files are sent to Load Vision, these status codes dictate whether the non-normal intervals are used in the calculation of average hourly loads. It is in this way, for example, that Load Vision can distinguish between a legitimate zero usage interval and one resulting from a power outage or data overflow. Files are sent to Load Vision in EU format for each sample point once a month. Data will be exported to Load Vision in 60-minute format.
A
number of steps are involved in building profiles within Load Vision. See Appendix A for an explanation of these
procedures.
ESTIMATION
OF LOADS
Chapter
321 (MPUC) requires that all customers of Bangor Hydro with demands of 500 kW
or greater must be telemetered. In
addition, a supplier may request that any of its customers be telemetered
rather than have a load profile applied.
Bangor
Hydro will poll each telemetered customer’s meter at some point after midnight
and before 6 a.m. each day. Actual
hourly loads for the previous day will be collected. These actual loads, with the addition of appropriate losses, will
be assigned to the customer’s Supplier.
Polling of the meters will be performed by MV-90.
Actual
data may not be available due to problems with remote meter reading equipment
or meter malfunctions. In these
instances, Load Vision will estimate data based on individual profiles. Each interval customer will have
individualized load profiles created which will be based solely on that
customer’s own historic usage. In the
event of missing data, gaps will be filled by applying these individual
curves. Later, if the data becomes
available, the estimated data will be replaced by actual. This approach allows
the estimate to be based on the customer's actual historical load, not a class
profile.
For
non-metered loads, such as street and area lighting, a deemed profile is
applied which is based on the number of hours of darkness occurring each day.
All other non-interval metered Bangor Hydro customers are assigned to one of
the three profile groups. Estimating
customer loads with profiles is performed in two steps. First, a usage factor is calculated for each
customer for the time period for which loads are to be estimated. A usage factor is the kWh consumed by a
customer for a given period of time divided by the kWh for the profile to which
the customer is assigned over the same period.
Next, the usage factor is applied to the profile to which the customer
is assigned. Each customer’s kWh usage
is spread across the time period to be estimated by multiplying the usage
factor by the hourly profile values.
Each
of the three Maine electric utilities – Bangor Hydro, Central Maine Power, and
Maine Public Service – are employing the same profiling methodology and the
same estimation software tool. For a
detailed explanation of how Load Vision performs the estimation process, see
CMP’s December 1, 1999 methodology report.
The relevant section is attached to this document as Appendix B, which
also contains an explanation of the month-end adjustment procedure.
REPORTING
Bangor Hydro will transmit estimated hourly loads to ISO-NE by Load Asset ID within 37 hours after the estimated day ends. At the same time, Bangor Hydro will send to each supplier its own estimated loads, again by Asset ID. The month-end adjusted figures will be reported within 90 days of the end of the month. ISO-NE will receive a revised total Megawatt figure, while suppliers will receive revised hourly numbers.
SUMMARY
Bangor
Hydro-Electric Company will estimate the hourly loads of all suppliers
operating within its service territory after retail competition begins on March
1, 2000. It will do so via a
methodology which combines collecting actual load data from interval metered
customers and by applying profiles adjusted by energy usage factors for all
other customers.
Appendix
A
The Day Type technique produces a series of typical
season/day-type profiles from historical interval data. To create a day-type profile, the user first
selects the Day Type as the Load Shape Representation on the Options for
Creating Load Profile window. Once the
user has selected the Day Type representation type and defined the desired
season/day types a load profile is created for each season/day-type.
To
view the profiles, the user clicks on the View Profile button on the Select
Segment for Load Profile Analysis window.
The Representation Viewer window will open and a Load Shape Libraries
file folder will be visible. When the user
clicks on the folder, or on the small box beside it, the folder will open, and
the associated season/day-types book icons will appear. To view the profile for a specific
season/day type the user clicks on the book icon associated with the season/day
type. The book will open, and the graph
will be displayed in the workspace. The
twenty-four hours for the profile are plotted on the horizontal axis and the
associated loads are plotted on the vertical axis as a bold line. The weighted average of the actual interval
data points used to create the profile are plotted as thin lines. The colors of the profile graph and the
actual interval data graphs can be adjusted on the Graph tab of the Load Vision
Settings.
Given
sets of 24 hour loads for each day of the year defined in the calendar, the
next step toward computing a 24 hour load shape for each set of days defined in
the first step is to average the hourly loads in chronological order. That is, the loads in each day’s first hour
are averaged to determine a first hour average load. Then the loads in each day’s second hour are averaged, and so
forth until the loads in all 24 hours are averaged. The result of this process is an average daily load shape.
Although
the average daily load shape may indicate the hour in which the average peak
occurs, and the average trough it also results in a flattened load shape. Therefore, this average load shape’s peak is
generally too low, and trough is generally too high. That is due to the fact that the peaks and troughs do not occur
in the same hour in each day of the averaged set of days. This flattening result is applicable to all
hours of the average daily load shape.
Therefore, this average daily load shape can be used to indicate the
ordering of the hours based upon load magnitude, but not the absolute load in
each hour.
Step 1: Group loads based on
season/day-type combination.
The
user will define the season and day-type structure to be used. The load values will be grouped into one of
the 12 bins. Holidays will be put into
the Sunday bin in this example.
Example:
4
Season/3 Day-Type
Season
1 (Winter): Dec, Jan, Feb
Season
2 (Spring): Mar, Apr, May
Season
3 (Summer): Jun, Jul, Aug
Season
4 (Autumn): Sep, Oct, Nov
Day-Type
1 (Weekday): Mon, Tue, Wed, Thurs, Fri
Day-Type
2 (Saturday): Sat
Day-Type
3 (Sunday): Sun
Step 2: Create a weighted
average of all observations for the profile for each day.
·
Get
list of StudyIDs and weights from the Segment_Weights table for the profile.
·
Get
actual observations from the .Interval file for the StudyIDs listed in the
Segment_Weights table.
·
Create
the weighted observation (2 options).
Example:
Weights
from Segment_Weights table:
|
Segment |
StudyID |
Weight |
|
RES |
123 |
1 |
|
RES |
456 |
1 |
|
RES |
789 |
1 |
Actual
Observations from the .Interval File:
|
StudyID |
Day |
Hour1 |
Hour2 |
Hour3 |
Hour4 |
… |
|
123 |
12/1/98 |
30 |
50 |
40 |
60 |
… |
|
456 |
12/1/98 |
35 |
55 |
45 |
65 |
… |
|
789 |
12/1/98 |
40 |
60 |
50 |
70 |
… |
Weighted
Observations:
Option
1: Divide by the Sum of Weights
WL
(p,s,d,h) = S (L (p,s,d,h) * W (p)) / S W (p)
Where:
WL
(p,s,d,h) = Weighted load for the
profile (p), season (s), day-type (d), and hour (h)
L
(p,s,d,h) = Load for the profile (p),
season (s), day-type (d), and hour (h)
W
(p) = Weight for the profile (p)
Option
2: Don’t Divide by the Sum of Weights
WL
(p,s,d,h) = S (L (p,s,d,h) * W (p))
Where:
WL
(p,s,d,h) = Weighted load for the
profile (p), season (s), day-type (d), and hour (h)
L
(p,s,d,h) = Load for the profile (p),
season (s), day-type (d), and hour (h)
W
(p) = Weight for the profile (p)
Result
(this is what will be seen graphically when looking at the profile data)
|
Segment |
Day |
Hour1 |
Hour2 |
Hour3 |
Hour4 |
… |
|
RES |
12/1/98 |
35 |
55 |
45 |
65 |
… |
|
RES |
12/2/98 |
60 |
70 |
50 |
55 |
… |
|
RES |
12/3/98 |
55 |
70 |
55 |
60 |
… |
|
RES |
12/4/98 |
32 |
60 |
45 |
50 |
… |
Step 3: Average the load
values to create an Average Load Shape
ALS
(p,s,d,h) = S L (s,d,h) / # observations
Where:
ALS
(s,d,h) = Average load shape for the season (s) , day-type (d), and hour (h).
L
(s,d,h) = Load for the season (s) , day-type (d), and hour (h).
S
= season
D
= day-type
H
= hour
#observations
= number of days in the observation for the season (s) , day-type (d), and hour
(h).
Season
1, Day-Type 1
|
Segment |
Day |
Hour1 |
Hour2 |
Hour3 |
Hour4 |
… |
|
RES |
12/1/98 |
35 |
55 |
45 |
65 |
… |
|
RES |
12/2/98 |
60 |
70 |
50 |
55 |
… |
|
RES |
12/3/98 |
55 |
70 |
55 |
60 |
… |
|
RES |
12/4/98 |
32 |
60 |
45 |
50 |
… |
Result:
|
Segment |
Day-Type |
Hour1 |
Hour2 |
Hour3 |
Hour4 |
… |
|
RES |
Weekday |
45.5 |
63.75 |
48.75 |
57.5 |
… |
Step 4: Sort the load values
in descending order to create a Load Duration Curve
Example:
Season
1, Day-Type 1
|
Segment |
Day |
Hour1 |
Hour2 |
Hour3 |
Hour4 |
… |
|
RES |
12/1/98 |
65 |
55 |
45 |
35 |
… |
|
RES |
12/2/98 |
70 |
60 |
55 |
50 |
… |
|
RES |
12/3/98 |
70 |
60 |
55 |
55 |
… |
|
RES |