You are on page 1of 6

Improvement of DT Load Estimation by Using

Typical Load Profiles


Andrew Charles D. Yang1, Ronoel M. Dellota2
Networks, Manila Electric Company
Pasig City, Philippines
1
acdyang@meralco.com.ph
2
rmdellota@meralco.com.ph

Abstract— This paper proposes a new methodology for single which costs time and additional operational expense. Hence,
distribution transformer (DT) load estimation using typical there is a need to explore other load estimation methodologies
customer load curves. The process described in this paper is an that give more accurate results.
improvement over regression-based formulas to properly
estimate DT loads. The results of the methodology are compared
and verified to the output of LP-capable meters installed at DTs.
In this study, the feasibility of using the typical customer
load profiles for transformer load estimation will be explored.
KVA loading will be computed using the aggregate loading of
Keywords— distribution, transformer, meter, DT, TLMS, load the connected customers.
profile
II. RELATED WORKS
I. INTRODUCTION
A load profile is a model of load characteristics represented
Transformer load monitoring is an important aspect of by parameters such as customer types, day/season and
electric distribution utility operation. An accurate and reliable temperature. The load profile is used in power generation to
load monitoring system enables system planners and operators monitor and plan their generation schedule. For the
to determine transformers that are overloaded, normally transmission system, load profiles are utilized for forecasting
loaded, under-loaded and idle. demand and system planning. Furthermore, the distribution
utilities make use of load profiles to enhance the operation
If properly identified, overloaded transformers can be efficiency and reliability of their facilities. Load profiles can
readily scheduled for replacement. Hence, DT failures due to also be used for load balancing and customer billing in a
overloading can be prevented. Under-loaded transformers can deregulated environment [1].
be removed and replaced with the proper rating, thereby
maximizing asset utilization. Idle transformers can be The importance and scope of the applications of load
removed and used as new stocks, thereby, saving purchase profiling lead to numerous researches related to the subject.
costs for new transformers. Most of these studies are targeted to seek out the most
accurate way to model the Typical Load Profile (TLP)
There are different methods of Transformer Load classification. We can categorize the classification techniques
monitoring. The most accurate method to monitor the load of into two. The first method is derived from the shape of the
transformers is to install indicating demand (ID) meters that load curves while the second method is derived from a pre-
register the kW or KVA demand. Because of the number determined set of consumers. Most of the papers that fall
distribution transformers (DT) installed in the field, this is not under the first category introduced a fuzzy clustering
practical due to the significant investment required. However, technique to group similar load curves ([1], [2]). In [1],
installation of ID meters can be justified in areas with high Measured Load Profile (MLP) is classified according to
incidence of DT failures due to overload. hierarchical clustering method and fuzzy logic. In [2], a two
stage FCM (Fuzzy C-Means) is introduced by grouping
Other methodologies of transformer load monitoring deal through load pattern and value. The second category wherein
with load estimation techniques. One particular method is by [3] and [4] belong retrieves data by sampling theory and
using an empirical formula based on regression to estimate the applies statistical analysis to develop the typical load profiles.
DT KVA loading given only the kWh consumption of
connected customers. This formula categorizes transformers The distribution transformer is the most crucial component
into three groups: residential, mixed residential and of the distribution system, customer wise. It bridges the gap
commercial. From this categorization, constants are obtained between the distribution utility and the consumer by stepping
for use with the formula. Currently, our experience with a down the distribution-level voltage to utilization voltage levels
regression based DT load estimation methodology has shown that are required by the customer. Thus, the loading level of
that the results are only 64% accurate. This necessitates the transformer should always be monitored in order to
further load testing on-site to verify overloaded transformers, maintain its good working condition and to ensure the
continuous supply of power to customers. To monitor For completeness, the methodology used in establishing the
accurately, the health and reliability of all distribution load profiles used shall be discussed along with the TLP DT
transformers, an ID (Indicating Demand) meters can be used load estimation methodology.
to record the peak load. However, given the number of
distribution transformers in the field, to put up ID (Indicating A. Development of Typical Load Profile
Demand) meters for each is not practical. Non-invasive digital
load loggers, on the other hand, can also be employed to The foundation of the whole methodology is the
estimate the loading on the transformer with <5% accuracy development of typical load profiles for use with the
but with limitations on storage capacity, load loggers can be estimation. Individual load profiles are developed for each
impractical as well. customer rate class following the process flow outlined in Fig.
2.
Manually, the loading of single phase distribution
transformers can be computed by converting kWh to kW Per Customer Hourly
using regression analysis. However, the accuracy of using this Readings
method is 60% i.e., the estimated DT loading is within ±10%
of the actual DT loading value 60% of the time only. This
inaccuracy may be attributed to 4 factors:
Average of the hourly readings
computed across all customers in
the rate class
A. Constants derived to determine DT peak load
B. Encoding of customers connected to the facility
C. Encoding of transformer capacity details
D. Power pilferage not accounted for by the formula Averaged Hourly
Readings
To address these problems, TLP is used to solve DT
estimation. Load Profile (LP) capable meters are used to
obtain the load profile of all customers connected to a
distribution transformer, as in [5]. The customer demand data Averaged Hourly Readings are
gathered through this activity are used in the development of a normalized based on peak load
new DT load estimation technique, using typical load profiles
for different rate classes as its basis.

III. METHODOLOGY
Typical Load Profile
The typical load profile (TLP) methodology for DT Load
Estimation is divided into three main steps, namely: 1).
Development of typical load profiles; 2). Application of Fig. 2. Flowchart of Typical Load Profile Development
customer kWh consumption to the typical load profile and 3).
Aggregation of demand profiles and DT load estimation. The Customers are classified according to type: residential,
typical load profiles used in this methodology are stratified by industrial and commercial. These customer segments are
rate classes, which also determine the pricing schedule for further stratified based on their average kWh consumption as
different customer segments. Fig. 1 is a graphical shown in Table I for residential and general services
representation of the whole scheme. (commercial) customers, or their monthly kW demand as in
Table II for industrial and Table III for large commercial
customers, respectively.

TABLE I
RATE CLASS STRATIFICATION (CONSUMPTION)

Rate Class kWh Range Rate Class kWh Range


1-50 kWh 1-50 kWh
51-70 kWh 51-70 kWh
71-100 kWh 71-100 kWh
101-200 kWh General 101-200 kWh
Residential Services
201-300 kWh (Commercial) 201-300 kWh
301-400 kWh 301-400 kWh
Fig. 1. DT Load Estimation Methodology Using Typical Load Profiles
401-800 kWh 401-800 kWh
Cons.>800 kWh Cons.>800 kWh
TABLE II
RATE CLASS STRATIFICATION (DEMAND)-INDUSTRIAL
Residential, Cons. = 301 - 401 kWh
Rate Class kW Demand Range 1.2

5kW-40kW 1

41kW-200kW

Per Unit, P.U.


0.8

Industrial 201kW-2000kW 0.6

0.4
2,000kW - 10,000kW
0.2
10,000kW<Demand
0

00

00

00

00

00

0
:0

:0

:0

:0

:0

:0

:0
1:

3:

5:

7:

9:

11

13

15

17

19

21

23
Time
TABLE III
RATE CLASS STRATIFICATION (DEMAND)-COMMERCIAL
Fig. 3. Typical Load Profile (Residential, Consumption = 301-
Rate Class kW Demand Range 400 kWh)
5kW-40kW
B. Application of Customer kWh Consumption to the Typical
Large 41kW-200kW Load Profile
Commercial 201kW-2,000kW
Demand>2,000 kW The process flow of the TLP methodology is shown in Fig.
4.
For each segment, customers are selected based on a
Monthly Customer
sample size computation of 90% confidence with a 10% kWh Consumption
margin of error. These customers have been installed with
load profile (LP) capable meters and their monthly kWh
demand recorded.
Divide by the number of billing
days to obtrain the approximate
We let cn Є C and n Є N where N is the number of elements in Daily kWh Consumption
C. Also, we let h be any number between {1…H} where H is
equal to 24. The recorded load profile for each customer is
normalized using (1). Apply the Daily kWh Consumption
Hourly kW
Streetlight Load
to the Typical Load Profile to get
Profile
Customer Hourly kW Load Profiles
kW h
PU hcn =
max( kWh ' ) (1)
Aggregate all Customer Hourly kW
Load Profiles connected to the
The normalized load profiles of each customer segment are transformer, the Streetlight Load
averaged hourly to obtain a typical load profile to represent and System Loss

that particular rate class as shown in (2).

N Aggregate DT Load
Profile
∑ PU
n =1
cn
h
PU h =
N (2) Take the Peak kW Demand from
the Aggregate DT Load Profile and
multiply it with the transformer’s
The hourly average of all customers in a particular power factor

customer group is normalized once again using (3) to obtain


the typical load profiles for use with the TLP DT estimation
DT Peak KVA = DT
methodology. Load Estimate
Fig. 4. Typical Load Profile (Residential, Consumption = 301-400 kWh)
PU h The monthly kWh consumption of each customer
LPh = connected to a distribution transformer is taken from the
max( PU h ' ) (3) customer load database. The monthly consumption is divided
by the number of billing days between meter readings to get
A sample typical load profile (normalized) for residential the daily kWh consumption, as shown in (4).
customers with monthly kWh consumption of 301 to 400 kWh
is shown in Fig. 3.
kWhmonthly Peak Total kWhour
kWhdaily = (4) Peak kVA = (8)
billing days pf

The approximate load profile, that is, the hourly kW load For (8), a power factor (pf) of 0.85 lag is assumed in the
per customer for a 24-hour period, is obtained using (5). calculation. As DT load estimates are commonly expressed as
a percentage of total transformer capacity, the Peak DT kVA
LPh load (%) is taken from (9).
kWhour = kWhdaily x H
(5)
∑ LP h
Peak kVA
h =1
Peak DT kVA % = X 100 (9)
Rated DT kVA
The load profile to be used in the above formula
corresponds to the rate class of the customer, as obtained from The Peak kVA value from (8) is analogous to the peak kVA
(3). Note that the load profile represents a typical day only. demand of the transformer, whereas the percentage obtained
in (9) is equivalent to the percent ratio of the peak DT kVA
In addition to customer loads, the effect of streetlight loads demand to the rated DT capacity. Hence, this is the resultant
and system loss on the total transformer loading must also be DT load estimate.
taken into account.
IV. RESULTS AND EVALUATION
Streetlight loads connected to the transformer, if any, are In developing the typical load profiles, load profile data
assumed to be 70W high pressure sodium (HPS) luminaries from 1,730 customers installed with LP-capable meters are
for residential transformers and 250W HPS luminaries for collected and analyzed. The gathered data was recorded for a
commercial transformers. Streetlight loads are assumed to period of one (1) year from November 2004 to October 2005,
operate between the hours of 1800H to 600H and the total and from these, typical load profiles for each rate class have
streetlight load is distributed within this period. Thus the been developed. Fig. 5, Fig. 6, Fig. 7 and Fig. 8 show the
streetlight consumption is added to the hourly kW load of the typical load profiles for the residential, general services
DT for the said range. The hourly streetlight load is denoted (commercial), large commercial and industrial rate classes
as STLhour. respectively.

The technical loss of the transformer should consider the Typical Load Profile of Re s ide ntial Custom ers

consumption of the connected customers and the incident 1.20 0-50 kWh
streetlight load. For the study, a technical loss of 2% is 1.00 51-70 kWh
Per Unit P.U.

assumed. Hence, the calculation of the hourly losses is 0.80 71-100 kWh
expressed as in (6): 0.60
101-200 kWh
0.40
201-300 kWh
0.20
Losses hour = ( kWhour + STLhour ) x DT Loss (6) 0.00
301-400 kWh

401-800 kWh
00

00

00

00

00

0
:0

:0

:0

:0

:0

:0

:0
5:

7:
1:

3:

9:
11

13

15

17

19

21

23

Cons. >800
The hourly streetlight load and DT losses obtained in (6) are Time kWh
to be used in the TLP load estimation:
Fig. 5. Typical Load Profiles for Residential Customers
C. Aggregation of Demand Profiles and DT Load Estimation
Typical Load Profile for General Services (Commercial)
Customers
The total hourly kW loading is obtained as the aggregate of 0-50 kWh

the hourly kW load, the streetlight load, if any and the hourly 1.20 51- 70 kWh
Per Unit P.U.

1.00
kW loss, also known as the Aggregate DT load profile. This 0.80 71-100 kWh
is also expressed as in (7). 0.60 101-200 kWh
0.40
0.20 201-300 kWh
0.00
301-400 kWh
Total kWhour = kWhour + STLhour + Losses hour (7)
00

00

00

00

00

0
:0

:0

:0

:0

:0

:0

:0

401-800 kWh
1:

3:

5:

7:

9:
11

13

15

17

19

21

23

Time Cons. >800


kWh
One may begin to obtain a DT load estimate from the
aggregate DT load profile. To do so, the Peak kVA load of Fig. 6. Typical Load Profiles for General Services Customers
the transformer is computed by converting the max. total
kWhour load to its kVA load equivalent, using (8).
Typical Load Profile of Large Com m ercial Cus tom e rs
TABLE IVV
RESULTS COMPARISON (TLP VS REGRESSION)
1.20
Per Unit P.U.
1.00 % Deviation from Actual Metered DT
5-40 kW Demand (Over the Study Period)
0.80
41-200 kW
0.60
201-2,000 kW Trafo Type Regression TLP
0.40
2,000 kW< Demand Formula Methodology
0.20
0.00 1 Residential 16.83% - 29.2% 0.18% - 14.31%
00

00

00

0
:0 2 Residential 0.08% - 17.94% 6.34% - 19.81%

:0

:0
:0

:0
1:

4:

7:

10

13

16

19

22
Time 3 Commercial 77.18% - 130.27% 15.37% - 26.00%
4 Commercial 2.88% - 8.44% 18.76% - 24.02%
Fig. 7. Typical Load Profiles for Large Commercial Customers
Based on our side-by-side comparison of the demand
estimates of both techniques against the actual metered
Typical Load Profile of Indus trial Custom ers
transformer peak demand, on the average, the results of the
1.20 TLP methodology are more precise than the regression
1.00 5-40 kW formula. Although there are instances where the regression
Per Unit P.U.

0.80 41-200 kW
formula is more accurate than the TLP methodology and vice-
0.60 201-2,000 kW
0.40 2,001-10,000 kW
versa, the TLP methodology provides more consistent results
0.20 10,000 kW< Demand that range within 20% of the actual DT loading value for
0.00 residential DTs and 26% of the actual DT loading for
commercial DTs. The regression formula on the other hand,
00

00

00

0
0

:0
:0

:0

:0

:0
1:

4:

7:

10

13

16

19

22

Time
provides results that range from 30% of the actual DT loading
for residential DTs, and 130% of the actual DT loading for
Fig. 8. Typical Load Profiles for Industrial Customers commercial DTs.

To validate the accuracy of the developed load profiles V. CONCLUSIONS


against the actual measured transformer loading, the TLP The load profile methodology for DT load estimation
Load Estimation methodology is applied to a representative shows a lot of potential based on the results of our data
transformer and the resultant aggregate transformer load analysis.
profile is superimposed against a plot of the actual metered
transformer loading, shown in Fig. 9. From observation, the Overall, the LP methodology is more precise than the
TLP demand profile closely approximates the shape of the regression formula for DT load estimation, as the results are
actual metered DT loading, although the most crucial point for much closer on the average to the actual metered demand than
the load estimate, which is the peak demand occurring at the regression formula. But still, there is a lot of room for
23:00 shows a deviation of around 13.2% from the actual improvement of the methodology’s accuracy, which shall be
transformer peak demand, as is with the test case. the subject of future studies.

Comparis on of TLP vs. Actual M ete red DT Loading For instance, the load profiles of commercial and industrial
customers could be stratified into different categories based on
16
14
the type of establishment and load behavior.
Metered
Demand
kW Demand

12
10
Furthermore, introducing load profiles for weekday and
TLP
8 Demand weekend loads and load profiles that differentiate load
6 behavior between wet and dry seasons may make the LP
4 methodology more accurate.
00

00

00

00

00

0
:0

:0

:0

:0

:0

:0

:0
1:

3:

5:

7:

9:

11

13

15

17

19

21

23

However, it should be noted that proper facility-customer


Time
connectivity in the customer load database is a prime
Fig. 9. Comparison of TLP Methodology with Actual Metered Load requirement for either technique to be accurate as well, as
customer kWh consumption, which is one of the requisite data
In comparing the accuracy of the TLP methodology against for both estimation techniques, comes from the load database.
the regression formula, the monthly transformer demand
estimates of both techniques were compared to the actual ACKNOWLEDGMENT
metered demand of four transformers installed with LP- The authors would like to thank the following: Jona D.
capable meters over a period of 6 months. The summary of Villanueva and Mercedo P. Delgado for being the initial
the findings are shown in Table IV. proponents of this project; the Meralco Metering Services
Asset Management group for lending their technical expertise
and use of their facilities; Manuel M. Galvan, Tomas A. [2] K. L. Lo, Z. Zakaria and M. H. Sohod, “Determination of Consumers’
Javellana, Carl G. Aquino and Ireneo Acuña for their Load Profiles Based on Two-stage Fuzzy C-means,” in Proc. 5th
WSEAS International Conference on Power Systems and
endorsement and support for the project; Ricardo V. Electromagnetic Compatibility , 2005, pp.212-217.
Buencamino and the staff of OH-Networks for their unending [3] C. S. Chen, J. C. Hwang, and C. W. Huang, “Application of Load
support; and Ma.Victoria B. Que, for her support and Survey Systems to Proper Tariff Design,” IEEE Trans. Power Systems,
guidance in the completion of this study. vol. 4, pp. 1746-1751, Nov. 1997.
[4] C. S. Chen, J. C. Hwang, Y. M. Tzeng, C. W. Huang, and M. Y. Cho
“Determination of Load Survey System at Taipower,” IEEE Trans.
Power Delivery, vol. 11, pp. 1430-1436, July 1996.
REFERENCES [5] I. H. Yu, J. K. Lee, J. M. Ko and S. I. Kim, “A Method for
[1] D. Gerbec, S. Gasperic and F. Gubina, “Determination and Allocation Classification of Electricity Demands using Load Profile Data,” in
of Typical Load Profiles to the Eligible Consumers”, in Proc. IEEE Proc. Fourth Annual ACIS International Conference on Computer and
Power Tech Conference, 2003, vol.1. Information Science, 2005, pp.164-168.

You might also like