1
Clustering based Short Term Load Forecasting
using Artificial Neural Network
Amit Jain, Member, IEEE, and B. Satish
Abstract-- A novel clustering based Short Term Load
Forecasting (STLF) using Artificial Neural Network (ANN) to
forecast the 48 half hourly loads for next day is presented in this
paper. The proposed architecture uses the historical load and
temperature to forecast the next day load. It is trained using back
propagation algorithm and tested. The daily average load of each
day for all the training patterns and testing patterns is calculated
and the patterns are clustered using a threshold value between
the daily average load of the testing pattern and the daily average
load of the training patterns. The results obtained from neural
network are presented and the results show that the clustering
based approach is more accurate.
Index Terms-- Artificial Neural Network, Back Propagation
Algorithm, Clustering, Short Term Load Forecasting.
I. INTRODUCTION
E LECTRICAL energy is a superior form of energy for all
types of consumer needs. The close tracking of the
system load by the system generation at all times is the basic
requirement in the operation of power systems [1]. There is a
3-7% increase of electric load per year for many years. The
increase of load depends on the population growth, local area
development, industrial expansion etc. The taxonomy of load
forecasting can be considered as Spatial forecasting &
Temporal forecasting. Forecasting future load distribution in a
particular region, such as a county, a state, or the whole
country is called Spatial forecasting. Temporal forecasting is
dealing with forecasting load for a specific supplier or
collection of consumers in future hours, days, months, or even
years. The temporal forecasting can be broadly divided into 4
types – long term, medium term, short term and very short
term. The long term forecast (5 to 20 years) is required as the
building of power plant requires many years. The forecast
ranging from few months to 5 years is termed as medium term
forecasting. Thus long and medium term forecasts help in
determining the capacity of generation, transmission or
distribution system expansions and the type of facilities
required in transmission expansion planning, annual hydro
thermal maintenance scheduling etc. Typically the short term
Amit Jain is with Power Systems Research Center, International Institute of
Information Technology, Hyderabad, Andhra Pradesh, India. (e-mail: amit
@iiit.ac.in).
B. Satish is with Power Systems Research Center, International Institute of
Information Technology, Hyderabad, Andhra Pradesh, India. (e-mail: satish_b
@research.iiit.ac.in).
978-1-4244-3811-2/09/$25.00 ©2009 IEEE
Authorized licensed use limited to: INTERNATIONAL INSTITUTE OF INFORMATION TECHNOLOGY. Downloaded on October 14, 2009 at 08:38 from IEEE Xplore. Restrictions apply.
load forecast covers a period of one week. The forecast
calculates the estimated load for each hour of the day, the
daily peak load and the daily/weekly energy generation. The
forecasted data is used for:
• Unit commitment (Selection of generators in operation,
start up/shut down of generation to minimize operation
cost)
• Hydro scheduling to optimize water release from
reservoirs
• Hydro-Thermal co-ordination to determine the least cost
operation mode (optimum mix)
• Interchange scheduling & energy purchase
• Transmission line loading
• Power system security assessment (load flow & transient
stability studies)
These offline network studies detect conditions under which
the system is vulnerable and warn for corrective actions like
load shedding, power purchase, starting up of peak units,
switching off interconnections and increasing spinning and
stand by reserve. Hence, the day to day operation of the
power system requires accurate short term load forecasting.
Bunn [2] reported that 1% increase in the forecasting error
leads to an increase of £10 million operating cost per year.
The purpose of very short term load forecasting (ranging from
minutes to hours) is for real time control & security
evaluation [3].
The introduction of deregulation in the electricity industry
made short term load forecasting much more important.
Because of its great economic importance and the high
complexity of electric power systems, short term load
forecasting has been subjected to constant improvements in
which numerous techniques have been used [1, 2, 4].
Different techniques for load forecasting: Time series
models (load is modeled as a function of its past observed
values), multiplicative auto-regressive models [5], dynamic
linear [6] or non-linear models [7], threshold auto-regressive
models [8], methods based on Kalman-filtering [9, 10, 11],
Box – Jenkins transfer functions [12, 13], ARMAX models
[14, 15], optimization techniques [16], non-parametric
regression [17], structural models [18] and curve – fitting [19]
procedures. The most popular ones are linear regression ones
[20, 21, 22, 23, 24]. Artificial Intelligence techniques include
Expert Systems [25, 26], Fuzzy inference [27] and Fuzzy –
neural models [7, 28].
In this paper, an attempt is being made to predict the next

day load by using artificial neural network by clustering the
training patterns with respect to the testing pattern. The paper
is organized as follows: Section – II discusses the basics of
ANN and past literature on ANN for STLF. Section – III
explains the proposed architecture and its description. The
solution methodology and the results are presented in Section
– IV and Section – V respectively. Section – VI contain the
conclusions & the future work.
II. BASICS OF ANN AND LITERATURE ON ANN FOR STLF
Among the Artificial Intelligence techniques available,
ANN is widely used for forecasting the electric load. ANN
can be defined as highly connected array of elementary
processors called neurons and is capable to perform non-
linear modeling and adaptation. Neural networks attempt to
learn by themselves the functional relationship between
system inputs and outputs.
In case of load forecasting, it uses previous load patterns as
in the cases of Time series and Regression approaches and
weather information as in the case of Regression approach;
thus ANN has advantages of both of Time series and
Regression methods. The feed forward back propagation
algorithm, which updates the weights in such a way that the
error is minimized, is used to train the neural networks. The
detailed explanation of back propagation algorithm is
available in any standard neural network textbook.
Park et al. [29] presented an ANN approach to electric load
forecasting in which the ANN is trained with the back
propagation algorithm. Peng et al [30] proposed a procedure
for choosing the training cases, which are most similar to the
forecasted inputs. Khotanzad et al [31] presented a load
forecasting system known as ANNSTLF, which predicts the
next 24 hours load. It includes two ANN forecasters. One of
them predicts the base load and the other forecasts the change
in load. The final forecast is computed by an adaptive
combination of these two forecasts. The effect of humidity
and wind speed is considered through a linear transformation
of temperature. Till date, several researchers dealt with the
application of various neural networks to Short Term Load
Forecasting with varying success [32-41]. Although neural
networks are capable of handling nonlinearity between the
electric load and the weather factors that affect the load, they
somehow lack to fully handle unusual changes that occur in
the environment. The topology of a neural network
determines the degrees of freedom available to model the
data. If the neural network is too simple then the network will
not be able to learn the function relating the input to the
output and an over-complex network will learn the noise in
the data and will not be able to generalize.
III. PROPOSED ARCHITECTURE
A model is proposed here to forecast the electric load one
day in advance. The aim is to prepare the model for real-time
forecasts by clustering the available past data.
The system’s electric load, which is the sum of the
Authorized licensed use limited to: INTERNATIONAL INSTITUTE OF INFORMATION TECHNOLOGY. Downloaded on October 14, 2009 at 08:38 from IEEE Xplore. Restrictions apply.
2
individual loads, has two components: Base component and
Random component. These components vary in the load curve
due to the following factors:
• Economical or Environmental
• Time
• Weather and
• Random disturbances
The economical/environmental factors include change in
Service area demographics (rural, residential), Industrial
growth, Emergence of new industry, Change of farming,
Penetration or saturation of appliance usage, Economical
trends (Expansion or Recession), Change of the price of
electricity and Demand side load management. The time
constraints of economical/environmental factors are slow,
measured in years. The time factors affecting the load are
seasonal variation of load (Summer, Winter etc.), start of
school year etc. This also include weekly cyclic variation like
significant reduction in load on weekends like Saturday &
Sunday, slight reduction in load on Monday & Friday, similar
pattern of load on the other days of the week and different
pattern of load on holidays like Christmas, New Year,
Vacation etc. The various weather factors that affect the load
are air temperature, dew temperature, wet bulb temperature,
relative humidity, thunderstorms, wind speed, rain, fog, snow,
cloud cover/sunshine. Not all weather factors are similar in
importance and among them temperature is the most
important as it has direct influence on many kind of electrical
consumption. The random disturbances include start or stop
of large loads (steel mill, factory or furnace), widespread
strikes, sporting events (football games, cricket matches etc.),
popular television shows and shut-down of industrial facility.
The proposed architecture is shown is Fig. 1. The objective
of the proposed architecture is to recognize the above factors
from the training data and predict the load accordingly. Thus
a suitable architecture along with appropriate inputs is
needed. There are no general rules to follow in the selection
of input variables. It depends largely on experience,
professional judgment and preliminary experimentation. The
demand for electricity is known to vary by the time of the day,
week, month, temperature and usage habits of the consumers.
Though usage habit is not directly observable, it may be
implied in the patterns of usage that have occurred in the past.
For solving a STLF problem all of these inputs are not needed
at the same time. Depending on the forecast to be made,
whether daily or hourly; the choice of input variables will
change.
Ld-1
Td-1
Neural Network Ld
Td
DOW
Fig. 1 Proposed architecture for STLF
 3

A. Description of Proposed Architecture

INPUTS: 53
Load (Ld-1): 48 half-an-hour loads
Temperature (Td-1): 1 (average temperature)
Forecasted day’s average Temperatures (Td):1
Day Of the Week (DOW) to be forecasted: 3
(Sunday-001, Monday-010, Tuesday-011, Wednesday100,
Thursday-101, Friday-110, Saturday-111)
OUTPUTS: 48
Forecasted Load (Ld): 48 half-an-hour loads

IV. SOLUTION METHODOLOGY

A. Data Analysis
The data considered for the proposed architecture include
electricity load and temperature. The load data set contains
the load per half hour of each day for two consecutive years Fig. 2 Training patterns for Sunday
while the temperature data set provides the average daily
temperature for the same two consecutive years. The 2 years
data contains 104 daily load curves for each day of a week.
The data is divided into 2 sets with 91 patterns for training
and 13 patterns for testing for each day of the week.
Fig. 2 and Fig. 3 represent the daily load curves of the
training set for Sunday and Wednesday, representing a
weekend day and a weekday, respectively. They clearly show
that the load is changing with season and furthermore the load
pattern of weekdays is different from that of the weekend.

Fig. 3 Training patterns for Wednesday

--- Training
patterns
*-- Testing
pattern

Fig. 4 Clustering of training patterns for different testing patterns for Tuesday

Authorized licensed use limited to: INTERNATIONAL INSTITUTE OF INFORMATION TECHNOLOGY. Downloaded on October 14, 2009 at 08:38 from IEEE Xplore. Restrictions apply.
 4

Fig. 4 shows the clustered training patterns for all the 13

B. Creating the Sample Set
testing patterns by considering a threshold value of 30MW for
The ANN is trained with the historic data before testing Tuesday. TABLE I contains the patterns matched for each
them. The first step for training them is obtaining an accurate Testing pattern for Tuesday corresponding to Fig. 4.
historical data. The data should be chosen that is relevant to
the model. How well the data is chosen is the defining factor V. RESULTS
in how well the networks output will match the event being
modeled. There should be some correlation between the A. Without clustering
training data and the testing data. In the load data, in general The neural network is trained initially with the
all the Sunday’s load data look alike, all the Monday’s data corresponding day patterns without doing clustering. Hence
look alike and this holds good for all the days of the week. for each day, the ANN is trained with 91 patterns.
Hence for testing a day, the training data considered is the
past data same as that of the testing day. B. With clustering
In this step, the training patterns are clustered by using a
C. Data Preparation threshold value of 30MW between the daily average load of
In this stage, the typical (raw) input data has to be arranged the testing pattern and the daily average load of the training
as input and output pattern pairs for training the ANN. The 53 patterns for all the days.
inputs for the ANN to be arranged as one column vector and TABLE II shows the comparison of maximum and average
the 48 outputs are to be arranged as another column vector. percentage errors of ANN (without clustering & with
This is to be done for all the days of the past data. clustering) for 5th Testing pattern for all the days. The results
show that the maximum and average % errors are less for all
D. Normalization the days when the training patterns are clustered.
Normalization is an important stage for training the neural
network. The data is normalized in such a way that the higher TABLE II
values should not suppress the lower values in order to retain COMPARISON OF MAXIMUM & AVERAGE % ERRORS OF ANN
(WITHOUT CLUSTERING & WITH CLUSTERING) FOR 5TH TESTING
the activation function [42]. Both the load and the PATTERN
temperature data should be normalized to the same range of
values. ANN ANN
E. Clustering Day % Error (Without (With
The daily average load is calculated for all the 104 patterns Clustering) Clustering)
for all the days of the week. The patterns are clustered based Max. % Error 16.5899 8.4659
Sunday
on the threshold value of the difference between the daily Avg. % Error 10.6365 3.3788
average load of the training patterns and the daily average Max. % Error 16.0491 11.6849
Monday
load of the testing pattern. Avg. % Error 7.8471 4.3927
Max. % Error 21.0562 14.5678
Tuesday
TABLE I Avg. % Error 12.0532 6.4338
PATTERNS CLUSTER FOR ALL THE 13 TESTING PATTERNS FOR TUESDAY
Max. % Error 21.7866 8.3199
Wednesday
Avg. % Error 13.1411 3.6995
Serial Testing pattern No. of Training patterns
Max.% Error 18.1500 12.9011
No. Number matched Thursday
Avg. % Error 11.6557 3.9104
1 1 7 Max. % Error 20.1283 15.4229
2 2 7 Friday
Avg. % Error 11.4397 3.4454
3 3 11 Max. % Error 10.0836 9.0649
Saturday
4 4 23 Avg. % Error 3.7211 3.3598
5 5 13
TABLE III
6 6 25 PATTERNS CLUSTER FOR 5
TH
TESTING PATTERN FOR DIFFERENT THRESHOLDS
7 7 22 Patterns Matched Patterns Matched
Day
8 8 18 (30 MW Threshold) (80 MW Threshold)
9 9 17 Sunday 16 45
Monday 21 45
10 10 11
Tuesday 13 38
11 11 9 Wednesday 20 41
12 12 18 Thursday 19 40
13 13 22 Friday 20 40
Saturday 18 42

Authorized licensed use limited to: INTERNATIONAL INSTITUTE OF INFORMATION TECHNOLOGY. Downloaded on October 14, 2009 at 08:38 from IEEE Xplore. Restrictions apply.
 5

The patterns are also clustered by considering 80MW
threshold. The no. of patterns matched for each day for the 5th
testing pattern is shown in TABLE III. From this table, we can
observe that the more the threshold, the more the number of
patterns matched.
respectively. The predicted loads obtained by using clustering
approach are more accurate than the loads obtained without
clustering.

TABLE IV
COMPARISON OF MAXIMUM & AVERAGE % ERRORS OF ANN (WITH
CLUSTERING FOR 30MW & 80MW) FOR 5TH TESTING PATTERN

ANN (30MW ANN (80MW

Threshold) Threshold)
Day % Error
(With (With
Clustering) Clustering)
Max. % Error 8.4659 10.3134
Sunday Fig. 6 Comparison of Actual & Predicted loads from ANN for Wednesday for
Avg. % Error 3.3788 3.4542
7th testing pattern
Max. % Error 11.6849 12.9257
Monday
Avg. % Error 4.3927 5.3571
VI. CONCLUSIONS & FUTURE WORK
Max. % Error 14.5678 16.7703
Tuesday A clustering based neural network approach for predicting
Avg. % Error 6.4338 8.6646
Max. % Error 8.3199 9.5505 the next day load is discussed. The training patterns for a
Wednesday particular day are generated by using a threshold between the
Avg. % Error 3.6995 4.5594
daily average loads of the all training patterns and the daily
Max.% Error 12.9011 14.6886
Thursday average load of the testing pattern. The predicted load
Avg. % Error 3.9104 5.5892
obtained by clustering the training patterns is following the
Max. % Error 15.4229 16.8100 actual load closely compared to the predicted load obtained
Friday
Avg. % Error 3.4454 3.6120 without clustering. Results are presented for different
Max. % Error 9.0649 9.1913 threshold values which result in forming different cluster
Saturday
Avg. % Error 3. 3598 3.6568 patterns. The data should be normalized before training the
ANN and the method of normalization also affect the error in
TABLE IV shows the comparison of maximum and forecasting. The proposed method does not require any heavy
average percentage errors of ANN (with clustering for 30MW computational burden and can be easily implemented
& 80MW threshold values) for 5th Testing pattern for all the compared to the conventional approaches. The most
days. The results show that the maximum and average % important conclusion from the present work is to show the
errors are less for all the days for 30MW threshold when applicability of clustering techniques for choosing training
compared to the 80MW threshold. It also shows that patterns for ANN based methods for getting better short term
maximum and average % errors are less for 80MW threshold load forecasting results. By using Pattern Recognition
compared to the errors when patterns are not clustered (refer techniques, the patterns, similar to that of the input of the
TABLE II for errors without clustering). testing pattern, can be chosen (instead of clustering based on
the daily average load corresponding to the day to be
forecasted) from the past data and then trained. This may be
better approach and may still reduce the forecasting error. The
authors are working on this direction and the results for that
study will be presented in a future publication.

VII. REFERENCES
[1] Gross, F. D. Galiana, “Short term load forecasting”, Proc. IEEE, Vol. 75,
No. 12, Dec. 1987, pp. 1558-1573.
[2] Bunn, D. W., “Short Term Forecasting: A review of procedures in the
electricity supply industry”, Journal of the Operational Research Society,
Vol. 33, 1982, pp.533-545.
[3] K. S. Swarup, S. Yamashiro, “Neural network based forecasting of daily
load”, FANATIC - 99, REC-WARANGAL, pp. 1-11.
[4] I. Moghram and S. Rahmq "Analysis and evaluation of five short-term load
Fig. 5 Comparison of Actual & Predicted loads from ANN for Sunday for 7th forecasting techniques", IEEE Transactions on Power Systems, Vol. 4, No.
testing pattern 4, October 1989, pp.1484-1491.
[5] G. A. N. Mbamalu and M. E. El-Hawary, “Load forecasting via
suboptimal seasonal autoregressive models and iteratively reweighted least
Fig. 5 and Fig. 6 shows the comparison of actual and squares estimation”, IEEE Trans. Power Systems, vol.8, no.1, pp.343–348,
predicted loads from ANN (without & with clustering – 1993.
30MW) for Sunday & Wednesday for 7th testing pattern,

Authorized licensed use limited to: INTERNATIONAL INSTITUTE OF INFORMATION TECHNOLOGY. Downloaded on October 14, 2009 at 08:38 from IEEE Xplore. Restrictions apply.

[6] A.P. Douglas, A.M. Breipohl, F.N. Lee, R. Adapa, “The impact of
temperature forecast uncertainty on Bayesian load forecasting”, IEEE
Trans. Power Systems, vol.13, no.4, pp.1507–1513,1998.
[7] R. Sadownik and E. P. Barbosa,“Short-term forecasting of industrial
electricity consumption in Brazil,” J. Forecast, vol.18, pp.215–224, 1999.
[8] S. R. Huang, “Short-term load forecasting using threshold autoregressive
models”, IEE Proc. Gener. Transm. Distrib., vol.144, no.5, pp. 477–
481,1997.
[9] D. G. Infield and D. C. Hill, “Optimal smoothing for trend removal in
short term electricity demand forecasting”, IEEE Trans. Power Systems,
vol.13, no.3, pp.1115–1120,1998.
[10] J. H. Park, Y. M. Park and K. Y. Lee, “Composite modeling for adaptive
short-term load forecasting”, IEEE Trans. Power Systems, vol.6, no.2,
pp.450–457,1991.
[11] S. Sargunaraj, D. P. Sen Gupta and S. Devi, “Short-term load forecasting
for demand side management”, IEE Proc. Gener. Transm. Distrib., vol.
144, no.1, pp.68–74,1997.
[12] M. T. Hagan and S. M. Behr, “The time series approach to short term load
forecasting”, IEEE Trans. Power Systems, vol. PWRS-2, no.3, pp. 785–
791,1987.
[13] G. M. Jenkins, “Practical experiences with modeling and forecast”, Time
Series, 1979.
[14] H. T. Yang and C. M. Huang, “A new short-term load forecasting
approach using self-organizing fuzzy ARMAX models”, IEEE Trans.
Power Systems, vol.13, no.1, pp.217–225,1998.
[15] H. T. Yang, C. M. Huang and C. L. Huang, “Identification of ARMAX
model for short term load forecasting: An evolutionary programming
approach”, IEEE Trans. Power Systems, vol.11, no.1, pp.403–408, 1996.
[16] Z. Yu, “A temperature match based optimization method for daily load
prediction considering DLC effect”, IEEE Trans. Power Systems, vol.11,
no.2, pp.728–733,1996.
[17] W. Charytoniuk, M. S. Chen and P. Van Olinda, “Non parametric
regression based short-term load forecasting”, IEEE Trans. Power
Systems, vol.13, no.3, pp.725–730,1998.
[18] A. Harvey and S. J. Koopman, “Forecasting hourly electricity demand
using time-varying splines”, J. American Stat. Assoc., vol.88, no.424,
pp.1228–1236,1993.
[19] J. W. Taylor and S. Majithia,“Using combined forecast swith changing
weights for electricity demand profiling”, J. Oper. Res. Soc., vol.51, no.1,
pp.72–82, 2000.
[20] R. F. Engle, C. Mustafa and J. Rice, “Modeling peak electricity demand”,
J. Forecast., vol.11, pp.241–251,1992.
[21] T. Haida and S. Muto, “Regression based peak load forecasting using a
transformation technique”, IEEE Trans. Power Systems, vol.9, no.4,
pp.1788–1794,1994.
[22] A. D. Papalexopoulos and T. C. Hesterberg, “A regression based approach
to short-term system load forecasting”, IEEE Trans. Power Systems, vol.5,
no.4, pp.1535–1547,1990.
[23] R. Ramanathan, R. Engle, C. W. J. Granger, F. Vahid-Araghi and C.
Brace, “Short-run forecasts of electricity loads and peaks”, Int. J. Forec.,
vol.13, pp.161–174,1997.
[24] S. A. Soliman, S. Persaud, K. El-Nagar and M. E. El-Hawary,
“Application of least absolute value parameter estimation based on linear
programming to short-term load forecasting”, Elect. Power & Energy
Syst., vol.19, no.3, pp.209–216, 1997.
[25] K. L. Ho, Y. Y. Hsu, C. F. Chen, T. E. Lee, C. C. Liang, T. S. Lai and K.
K. Chen, “Short term load forecasting of Taiwan power system using a
knowledge-based expert system”, IEEE Trans. Power Systems, vol.5, no.4,
pp.1214–1221,1990.
[26] S. Rahman and O. Hazim, “A generalized knowledge-based short-term
load-forecasting technique”, IEEE T. Power Syst, vol.8, no.2, pp.508–
514,1993.
[27] H. Mori and H. Kobayashi, “Optimal fuzzy inference for short-term load
forecasting”, IEEE Trans. Power Systems, vol.11, no.1, pp.390–396,1996.
[28] S. E. Papadakis, J. B. Theocharis, S. J. Kiartzis and A. G. Bakirtzis, “A
novel approach to short-term load forecasting using fuzzy neural
networks”, IEEE Trans. Power Systems, vol.13, no.2, pp.480–492,1998.
[29] Park D.C., El-Sharkawi M.A., Marks II R.J., Atlas L.E. and Damborg
M.J. , “Electric load forecasting using an artificial neural network”, IEEE
Transactions on Power Systems, Vol. 6, No.2, May 1991, pp. 442 – 449.
Authorized licensed use limited to: INTERNATIONAL INSTITUTE OF INFORMATION TECHNOLOGY. Downloaded on October 14, 2009 at 08:38 from IEEE Xplore. Restrictions apply.
6
[30] Peng T.M., Hubele N.F. and Karady G.G., “Advancement in the
application of neural networks for short-term load forecasting”, IEEE
Transactions on Power Systems, Vol.7, No.1, 1992, pp.250 – 257.
[31] A. Khotanzad, R. A. Rohani and D. Maratukulam, ”ANNSTLF-Artificial
neural network short term load forecaster-Generation Three”, IEEE
Trans. PAS, Vol. 13,No. 4, Nov. 1998, pp: 1413-1422.
[32] M. Beccali, M. Cellura, V. Lo Brano, A. Marvuglia, “Forecasting daily
urban electric load profiles using artificial neural Networks”, Elsevier,
Energy Conversion and Management, 45, 2004, pp. 2879 – 2900.
[33] Lee K.Y., Cha Y.T., and Park J.H., “Short-term load forecasting using an
artificial neural network”, IEEE Transactions on Power Systems, Vol.7,
No.1, 1992, pp.124 –132.
[34] Chen S. -T., Yu D.C.and Moghaddamjo A.R., “Weather sensitive short-
term load forecasting using nonfully connected artificial neural network”,
IEEE Transactions on Power Systems, Vol.7, No.3, Aug 1992, pp.1098-
1105.
[35] Lu C. –N., Wu H. -T. and Vemuri S., “Neural Network based short term
load forecasting”, IEEE Transactions on Power Systems, Vol.8, No.1, Feb.
1993, pp. 336-342.
[36] Bakirtzis A.G., Petridis V., Kiartzis S.J., Alexiadis M.C. and Maissis
A.H., “A neural network short-term load forecasting model for the Greek
power system”, IEEE Transactions on Power Systems, Vol.11, No.2, May
1996, pp. 858-863.
[37] Lamedica R., Prudenzi A., Sforna M., Caciotta M.and Cencellli V.O., “A
neural network based technique for short-term forecasting of anomalous
load periods”, IEEE Transactions on Power Systems, Vol. 17, No.3,
1996, pp.1749-1756.
[38] Topalli A.K. and Erkmen. I., “A hybrid learning for neural networks
applied to short term load Forecasting”, Neuro Computing, Vol.51, April
2003, pp. 495-500.
[39] Carpinteiro O.A.S., Reis A.J.R. and da Silva A.P.A., “A hierarchical
neural model in short-term load forecasting”, Applied Soft Computing,
Vol. 4, No.4, 2004, pp.405 – 412.
[40] Satish B., Swarup K.S., Srinivas S. and Hanumantha Rao A., “Effect of
temperature on short term load forecasting using an integrated ANN”,
EPSR Journal, Vol.72, No.1, 2004, pp.95-101.
[41] Topalli A.K., Erkmen I. and Topalli I., “Intelligent short-term load
forecasting in Turkey”, International Journal of Electrical Power and
Energy Systems, 2006, Vol.28, No.7, pp.437-447.
[42] T. Dillon, S. Sestito, ’’Short term load forecasting using neural
networks”,1996.
VIII. BIOGRAPHIES
Amit Jain graduated from KNIT, India in Electrical
Engineering. He completed his masters and Ph.D.
from Indian Institute of Technology, New Delhi,
India.
He was working in Alstom on the power SCADA
systems. He was working in Korea in 2002 as a
Post-doctoral researcher in the Brain Korea 21
project team of Chungbuk National University. He
was Post Doctoral Fellow of the Japan Society for
the Promotion of Science (JSPS) at Tohoku
University, Sendai, Japan. He also worked as a Post Doctoral Research
Associate at Tohoku University, Sendai, Japan. Currently he is an Assistant
Professor in IIIT, Hyderabad, India. His fields of research interest are power
system real time monitoring and control, artificial intelligence applications,
power system economics and electricity markets, renewable energy, reliability
analysis, GIS applications, parallel processing and nanotechnology.
B. Satish is a Ph. D. candidate in Power Systems
Research Center, International Institute of Information
Technology, Hyderabad, India. He received his B. Tech
degree from Koneru Lakshmaiah College of Engineering,
Vijaywada, India and M. Tech degree form IIT Madras,
India. He worked for three and half years as a faculty in
Department of EEE, Vellore Institute of Technology,
Vellore and published 12 papers at International and
national levels including IEEE and ELSEVIER. His areas of interest include
Applications of Artificial Neural Networks and Fuzzy logic to power systems,
 7

SCADA and Condition Monitoring, Diagnosis and Prognostics of Electrical

equipment.

Authorized licensed use limited to: INTERNATIONAL INSTITUTE OF INFORMATION TECHNOLOGY. Downloaded on October 14, 2009 at 08:38 from IEEE Xplore. Restrictions apply.