Thanks to visit codestin.com
Credit goes to www.scribd.com

0% found this document useful (0 votes)
48 views6 pages

Load Forecasting Using Artificial Neural Network

This paper presents a method for short-term electricity load forecasting using an Artificial Neural Network (ANN), addressing the limitations of traditional econometric approaches that assume linear relationships. The proposed single-layer ANN model incorporates key input variables such as hour, day indicators, and weather-related inputs, demonstrating improved forecasting accuracy. The findings highlight the adaptability of ANNs in optimizing power system operations, including unit commitment and power trading.

Uploaded by

Shivam Koli
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
48 views6 pages

Load Forecasting Using Artificial Neural Network

This paper presents a method for short-term electricity load forecasting using an Artificial Neural Network (ANN), addressing the limitations of traditional econometric approaches that assume linear relationships. The proposed single-layer ANN model incorporates key input variables such as hour, day indicators, and weather-related inputs, demonstrating improved forecasting accuracy. The findings highlight the adaptability of ANNs in optimizing power system operations, including unit commitment and power trading.

Uploaded by

Shivam Koli
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 6

Short term load forecasting using artificial neural network

Shivam A. Koli and Arya V. Joshi


Department of Electrical Engineering, walchand college of engineering, Vishrambag sangli, Maharashtra
416415, India

Abstract.
Short-term electricity demand forecasting is crucial for the effective control and
scheduling of power systems. Traditional econometric approaches, which often assume
linear relationships between weather variables and load demand, may not provide
accurate results due to the nonlinear and non-stationary nature of these relationships. This
paper proposes an adaptable technique using an Artificial Neural Network (ANN) for load
forecasting. ANNs, utilize supervised learning to map historical data and various inputs to
forecast future loads. The proposed model employs a single-layer neural network with
one input layer, one hidden layer, and one output layer. Key input variables include hour
and day indicators and weather-related inputs. The ANN’s ability to adapt and learn from
data makes it a valuable tool for optimizing unit commitment, startup and shutdown of
power plants, control of spinning reserves, and power trading in interconnected systems.

Keywords: load forecasting; artificial neural network

1.Introduction- assuming a linear relationship. However,


as Park et al. (1991) indicate, the
Short-term forecasts of electricity
econometric approach may not give
demand are needed for control and
sufficiently accurate results because of
scheduling of power systems, and hence
nonlinear and non-stationary relationships
they are very important for dispatchers.
between the load data and weather
Short-term electric load demand is often
variables. Therefore, an adaptable
considered as a function of weather
technique is needed.
variables and human social activities.
An artificial neural network (ANN)
Traditional econometric approaches
approach is such an adaptable technique.
establish functional relationships between
The Artificial neural network uses a
weather variables and current load demand
supervised learning technique for training
for the forecasting function, often
neural networks. The goal of this type of
network is to create a model that correctly
maps the input to the output using
Ô
historical data and various inputs so that
the model can then be used to produce the
forecast. These load forecasts play a vital
role in optimum unit commitment, start up
and shut down of thermal plants, control of
spinning reserve and buying and selling of
power in interconnected systems.an Figure 1:neural network
structure.
advantage of the Artificial neural network
The output is a sum of the weighted
is that it does not require any assumption
nonlinear terms given by
between load and weather variables in
advance. Moreover, the literature suggests where
that the accuracy of forecasts is better than
traditional econometric forecasting.
In this literature we propose a
method of load forecasting based on a Wᵢⱼ (i = l,..., n ; j = l ,..., nᵢ) is the weight
single-layer neural network and input factor between the ith input node and the
variables to be used for such neural ith hidden node; v ᵢ is the threshold value
network. at the ith hidden node defined by the
2.Artificial neural network algorithm; and xⱼ is the jth input element.

Artificial neural network is an The learning of a network is a


information paradigm that is inspired by recursive process which adapts the
biological nervous systems. The system connection strengths, i.e. the weights
structure is composed of a large number of among the nodes and the threshold values
highly interconnected processing elements at each iteration. The error function E is to
(neurons) working together to solve update the weights at each recursive
specific problems. Each unit operates only minimum:
on its local data and on the inputs, they
receive via the connections. The proposed
neural network consists of one input layer,
one hidden layer and one output layer. The
where K is the number of cases used to
neural network maps the input vector X to
train the network O(m) is the actual output
the output vector Ô (Fig. 1).
of the training case m, and Ô(m) is the
forecasted output of the network.
3.selection of input variables

The most important work in 3.2. Day indicator D(n)


building an Artificial neural network load
Load also changes throughout the
forecasting model is the selection of input
week it is clear from the fig.3 that power
variables. Selection of inputs is based on
consumption in the weekend (day 6,7)
judgement of experience of the engineer
drops compared to power consumption in
and it is carried out by trial and error since
the weekdays. To take this into account
there is no rule of thumb that can be
while calculating load forecasting, variable
followed to determine the which inputs to
associated with day of week D(n) where n
use. However, some statistical analysis can
changes from 1 to 7 representing days of
be done to determine the variables with
week
most influence on power consumption.
These inputs are (a) hour and day, (b)
weather associated inputs and (c) historical
loads

3.1. Hour indicator H(i)

Load changes throughout the day from one


Figure 3: load changes during
hour to another. On a typical day in India week
peak load occurs generally between
06:00PM to 08:00PM. And minimum load
3.3. Weather indicating variables
occurs at 03:00AM(Fig.2). We can
conclude that the magnitude differs
between various hours by a large margin
there for an hour indicator H(i) (where i
changes between 1-24) is very helpful
while calculating short term load

Figure 2: daily load curve in


India
Weather variables primarily amplitude and offset are often referred to
include temperature, wind velocity cloud as normalization coefficients. The

normalization coefficients are computed


based on the minimum and maximum
values found across all of the data sets.
The formulae used to compute these
coefficients are

Figure 4: monthly power consumption and


temperature graph
where Max(i) and Min(i) are the maximum
coverage, rain etc. out of all the above
and minimum values The input
variables temperature seems to have
components normalize the data using the
strongest correlation to power
following formula:
consumption as displayed by Fig.4.
Although wind velocity and cloud 4.Training of ANN

coverage have their own effects on load Training is a process to find the
due to its effect being very small no input–output map by repeatedly analyzing
significant relation is seen. therefore, only the training set as the network learns. The
temperature as a weather variable Is parameters of the network obtain optimal
considered while calculating load values during this process and the
forecasting performance checked with the test set. The

3.5. Normalized data ingredients for supervised learning are the


input, the desired response, the definition
Normalization or scaling is not
of error and a learning law.
really a functional requirement for the
neural networks to learn, but Training can be implemented in

it significantly helps as it transposes the two ways. Either we present a pattern and

input variables into the data range that the adapt the weights (on-line training), or we

sigmoid activation functions lie in (i.e. for present all the patterns in the input file (an

logistic [0, 1] and tanh [-1, 1]. And it epoch), accumulate the weight updates,

makes it so that all inputs are in then update the weights with the average

comparable range weight update, called batch learning.in


proposed ANN backpropagation can be
Each input data is multiplied by an
implemented to effectively train the
amplitude and shifted by an offset. The
artificial neural network. Backpropagation
is an iterative algorithm, that helps to as well. These trends make it difficult to
minimize the cost function by determining correctly model load forecasting data.
which weights and biases should be
Artificial neural networks can
adjusted. During every epoch, the model
become too complex because of their
learns by adapting the weights and biases
architecture and the huge datasets used to
to minimize the loss by moving down
train them. They can also memorize the
toward the gradient of the error. Thus, it
training data. These can lead to poor
involves the two most popular
generalization of new data.
optimization algorithms, such as gradient
descent or stochastic gradient descent. They can also be vulnerable to
adversarial examples or minor changes in
the input data. These changes can cause
5. Performance measures of the network the particular artificial neural network to
make incorrect decisions and irrelevant
In order to evaluate the
outcomes.
performance of the network model, the
following common performance measures 7. Conclusion
are used automatically by the software:
The application of Artificial Neural
mean square error (MSE), normalized
Networks (ANNs) for short-term
mean square error (NMSE), correlation
electricity load forecasting presents a
coefficient, % error, Akaike’s information
significant advancement over traditional
criterion (AIC) and the minimum
econometric methods. By leveraging the
description length (MDL). The most
adaptability and learning capabilities of
common measure is to minimize MSE in
ANNs, this approach addresses the
the forecasting but this measure has some
nonlinear and non-stationary relationships
drawbacks. Alternative measures such as
between load demand and influencing
percentage error, mean percentage error
variables such as weather and human
(MPE) and mean absolute percentage error
activities. The proposed single-layer neural
(MAPE) are also presented in this study
network model, which incorporates key
6.limitations of ANN input variables like hour and day
indicators and weather-related inputs
A major problem in modelling load
demonstrates improved forecasting
forecasting is that there are usually many
accuracy.
trends in the data. For example, there may
be a seasonal trend reflecting the seasons This enhanced accuracy is critical
of the year. As well as there may be for optimizing power system operations,
monthly and/or weekly trends in the data including unit commitment, thermal plant
management, spinning reserve control, and
power trading. The findings underscore the
potential of artificial neural networks to
enhance load forecasting, offering a
efficient and flexible solution that can
adapt to the dynamic nature of electricity
demand.

References
(4) N. Kandil et al. / Electrical Power and
(1) Ongsakul, Weerakorn_ Vo, Dieu Ngoc - Energy Systems 28 (2006) 525–530
Artificial intelligence in power system
optimization (2013, CRC Press, Taylor & (5) S. C. Bhattacharya and L. T. Tha Short -
Francis) Term ELECTRIC LOAD forecasting

(2) P.K. Dash et al./Electric Power Systems (6) S.M. Islam et al./Electric Power Systems
Research 32 (1995) 19-23 Research 34 (1995) 1-9

(3) O.A.S. Carpinteiro et al. / Applied Soft (7) M. Ghiassi et al. / Electric Power Systems
Computing 4 (2004) 405–412 Research 76 (2006) 302–316

You might also like