Page 139 - IJEEE-2022-Vol18-ISSUE-1

Al-Nasiri, Al-bayaty & Al-Hafidh
Fig. 3: Mosul Data
Electricity is consumed in the residential sector, the
government sector, the industrial sector, the agricultural
sector, the commercial sector, and the tourism sector. The
residential sector represents the largest electricity
consumption sector in Iraq. The residential electrical load
consists of many components. They are constantly changing
due to many factors affecting these components.
Temperature is the most important climatic factor affecting
the change of load. The electrical load in the residential
sector can be classified into five main components according minimum temperatures, and day sequence. The output is
to the level of consumption as follows:
• Lighting.
• Domestic.
• Cooling.
• Air Heating.
• Water Heating.
Figure 4 shows the consumption rates of electric load
components in the residential sector for selected months of
summer and winter, August 2010 and January 2011.
100.00%
80.00%
60.00%
40.00%
20.00%
0.00%
Lighting Domestic Cooling Water Air
heater heating
Jan Aug
Fig. 4: Percentages of electrical load components
D. Training and Test ANN
The ANN model must go through a training phase, which
is the second phase before it can be applied in predicting
| 135
electrical loads. The goal of the training process is to adjust
the network weights and biases and reduce the error between
the network output and the desired output [19]. A feed-
forward backpropagation neural network was used and
trained with the Levenberg- Marquardt back propagation
(MLP) algorithm. It adjusts the network's weights and biases
very quickly. Training is an iterative process that continues
until an acceptable level of error is gained.
Input data include daily peak load, daily maximum and
three compounds load predicting as shown in Fig 5. The load
and weather data for the period (April 1, 2010, to March 31,
2011) were used to train and test the network. The data set
was divided into two groups:
• 50% of the data set is used for training.
• The other 50% is used to test the model.
After training the network and obtaining the lowest error
value, the accuracy of the model is tested with a set of data.
After the network is trained and tested, it can be given new
input information to predict the desired output.
%Load percentage
of Five components
Daily Peak load of ANN 5 components
current day and next week of Load
forecast
next week
Tmin ,Tmax today
Tmin , Tmax for next week
Day number in data series
Fig. 5: Data set of ANN model
III. ANN PREDICTION RESULTS
The artificial neural network model was applied to the
weather and load data for the city of Mosul for the period (1
April 2010 - 31 March 2011). The input data includes the
maximum and minimum temperatures, the daily peak load,
load percentage of Five components and the day sequence.
Network training performance was evaluated using mean
square error (MSE).
Figure 6. shows the training performance of the neural
network. Network training continues to reach maximum
convergence of the output values and the target. The training
stops at (1000 iterations). Figure 7. shows the regression of
the neural network in the training and testing, and the value
of the Correlation coefficient (R) is (0.98) for training and
testing respectively, Where the results of the model (Output)
fit with the original (Target) values very much.