Wind Power Forecasting using Artificial Neural Network

MOHAMMAD A. OBEIDAT

Department of Electrical Engineering, College of Engineering, Al-Ahliyya Amman University,

JORDAN

Department of Electrical Power and Mechatronics Engineering, College of Engineering,

Tafila Technical University, Tafila 66110, JORDAN

BAKER N AL AMERYEEN

Kepco Plant Service & Engineering, Amman, JORDAN

AYMAN M MANSOUR

Faculty of Computer Studies (FCS), Arab Open University (AOU),

Amman – Tareq, JORDAN

Department of Computer and Communications Engineering, College of Engineering,

Tafila Technical University, Tafila 6611, JORDAN

HESHAM AL SALEM

Department of Mechanical Engineering, College of Engineering, Tafila Technical University,

Tafila, 66110, JORDAN

ABDULLAH M. EIAL AWWAD

Department of Electrical Power and Mechatronics Engineering, College of Engineering,

Tafila Technical University, Tafila 66110, JORDAN

Abstract: - The electric energy generated from wind resources is now one of the most important sources in the

electrical power system. Predicting wind speed is difficult because wind characteristics are unpredictable,

highly variable, and dependent on many factors. This paper presents the design of an artificial neural network

used in wind energy forecasting that has been trained using weather data that influences wind energy

generation. Artificial Neural Network (ANN) has gained popularity in recent years due to its superior

performance. The main objective of the developed model is to improve the forecasting of energy generated

from wind farms. The developed system allows the power system operator to determine the best time to rely on

the wind farm to produce power for the electrical system without affecting the stability of the system and

reducing the cost of electricity generation due to the traditional method. The analysis is performed by

investigating wind potential and collecting data from a highly recommended source. The heatmap, covariance

and correlation methods are used to analyze the data, and then the data is used to build an Artificial Neural

Network (ANN) in MATLAB 2020. The results show very high accuracy 99.9%.

Key-Words: - Wind power, Artificial Neural Network ANN, Wind Turbine, Feed forward ANN

Received: June 18, 2021. Revised: June 16, 2022. Accepted: July 26, 2022. Published: September 23, 2022.

1 Introduction

Renewable Energy sources (RES) are very important

in the last few years, It occupies a large field of

study and development with increasing the fire about

global warming, increasing fossil fuels cost, air

pollution and worldwide reducing of CO2 are

forcing to switch from traditional energy source to

Renewable energy source (RES) , such as: wind ,

solar , hydro ,bio ,and nuclear energy source[1].

Using artificial intelligence is used in many

applications recently. It has been used in recent

research in electric vehicle charging loads on

WSEAS TRANSACTIONS on POWER SYSTEMS

DOI: 10.37394/232016.2022.17.28

Mohammad A. Obeidat, Baker N Al Ameryeen,

Ayman M Mansour, Hesham Al Salem,

Abdullah M. Eial Awwad

E-ISSN: 2224-350X

269

Volume 17, 2022

realistic residential distribution system [2]. Beside

this, it has used to make decision regarding selection

of wind farms [3]. One of the used techniques is

neural networks which has been used in power

applications [4] and prediction [5]. Other artificial

intelligence techniques such as decision tree has

been used in power applications [6]

Increase diversity in energy sources it will be

significantly reflected on developing countries,

where RES will be the motivation of enhancing

business, green economy and creating new qualified

jobs [7].

The most type of RES use in Jordan are wind, solar

,hydro and biogas divided in transmission and

distribution network in different size from few KW

to a few MW, the Jordan government Planning to

increase accreditation on RES in the future energy

consumption plan ,Jordan’s electrical load of about

3500 MW which may reach more than 6200 MW by

2025.Wind energy is one of the strongest

possibilities to cover a portion of Jordan's energy

demands, as wind availability rates at speeds suited

for wind energy applications are deemed adequate

and perfect for energy security applications. The

wind speed reaching 7-11 m/s which is very

applicable for wind energy generation [8-9].

2 Problem Formulation

The generation power from the wind turbine

depends on weather conditions included (wind

speed, temperature, humidity, pressure and either)

this factor will effect of the generation planning.

Collecting all type of data that effect on wind power

generation and use it in Models that help predict

energy output through data, this process will help the

operator to know how much energy which will got

from the wind farm [10-11].

2.1 Wind Power Equation

The power available from the wind is given by [9]:

Pa = (1/2) ρ A V^3 (2.1)

whereas:

Pa is the amount of wind energy available in [W],

ρ is the air density in [Kg/m3],

A is the cross-sectional area in [m],

V is the wind velocity in [m/s]

The air density will change with change the weather

condition we can found it by equation (2.2):

ρ=P/(R T) (2.2)

where: P air pressure in (pa)

R specific gas constant for air in (J/(Kg.C)

T temperate in (C)

The rotor cannot extract all of the power available

from the wind stream, and the value of the extracted

power is determined by the total efficiency ŋt

equation (2.3) shown the final form of wind power

equation [9,10]:

Pa= (1/2) ρ A V3 ŋt (2.3)

From the equation (2.3) showed the variables

effecting the wind power which are Pressure,

temperate, wind velocity.

2.2 Wind Power Forecasting

The growing importance of wind power raises the

issue of understanding its behavior and its impact in

electrical sector [11]. Wind power production, being

subject to the available wind, can only controlled

within the margin of the possible production

correspondent to the available wind, thus it has

reduced control capacities [12,13]. The extraction of

energy from the wind should thus be maximized for

economic and environmental reasons. This paper

will demonstration that illustrates how to use the

proposed Artificial Neural Network (ANN) to

estimate wind power based on the input parameters

[14-15].

 Neural Network Design and Methodology

A neural network is a machine learning algorithm

that mimics the human brain and is inspired by

biology. The idea of machine learning refers to a

computer's ability to learn from raw data by

automatically finding a meaningful pattern [16].

Both supervised and unsupervised learning problems

can be addressed by the NN. Supervised learning is

used in fitting situations when the input and desired

output are known . The NN works with numeric data

in these applications, and its goal is to approximate

the relationship between the input and output data.

ANNs consist of simple processing units - neurons -

subdivided into layers - an input, an output, plus one

or more hidden layers - with each connection having

a weight factor that varies according to the input and

the output [17-19].

There are two types of networks: static and

dynamic. Static networks depend only on the current

value of inputs, while dynamic networks are

governed by differential equations, which show

memory [20-23].

During supervised or unsupervised learning, the

weights associated with the connections are adjusted

based on inputs and outputs that have been presented

WSEAS TRANSACTIONS on POWER SYSTEMS

DOI: 10.37394/232016.2022.17.28

Mohammad A. Obeidat, Baker N Al Ameryeen,

Ayman M Mansour, Hesham Al Salem,

Abdullah M. Eial Awwad

E-ISSN: 2224-350X

270

Volume 17, 2022

to the network[24]. The network is given only the

inputs in an unsupervised learning process, and it is

trained to distinguish different classes of data.

ANNs may also be classified as feed forward and

recurrent [25]. Feed forward networks, which are

without directed cycles in their topology graph, are

typically used to predict wind speed and wind

power. Figure 1 shows the typical structure of a

multilayer perceptron, which is typically trained by

supervised learning and is often used for wind

forecasting [26].

Fig. 1: Structure of a multilayer perceptron network.

 Feed forward ANN

The generic feed forward ANN model consists of

numerous layers of neurons, each of which is

referred to as a perceptron. It's worth noting that the

input layer isn't named a perceptron because it

doesn't do any calculations [27-29]. Furthermore,

each neuron gets a large number of input signals

while only producing a single output signal [30-34].

The weighted sum of the neuron's inputs is

computed, and the output is then determined using

one of the typically used activation functions, such

as sign, step, sigmoid, or linear feature. The sign and

step measures are known as hard boundary functions

because their values abruptly shift at a specific point,

making them ideal for decision-making applications

like categorization and pattern recognition [35-38].

• Recurrent ANN

Feed forward ANN does not entirely replicate the

human brain. As a result, literature has proposed a

recurrent ANN with feedback loops from its outputs

to the input. The insertion of such loops improved

the ANN's learning capabilities. The Hopfield

network is a sort of recurrent ANN that works by

feeding each network outcome back to all inputs. In

order to recover a huge number of essential memory

storage, this sort of ANN demands larger storage

capacity [39-42].

2.3 Data Processing

The data for this thesis was obtained for the south of

Jordan and included a variety of weather data it was

4735 column and 41 row it was need analyse and

filtering. The information was gathered from an

internet source (World Weather Online), which

provides a wide range of information. The data was

collected between July 1, 2008, and June 16, 2021.

The data Contains many factors that have no effect

on wind power such as weather Disc, moon phase,

moonset, UV index …etc. we began by analysing all

of the variables that affect wind energy output,

including wind speed, average temperature,

humidity, visibility, pressure, dew point, wind gusts,

and total precipitation.

Fig. 2: The analyzing Weather data With Power

The data are now ready to be analyzed using two

techniques: heat map, correlation matrix, and

covariance matrix. A heat map is a data visualization

approach that displays the magnitude of a

phenomenon as color in two dimensions; through it,

we can observe the impact of each element on wind

power.

2.4 MATLAB Simulation

In this section, the steps of implementing the ANN

for wind power forecasting was illustrated in details:

 Step 1: read the excel sheet with the following

parameters: temperature, pressure, Dew point,

wind speed, gas constant, power coefficient, and

turbine blade length.

 Step 2: Using a Matlab script, code the wind

power equation and compute it for all data (from

year 2008 - 2021).

 Step 3: Using a Matlab script, code the heat map

and compute it for all parameters and observe

the relation between them.

WSEAS TRANSACTIONS on POWER SYSTEMS

DOI: 10.37394/232016.2022.17.28

Mohammad A. Obeidat, Baker N Al Ameryeen,

Ayman M Mansour, Hesham Al Salem,

Abdullah M. Eial Awwad

E-ISSN: 2224-350X

271

Volume 17, 2022

 Step 4: Code the normalization method for

inputs (recorded parameters) and output (wind

power) in a Matlab script and compute it for all

data (from year 2008 - 2021)

 Step 5: To prepare the data for training, utilize

the nntool box in MATLAB program.

Furthermore, choose between two training

methods: feed forward back propagation and

layer recurrent ANN.

Fig. 3: nntool window of Matlab software

 Step 6: After establishing the layers and

neurons counts, train the selected ANN with

the needed parameters.

Fig. 4: ANN structure after design process.

Fig. 5: ANN design process

 Step 7: repeat step 6 for parameters of

(temperature, pressure, wind speed, wind gust,

humidity and visibility) and neurons for layer

recurrent ANN and compare the results with the

feed forward method.

 Step 8: repeat step 6 for different number of

parameters and neurons for feed forward

method. (28 training)

 Step 9: Compare the predicted results with the

recorded data for each train to see the training

attempts.

 Step 10: In terms of regression and MSE,

distinguish between the training trials.

The following is the results of feed forward ANN

training for 6 parameters mentioned in step 8:

Fig. 6: Regression results of ANN training.

WSEAS TRANSACTIONS on POWER SYSTEMS

DOI: 10.37394/232016.2022.17.28

Mohammad A. Obeidat, Baker N Al Ameryeen,

Ayman M Mansour, Hesham Al Salem,

Abdullah M. Eial Awwad

E-ISSN: 2224-350X

272

Volume 17, 2022

Fig. 7: ANN training performance

Figures 6 show that the tool box had four regression

plots, the first of which used 70% of the data and

indicated that the training procedure was excellent.

The second figure shows how to validate the

Forecasting findings, and the third plot shows how

to test the regression with new data. The last graph

depicts the overall regression.

Figures 7 shows that the validation and testing

processes have a low mean squared error,

demonstrating the training's effectiveness.

3 Results and Discussions

In this section, the obtained results from the

covariance and correlation matrices, as well as the

ANN training of feed forward and layer Recurrent

methods that estimate the produced power from six

parameters, were presented and analyzed.

Furthermore, we discussed the impact of ANN

training on a number of parameters in order to

determine which was the most beneficial.

Table 1. Covariance matrix of the estimated parameters

Avg.

Temp.

Wind

speed

Visibility

Pressure

Dew

Point

Wind Gust

Total precip

Wind power

Avg.

Temp.

0.071710

0.001545

0.54898

0.007671

0.0408226

0.018231

0.000697

0.003427

Wind

speed

0.001545

0.022253

0.003967

-0.00546

-0.006607

0.002960

0.023844

0.004637

Visibility

0.054898

0.003967

0.059135

-0.0078

0.028523

0.005316

0.002635

0.004299

Pressure

0.007671

-0.00596

-0.00978

0.013924

-0.00241

-0.00053

-0.05504

-0.0049

Dew Point

0.040826

-0.00607

0.028523

-0.00246

0.034730

-0.1494

-0.00969

-0.00734

Wind

Gust

0.018231

0.002960

0.005316

0.000583

-0.014940

0.030538

0.003708

0.000735

Total

precip

0.000697

0.023844

0.002635

-0.00554

-0.007969

0.003708

0.026731

0.004414

Wind

power

0.003427

0.004637

0.004299

-0.00490

-0.000734

0.000735

0.004414

0.005278

3.1 Covariance and Correlation Matrices

Covariance is a measure of the relationship between

two random variables, and statistics principles can

help investors monitor it. The statistic measures how

much – and how far – the variables change in

tandem. To put it another way, it's a measure of the

variance between two variables.

matrix of covariance Determines the nature of

relationship between the variables, with a negative

value indicating an inverse relation and a positive

value indicating a direct relationship between the

two variables being investigated. Except for pressure

(-0.004900) and dew point(-0.000734), all

parameters in Table 1 exhibit a positive relationship

with wind power. Also the Pressure have negative

relationship with wind speed (-0.005496), Visibility

(-0.009578) , Dew point (-0.002461), wind gust (-

0.000583) and total Precip (-0.005504). Another

WSEAS TRANSACTIONS on POWER SYSTEMS

DOI: 10.37394/232016.2022.17.28

Mohammad A. Obeidat, Baker N Al Ameryeen,

Ayman M Mansour, Hesham Al Salem,

Abdullah M. Eial Awwad

E-ISSN: 2224-350X

273

Volume 17, 2022

factor have negative relationship with the other

factor which is the Dew point like (-0.014940) with

wind gust and (-0.007969) with total Precip.

Table 2. Correlation matrix of the estimated parameters

Avg.

Temp.

Wind

speed

Visibility

Pressure

Dew

Point

Wind

Gust

Total

precip

Wind

power

Avg.

Temp.

-0.03867

-0.84303

0.242753

-0.81807

0.389576

0.015919

-0.17614

Wind

speed

-0.03867

0.109352

-0.31224

-0.23765

0.113546

0.977632

0.427899

Visibility

-0.84303

0.109352

-0.33379

0.629387

0.125094

0.066278

0.243315

Pressure

0.242753

-0.31224

-0.33379

-0.1119

-0.02829

-0.28528

-0.57161

Dew Point

-0.81807

-0.23765

0.629387

-0.1119

-0.45876

-0.26154

-0.0542

Wind

Gust

0.389576

0.113546

0.125094

-0.02829

-0.45876

0.129784

0.057869

Total

precip

0.015919

0.977632

0.066278

-0.28528

-0.26154

0.129784

0.371586

wind

power

-0.17614

0.427899

0.243315

-0.57161

-0.0542

0.057869

0.371586

A totally negative linear correlation between two

variables is shown by a -1 value in the table 3. A

score of 0 shows that there is no linear association

between two variables. Moreover, a value of 1

denotes a perfect positive linear correlation between

two variables.

3.2 ANN Training with Different Methods

Table 3 ANN training methods

Number of inputs = 6

ANN Method

Feed forward back-prop

Layer Recurrent

N in layer

(1)

N in

layer (2)

SSE

MSE

SSE

MSE

5.101E

-03

1.080E-06

2.365E-

5.008

E-05

4.867E

-02

1.031E-05

2.831E-

5.995

E-05

1.573E

-03

3.332E-07

3.316E-

7.022

E-05

1.441E

-03

3.051E-07

2.664E-

5.641

E-06

3.838E

-03

8.128E-07

2.224E-

4.710

E-05

1.421E

-03

3.010E-07

2.791E-

5.911

E-06

Table 3 presented two approaches for training neural

networks, with the feed forward method

outperforming the second strategy in terms of SSE

and MSE. To get the lowest possible error in

training, the ANN should include two layers, each

with 50 and 10 neurons.

Table 2 showed that the high degree of correlation is

between wind speed (-0.427899) and wind power

(strong proportional relation), whereas the poorest

relationship with power is Dew point (-0.0542)

parameter (weak inverse relation).

WSEAS TRANSACTIONS on POWER SYSTEMS

DOI: 10.37394/232016.2022.17.28

Mohammad A. Obeidat, Baker N Al Ameryeen,

Ayman M Mansour, Hesham Al Salem,

Abdullah M. Eial Awwad

E-ISSN: 2224-350X

274

Volume 17, 2022

Fig. 8: Comparison between the two ANN

approaches in terms of error.

3.3 Feed Forward Back-Propagation

Training for 6 Inputs with Different Number

of Layers and Neurons

Table 4 shows the Feed forward back-prop

technique training results for six parameters Avg.

Temp, Wind speed, Visibility, Pressure , Dew Point

and Total precip, (which gave better outcomes in the

comparison). The number of neurons in each layer

was changed started with 10 neurons in layer1 and 1

neurons in layer2. First we started changed the

number of neurons in layer1 and keep the neurons

constant in layer2 , after that we keep neurons in

layer1 and changed neurons in layer2 in three step 5,

10 and 15, and each training's regression results

were recorded. The best regression results (0.9995)

were obtained with 50 neurons in the first layer and

10 neurons in the second.

Table 4. Feed forward back-propagation training

for 6 inputs

N in layer (1)

N in layer (2)

Regression

0.9983

0.9254

0.9927

0.9767

0.9992

0.9679

0.9995

0.9987

0.9995

0.9427

100

0.9430

100

0.9721

100

0.9288

200

0.5952

200

0.9669

200

0.8797

3.4 ANN Training with Different Parameters

and Neurons

Many training sessions employing the Feed forward

back-prop method were held in this part. Each

training contains a different amount of variables as

well as neuron count. These completed trainings

assist us in illustrating the optimal variable to use in

the wind power generation forecast procedure. The

best training was trial 27, which included

Temperature, Humidity, Visibility, and Pressure as

training variables with 40 neurons, as shown in

Table 5.

Table 5. ANN training results for different

variables and different neurons

Trial

Training

inputs

Neurons

number

Regression

MSE

Temperature

0.48827

0.0022

Temperature

0.48356

0.0018

WSEAS TRANSACTIONS on POWER SYSTEMS

DOI: 10.37394/232016.2022.17.28

Mohammad A. Obeidat, Baker N Al Ameryeen,

Ayman M Mansour, Hesham Al Salem,

Abdullah M. Eial Awwad

E-ISSN: 2224-350X

275

Volume 17, 2022

Temperature

0.48030

0.0022

Humidity

0.28159

0.0031

Humidity

0.31025

0.0030

Humidity

0.30274

0.0026

Visibility

0.30888

0.0029

Visibility

0.31531

0.0031

Visibility

0.30758

0.0028

Pressure

0.31442

0.0028

Pressure

0.28790

0.0025

Pressure

0.27470

0.0030

Temperature

& Pressure

0.64869

0.0019

Temperature

& Pressure

0.63456

0.0020

Temperature

& Pressure

0.67036

0.0013

Temperature

& Visibility

0.51451

0.0021

Temperature

& Visibility

0.55484

0.0020

Temperature

& Visibility

0.51416

0.0022

Temperature

& Humidity

& Visibility

0.55793

0.0020

Temperature

& Humidity

& Visibility

0.57050

0.0019

Temperature

& Humidity

& Visibility

0.57870

0.0016

Temperature

& Humidity

& Pressure

0.66780

0.0016

Temperature

& Humidity

& Pressure

0.65606

0.0014

Temperature

& Humidity

& Pressure

0.67130

0.0016

Temperature

& Humidity

& Visibility

& Pressure

0.66030

0.0014

Temperature

& Humidity

& Visibility

& Pressure

0.68860

0.0014

Temperature

& Humidity

& Visibility

& Pressure

0.62560

0.0017.

• Training using one variable

Predicting quality of wind power using one variables

have the best regression and MSE when training

using the temperature with 10 neurons the regression

was (0.48827) and the MSE was (0.00226) as

showed in the figure 9 , All the other variables have

low regression and MSE.

Fig. 9: Power Forecasting using temperature with

10 neurons

• Training using two variables

As shown in Table (5) the training using two

variables has the best regression and MSE when

used the Temperature & Pressure with 50 neurons

with (0.67036) regression and (0.00132) MSE,

showed in figure10.

WSEAS TRANSACTIONS on POWER SYSTEMS

DOI: 10.37394/232016.2022.17.28

Mohammad A. Obeidat, Baker N Al Ameryeen,

Ayman M Mansour, Hesham Al Salem,

Abdullah M. Eial Awwad

E-ISSN: 2224-350X

276

Volume 17, 2022

Fig. 10: Power Forecasting using temperature and

pressure with 50 neurons.

 Training using three variables

As shown in Table (5) it can be seen that the best

three variables can use for training are Temperature

& Humidity & Pressure with 50 neurons the

regression was (0.67130) and the MSE was

(0.00161), the result showed in figure.11.

Fig. 11: Power Forecastingusing temperature,

humidity and pressure with 50 neurons.

 Training using Four variables

The best Forecasting of all training is the forecast of

wind power using the combination of four specified

parameters (temperature, humidity, visibility, and

pressure), and as shown in figure 12, increasing the

number of neurons up to 40 had a favorable

influence on the quality of the forecasting. However,

in terms of amplitudes and frequency of calculated

wind speeds, the combined parameters

Forecastingsurpasses the other techniques in terms

of regression and MSE.

Fig. 12: Power Forecasting using temperature,

humidity, visibility and Pressure with 40 neurons.

4 Conclusions

In this paper, a model of ANN with Weather

conduction was built via MATLAB-SIMULINK as

a real case study. A covariance and correlation

matrices are proposed to study the covariance

between each pair of components and the correlation

between variables of the data used. All parameters,

temperature, humidity, visibility, and pressure

exhibit a positive relationship with wind power.

Feed forward back-prop technique and Layer

Recurrent ANN are used. The performance of the

proposed approach was assessed and validated by

the Regression, SSE and MSE. The best regression

results were obtained with 50 neurons in the first

layer and 10 neurons in the second using feed

forward technique. For Different variables it can be

seen that, the best training was the trial 27, which

included temperature, humidity, visibility, and

pressure as training variables with 40 neurons. The

accuracy of the proposed system reaches 99.9%.

References:

[1] Panwar, N. L., S. C. Kaushik, and Surendra

Kothari. "Role of renewable energy sources in

environmental protection: A review."

Renewable and sustainable energy reviews

15.3 (2011): 1513-1524.

[2] Obeidat, Mohammad A., Abdulaziz Almutairi,

Saeed Alyami, Ruia Dahoud, Ayman M.

Mansour, Al-Motasem Aldaoudeyeh, and

Eyad S. Hrayshat 2021. "Effect of Electric

Vehicles Charging Loads on Realistic

WSEAS TRANSACTIONS on POWER SYSTEMS

DOI: 10.37394/232016.2022.17.28

Mohammad A. Obeidat, Baker N Al Ameryeen,

Ayman M Mansour, Hesham Al Salem,

Abdullah M. Eial Awwad

E-ISSN: 2224-350X

277

Volume 17, 2022

Residential Distribution System in Aqaba-

Jordan" World Electric Vehicle Journal 12, no.

4: 218. https://doi.org/10.3390/wevj12040218

[3] Mansour, Ayman M., Abdulaziz Almutairi,

Saeed Alyami, Mohammad A. Obeidat,

Dhafer Almkahles, and Jagabar Sathik. 2021.

"A Unique Unified Wind Speed Approach to

Decision-Making for Dispersed Locations"

Sustainability 13, no. 16: 9340.

https://doi.org/10.3390/su13169340.

[4] M. A. Obeidat, A. M. Mansour, B. Al

Omaireen, J. Abdallah, F. Khazalah and M.

Alaqtash, "A Deep Review and Analysis of

Artificial Neural Network Use in Power

Application with Further Recommendation

and Future Direction," 2021 12th International

Renewable Engineering IEEE Conference

(IREC), 2021, pp. 1-5, doi:

10.1109/IREC51415.2021.9427846.

[5] Suhail Sharadqah, Ayman M Mansour,

Mohammad A Obeidat, Ramiro Marbello and

Soraya Mercedes Perez, “Nonlinear Rainfall

Yearly Forecastingbased on Autoregressive

Artificial Neural Networks Model in Central

Jordan using Data Records: 1938-2018”,

International Journal of Advanced Computer

Science and Applications (IJACSA), 12(2),

2021.

[6] Mansour, A.M., Abdallah, J., Obeidat, M.A.,”

An efficient intelligent power detection

method for photovoltaic system,” International

Journal of Circuits, Systems and Signal

Processing, vol. 14, pp. 686–699, 2020.

[7] https://ourworldindata.org/energy-mix.

[8] https://www.pub.iaea.org/mtcd/publications/p

df/cnpp2013_cd/countryprofiles/Jordan/

Jordan.htm

[9] Mathew, S. (2006). Wind energy:

fundamentals, resource analysis and

economics. Springer.

[10] Rashad, Ahmed, Salah Kamel, and Francisco

Jurado. "The basic principles of wind farms."

Distributed Generation Systems (2017): 21-

67.

[11] wind electricity generation and share of total

U.S.A electricity generation.[online]

https//www.eia.gov/energyexplained/wind/hist

ory-of-wind-power.

[12] Kisvari, Adam, Zi Lin, and Xiaolei Liu.

"Wind power forecasting–A data-driven

method along with gated recurrent neural

network." Renewable Energy 163 (2021):

1895-1909.

[13] Wang, Shuangxin, et al. "Small-world neural

network and its performance for wind power

forecasting." CSEE Journal of Power and

Energy Systems 6.2 (2019): 362-373.

[14] Khodayar, Mahdi, and Jianhui Wang. "Spatio-

temporal graph deep neural network for short-

term wind speed forecasting." IEEE

Transactions on Sustainable Energy 10.2

(2018): 670-681.

[15] Shahzad, Mirza Naveed, Saiqa Kanwal, and

Abid Hussanan. "A New Hybrid ARAR and

Neural Network Model for Multi-Step Ahead

Wind Speed Forecasting in Three Regions of

Pakistan." IEEE Access 8 (2020): 199382-

199392.

[16] Alencar, David B., et al. "Hybrid approach

combining SARIMA and neural networks for

multi-step ahead wind speed forecasting in

Brazil." IEEE Access 6 (2018): 55986-55994.

[17] Chen, Gang, et al. "Research on wind power

Forecastingmethod based on convolutional

neural network and genetic algorithm." 2019

IEEE Innovative Smart Grid Technologies-

Asia (ISGT Asia). IEEE, 2019.

[18] Paidi, ESN Raju, et al. "Development and

Validation of Artificial Neural Network-Based

Tools for Forecasting of Power System Inertia

With Wind Farms Penetration." IEEE Systems

Journal 14.4 (2020): 4978-4989.

[19] Wang, Cong, Hongli Zhang, and Ping Ma.

"Wind power forecasting based on singular

spectrum analysis and a new hybrid Laguerre

neural network." Applied Energy 259 (2020):

114139.

[20] Lin, Zi, and Xiaolei Liu. "Wind power

forecasting of an offshore wind turbine based

on high-frequency SCADA data and deep

learning neural network." Energy 201 (2020):

117693.

[21] Zhou, Min, et al. "Multi-objective

Forecastingintervals for wind power forecast

based on deep neural networks." Information

Sciences 550 (2021): 207-220.

[22] Wu, Wenbin, and Mugen Peng. "A data

mining approach combining $ k $-means

clustering with bagging neural network for

short-term wind power forecasting." IEEE

Internet of Things Journal 4.4 (2017): 979-

986.

[23] Khodayar, Mahdi, Okyay Kaynak, and

Mohammad E. Khodayar. "Rough deep neural

architecture for short-term wind speed

WSEAS TRANSACTIONS on POWER SYSTEMS

DOI: 10.37394/232016.2022.17.28

Mohammad A. Obeidat, Baker N Al Ameryeen,

Ayman M Mansour, Hesham Al Salem,

Abdullah M. Eial Awwad

E-ISSN: 2224-350X

278

Volume 17, 2022

forecasting." IEEE Transactions on Industrial

Informatics 13.6 (2017): 2770-2779.

[24] Khodayar, Mahdi, Jianhui Wang, and

Mohammad Manthouri. "Interval deep

generative neural network for wind speed

forecasting." IEEE Transactions on Smart

Grid 10.4 (2018): 3974-3989.

[25] Saroha, Sumit, and S. K. Aggarwal. "Wind

power forecasting using wavelet transforms

and neural networks with tapped delay."

CSEE Journal of Power and Energy Systems

4.2 (2018): 197-209.

[26] Zhang, Yagang, et al. "Wind speed

Forecastingof IPSO-BP neural network based

on lorenz disturbance." Ieee Access 6 (2018):

53168-53179.

[27] Shi, Zhichao, Hao Liang, and Venkata

Dinavahi. "Wavelet neural network based

multiobjective interval Forecastingfor short-

term wind speed." IEEE Access 6 (2018):

63352-63365.

[28] Mezaache, Hatem, and Hassen Bouzgou.

"Auto-encoder with neural networks for wind

speed forecasting." 2018 International

Conference on Communications and Electrical

Engineering (ICCEE). IEEE, 2018.

[29] Hur, Sung-Ho. "Estimation of useful variables

in wind turbines and farms using neural

networks and extended Kalman filter." IEEE

Access 7 (2019): 24017-24028.

[30] Zhang, Yagang, Yuan Zhao, and Shuang Gao.

"A novel hybrid model for wind speed

Forecastingbased on VMD and neural network

considering atmospheric uncertainties." IEEE

Access 7 (2019): 60322-60332.

[31] Medina, Sergio Velázquez, and Ulises Portero

Ajenjo. "Performance improvement of

`1artificial neural network model in short-term

forecasting of wind farm power output."

Journal of Modern Power Systems and Clean

Energy 8.3 (2020): 484-490.

[32] Tong, W. (2010). Fundamentals of wind

energy. Wind power generation and wind

turbine design, 3-42.

[33] https://www.eepowerschool.com/wind/wind-

turbine-working-principle.

[34] Alexander Kalmikov, Chapter 2 - Wind Power

Fundamentals, Editor(s): Trevor M. Letcher,

Wind Energy Engineering, Academic

Press,2017,Pages 17-24 ISBN

9780128094518.

[35] Licari, J. (2013). Control of a variable speed

wind turbine (Doctoral dissertation, Cardiff

University).

[36] Zakaria." Robust Control of Wind Turbine via

Pitch Angle Manipulation"(2020).

[37] G. Giebel, R. Brownsword, and G.

Kariniotakis, (2003), “The State-of-the-Art in

Short-Term Forecastingof Wind Power – A

Literature Overview,” Project ANEMOS,

Tech. Rep.

[38] L. Landberg, G. Giebel, H. Nielsen, T.

Nielsen, and H. Madsen, (2003), “Short-Term

Forecasting– An Overview,” Wind Energy,

vol.6, no.3.

[39] G. Giebel, G. Kariniotakis, and R.

Brownsword, (2003), “State-of-the-Art on

Methods and Software Tools for Short-Term

Forecastingof Wind Energy Production,” in

European Wind Energy Conference &

Exhibition – EWEC2003, Madrid.

[40] Gomes, Pedro, and Rui Castro. "Wind speed

and wind power forecasting using statistical

models: autoregressive moving average

(ARMA) and artificial neural networks

(ANN)." International Journal of Sustainable

Energy Development 1.1/2 (2012).

[41] G. Zhang, B. Patuwo, and M. Hu, (1998),

“Forecasting with Artificial Neural Networks:

The State of the Art,” International Journal of

Forecasting, vol.14, no.1.

[42] M. Gardner and S. Dorling, (1998), “Artificial

Neural Networks (The Multilayer Perceptron)

– A Review of Applications in the

Atmospheric Sciences,” Atmospheric

Environment, vol.32, no.14-15

Creative Commons Attribution License 4.0

(Attribution 4.0 International, CC BY 4.0)

This article is published under the terms of the

Creative Commons Attribution License 4.0

https://creativecommons.org/licenses/by/4.0/deed.en

_US

WSEAS TRANSACTIONS on POWER SYSTEMS

DOI: 10.37394/232016.2022.17.28

Mohammad A. Obeidat, Baker N Al Ameryeen,

Ayman M Mansour, Hesham Al Salem,

Abdullah M. Eial Awwad

E-ISSN: 2224-350X

279

Volume 17, 2022