Application of Artificial Intelligence (AI) in Wind Energy System with a Case
Study
Keywords: Wind Energy, Artificial Intelligence, Wind Turbine, Wind Mill, Genetic
Algorithm, Wind Speed, Current Output
Abstract
Renewable energy
is the fastest growing source of clean energy worldwide. The
employment of wind energy is expected to increase dramatically over the next few years. There is
a good source of wind power on the highways due to the movement of vehicles. A small windmill
could utilize the wind power generated by passing vehicles and produce electricity that can power
the lights on the highway. This paper presents the application of artificial intelligence to predict
the current output from a small windmill placed on the highway. The results show a good
concurrence between the experimental and predicted values.
Introduction
The main energy source is from fossil fuels, which is extensively used to meet the demand.
The usage of fossil fuels directly harms the clean environment and also leads to global warming.
Fossil fuels are non-renewable and get depleting, which makes people to focus on renewable
energy sources. All over the world harnessing of solar and wind as a sustainable source of energy
gained popularity to curtail the heavy dependency on fossil fuels and also to counter the global
warming. When wind energy is used to produce electricity, less pollution from conventional power
plants will be released into the environment. The need of concentrating on renewable energy
resources has increased, particularly in the wake of the Gulf of Mexico oil leak and the Japanese
nuclear accident. The installed wind energy capacity reached 196,630 MW globally in 2010, with
37,642 MW added in that year, according to the World Energy Association's report on wind energy
for 2010. Enhancing wind farm design and layout; boosting wind turbine accessibility,
dependability, and efficiency; streamlining the upkeep, assembly, and installation of offshore and
onshore turbines and their substructures; showcasing massive wind turbine prototypes and
expansive, interconnected wind farms, etc. are the main research areas that should be prioritized
in the wind energy industry [1]. The first wind-powered generator was invented by Charles F.
Brush, an electrical pioneer from America, and it produced energy in his backyard. He built a
windmill that was 40 tons in weight and stood 60 feet tall. The actual wind mill measured 56 feet
in diameter. The wind mill had a total of 144 separate blades. 500 revolutions per minute was the
turbine's peak rotational speed. Everything in his basement was wired up to 408 batteries. With
this technology, he was able to power his entire house, including the lab. Up until 1909, his wind
mill operated for 20 years [2]. The wind turbine's size is determined by its intended use. Typically, tiny turbines have a
power output between 20 and 100 kW. The 20- to 500-watt "micro" turbines are smaller and have
a wider range of uses, including the charging of sailboat and recreational vehicle batteries. Water
pumping is one use for turbines ranging from one to ten kW. Grain mills and water pumps have
been powered by wind for millennia. While mechanical windmills remain a cost-effective and
practical choice for water pumping in wind-free regions, farmers and ranchers are discovering
that wind-electric pumping offers greater versatility and doubles the volume of water pumped for
the same initial outlay. Furthermore, mechanical windmills have to be positioned straight above
the well, which could not maximize the wind resources that are available. Electric cables can be
used to link wind-electric pumping systems to the pump motor, which can be installed where the
best wind resource is available. Depending on how much power you wish to create, household
turbines can range in size from 400 watts to 100 kW (100 kW for extremely big loads). An average
household consumes around 10,000 kWh (kilowatt-hours) of power year, or 830 kWh each
month. To significantly meet this requirement, a wind turbine with a rating of between 5 and 15
kW would be needed, depending on the typical wind speed in the region. If the average yearly
wind speed in the area is 14 miles per hour (6.26 meters per second), a 1.5 kW wind turbine can
supply the energy needed for a house that uses 300 kWh per month. Automatic overspeed-
governing mechanisms are included in most turbines, which prevent the rotor from spinning
uncontrollably in extremely strong winds [3].
Harrous and Ahshan [4], [5] developed a hybrid solar/wind system for his home. The
hybrid system consists of Bergey XL-1, a 1000-watt wind generator mounted on a tower 104 feet
tall along with 300 watts of solar, which is a stand-alone system with batteries. The battery bank
is a 220-amp system made up of eight 6-volt batteries wired as a 24-volt system. The system runs
incandescent lights and a well pump at the barn, as well as water through heaters. The cost of the
complete system was around $ 10,000 including equipment, trenching for wires, building permit,
etc. For wind energy uses, there must be open space or accessible coastlines for wind energy
plants. Saudi Arabia is a large nation with extensive coastlines and open spaces. In the majority
of these locations, the wind speed is sufficiently high to make using wind energy cost-effective.
Saudi Arabian authorities will invest billions in this potential field of electricity since they
understand the value of renewable energy, particularly wind energy. Despite its vast oil reserves,
Saudi Arabia is very interested in actively participating in the development of new technologies
for the exploitation and use of renewable energy sources. Despite Saudi Arabia's substantial wind
resource potential, there are several obstacles to its development. These comprise the resource's
erratic nature, its seasonal and diurnal fluctuations, its isolated geographic position, and the
electrical grid infrastructure required to transfer wind energy to load regions. Significant
technological obstacles must be overcome in order to fully utilize Saudi Arabia's wind potential.
The energy balance between the needed load and the generated power, as well as the matching of
the wind turbine and location with an appropriate economic position, remain a significant
problem. By matching the locations and wind turbines, the researcher created an extensive
computer program that does all the calculations and optimization needed to precisely build the
Saudi Arabian wind energy system [6].
Eltamaly et al [7] built and examined the dynamic performance of a novel wind turbine
producing system using a thyristor inverter. The system is basically based on shaft generators,
which are highly reliable and produce high-quality power output and are frequently employed in
big ships. It was looked into if this innovative method could provide low-distortion electric power
at a steady frequency even when the natural wind's velocity fluctuated. Additionally, a dynamic
model was created, and it was discovered to have good agreement with the system's experimental
and simulated results. Zemamou et al [8] investigated the remarkable performance of savonius wind turbines and how they might be used as an alternative to normal wind turbines to extract
valuable energy from air streams. Some benefits of employing this kind of machine include its
straightforward design, high starting and full operation moment, ability to receive wind from any
direction, minimal noise and angular velocity when operating, and reduced wear on moving
components. There have been many suggested modifications for this gadget over time. Another
benefit of employing such a machine is the range of possible rotor designs. The performance of a
Savonius rotor is impacted by each configuration. The performance of a Savonius rotor is
influenced by air flow, geometric, and operational factors. For the majority of settings, the quoted
range for the highest averaged power coefficient is between 0.05 and 0.30. The usage of stators
has also been shown to result in performance increases of up to 50% for the tip speed ratio of the
highest averaged power coefficient.
Renewable energy technologies affect how household power demands are met. Since most
of the energy produced by fossil fuels is used in buildings and their unchecked use is linked to
environmental risks, global warming, and the possibility of their depletion, it will be advantageous
to replace the conventional energy generation system with renewable energy sources [9].
Globally, there is a growing need to transition from fossil fuels to renewable energy sources. The
main causes of this transformation are the lack of fossil fuels and their detrimental consequences
on the environment, particularly the climate. As a result, interest in renewable energy sources such
as solar, wind, and wave energy is growing around the world [10]. Converters for multiphase
generators, back-to-back linked converters, passive generator-side converters, and converters
without an intermediary dc-link for high-power wind energy conversion systems (WECS) are all
included in the low and medium voltage category. The series/parallel connection of wind turbine
ac/dc output terminals and high voltage ac/dc gearbox are taken into consideration while
evaluating the onshore and offshore wind farm layouts [11].
Artificial Intelligence (AI) in Wind Energy Systems
The majority of wind farms are situated in isolated areas or several miles offshore, thus it
is vital to monitor their mechanical parts for maintenance in order to keep them from breaking
down mechanically and perhaps cutting themselves off from the electricity grid [12]. Machine
learning algorithms, particularly artificial neural network ANNs, are commonly used to process
gathered data. The ANN's structure is inspired by real neurons, with basic processing units coupled
by weighted linkages. It contains three major layers: input, concealed, and output. Furthermore,
the number of hidden layers may be increased to construct the deep neural network (DNN)
architecture [13]. The Artificial Neural Network (ANN), Backpropagation Neural Network
(BPNN), Adaptive Neuro-Fuzzy Inference System (ANFIS), and Genetic Algorithm (GA) are
some of the most often used and proven AI approaches. The level of technology today and
potentially uses tried-and-true methods to create AI-powered renewable energy systems,
particularly solar energy systems. To ascertain the state and progress of AI approaches in the field
of renewable energy systems (RES), particularly solar power systems, a number of peer-reviewed
journal publications were analyzed [14].
The physical methods forecast wind energy using meteorological data, such as topography,
atmospheric pressure, and ambient temperature; the hybrid methods combine the advantages of
multiple single forecasting models to obtain the final prediction results through various weighting
strategies; and the intelligent methods process and optimize the integration of external and internal
big data to estimate future wind energy. The statistical approaches anticipate wind energy time
series by an assessment of the probability distribution and random process of the samples. Since
intelligent approaches and AI-based hybrid methods are more efficient at analyzing the complex
connections present in huge data sets, they are essential for increasing energy efficiency, decreasing energy usage, and allowing real-time decision-making in the wind energy business.
[15].
Case Study
A small wind mill was fabricated using wind turbine mounted inside the tube, generator
and a battery pack. The fabricated tubular wind mill was flexible and can withstand turbulence.
The turbine inside the tube rotates in the direction of wind turbulence. Standard generator system
was used which can deliver a power output of 1 kW along with a maintenance free battery pack,
inverter and charge controller. Figure 1 shows the schematic diagram of the wind mill.
Experimental data was recorded by keeping the wind mill on road side platform based on vehicular
movements for 7 days. Duration of data recording on each day varies from 30 to 180 min.
Fig 1. Schematic Diagram of Wind Mill
Artificial Neural Networks (ANN) use genetic algorithms that make use of adaptive
heuristic search methods. A genetic algorithm is a better method for achieving the global
optimum's convergence. Chromosome initialization is the first step in the genetic algorithm's
operation, after which fitness is assessed using an objective function [16]. Chromosomes are
genetically propagated by first selecting the most fit individuals and then using operators such as
crossover and mutation. A multi-objective solution from the optimization toolkit and a genetic
algorithm were used to optimize the process output variable models [17]. Ten neurons or nodes
made up the hidden layer, the output current serving as the dependent output neuron, and time and
wind speed serving as independent input variables were used to create the ANN network model
[18]. Ten neurons made up the hidden layer of the neural network, which was trained until the
mean squared error between the target and model output was as little as possible. The comparison
between the goal values of the present output and the output values of the ANN network model is
displayed in Figure 2 [19]. A high correlation coefficient value across training, validation, testing,
and overall comparison shows that the model can accurately forecast the wind mill's current
production value. The comparison output variables between the experimental investigations and
the ANN-GA projected values are displayed in Table 1 [20], [21]. Output 0.95 Target + 21
Output =0.87*Target +63
Training: R=0.96054
490
°
Data
Fit
480
470
Y T
460
450
440
430
440
460
480
Target
Test: R=0.99205
Output = 0.96*Target + 21
Validation: R=0.99924
480
470
о
Data
Fit
Y T
460
450
440
430
430 440 450 460 470 480
Target
490
480
470
° Data
Fit
........Y=T
All: R=0.97457
460
450
°
440
Output =0.9*Target +46
490
о Data
Fit
480
470
Y T
460
о
450
0
440
°
°
440 450 460 470 480 490
Target
430
440
460
480
Target
Fig 2. ANN network model output values and the target values of the current output
Table 1. Experimental data and ANN-GA (Genetic Algorithm) Predicted Data
Current
[Amp]
(Predicted
Wind Speed
Current [Amp]
Time
[m/s]
(Experimental)
from ANN)
30
6.44
460
459.562
60
6.58
470
469.965
90
6.86
490
487.948
120
6.72
480
475.955
150
6.58
470
470.052
180
6.78
485
485.718
30
6.3
450
450.116
60
6.37
455
455.799
90
6.52
466
466.029
120
6.47
462
463.786
150
6.44
460
460.033
180
6.59
471
471.153
30
6.88
492
489.558/n I have a research paper and I want to transfer it into a clear and engaging PowerPoint
presentation.
I will be presenting in a conference so I need it to look professional and elegant.
PRESENTATION FORMAT: The format for your paper presentation will be oral presentation
(power point).
Attached video in which professor has explained what is required
1st page will be TITLE.
Presentation will be for 15 Minutes (20 slides minimum) + speaker notes required
