Proceedings of the 2nd International Conference on Functional Materials and Civil Engineering

DOI: 10.54254/2755-2721/66/20240949

Research on the application of computer science in wind

power generation and the future trends

Haipeng Shou
School of Energy and Power Engineering, Huazhong University of Science and
Technology, Wuhan, 430074, China

[email protected].

Abstract. In the context of the growing global demand for clean energy, wind power, as an
important form of renewable energy, is gradually occupying a dominant position. At the same
time, the continuous development of computer technology has also brought new breakthroughs
in wind power generation. Traditional wind power generation suffers from low power generation
efficiency and large fluctuations in power generation. However, through the integration of wind
power generation with advanced computer technologies, the efficiency and cost of wind power
generation have been greatly optimized. This paper seeks to provide an overview of how
computer science is utilized in wind power generation, focusing on the integration of
computational modeling, simulation, and machine learning. It highlights future trendsand
emphasizes the significance of developing digital twins and integrating machine learning with
physical models for efficient wind farm management. Combining findings from various existing
studies, this paper delves into the profound implications for the future of energy, highlighting the
transformative role of computer science in advancing sustainable power solutions and shaping
the landscape of renewable energy sources.

Keywords: Wind power, simulation, computer science, machine learning

1. Introduction
With the continuous advancement of global energy transition and carbon neutrality work, the global
power system is decarbonizing, led by the increasing dominance of wind and solar energy [1]. Since
ancient times, people have been able to use wind energy to promote production. In modern times, wind
energy has gradually become one of the main primary energy sources for renewable energy generation.
It is expected that, by 2050, the annual wind power generation will reach at least 20,000 TWh [1].
However, due to the natural volatility and temporal and spatial instability of wind energy, the low
utilization rate of wind energy and the unstable impact on the power grid hinder the development of
wind power generation. But with the development of computer science in recent decades, many
computer technologies have been successfully applied to wind power generation, greatly reducing the
potential impact of wind energy on the power grid and improving the efficiency of wind energy
utilization. This paper focuses on summarizing the applications of computer modeling and simulation,
machine learning, and neural networks in wind power generation. Meanwhile, it analyzes and
summarizes the shortcomings and prospects of existing research on how computer science can be better

© 2024 The Authors. This is an open access article distributed under the terms of the Creative Commons Attribution License 4.0
(https://creativecommons.org/licenses/by/4.0/).

192
 Proceedings of the 2nd International Conference on Functional Materials and Civil Engineering
DOI: 10.54254/2755-2721/66/20240949

applied in the field of wind power generation in the future, so as to approach the goal of carbon neutrality
in human society by the middle of this century.

2. Computational Modeling in Wind Power Generation

2.1. Computational Modeling for Wind Turbine Design

There are mainly two types of wind turbines in the world today: vertical and horizontal, and the
horizontal three blade structure wind turbine (HAWT) accounts for the vast majority of them [2]. Due
to the fact that its three blades rotate around the central axis, the line velocity at the point further away
from the axis of rotation on the blades increases while the angular velocity remains constant. This results
in different relative speeds of the blades at different radii for the same wind speed. Therefore, the blade
geometry of wind turbines should be different at different radii. In the past, a blade with high wind
energy utilization could only be obtained through more complex calculations. With the maturity of
computer modeling technology, the blade design work of wind turbines can now be completed on
computers. At the same time, computer simulation technology can be used for verification and model
optimization, in order to maximize the wind energy capture rate of wind turbines.
In the design process of wind turbine blades, the material of the blades plays a crucial role. Its
physical and chemical properties, such as strength, quality, and corrosion resistance, will affect the
lifespan and power generation efficiency of wind turbines. By using computers for modeling and
simulation, existing digital models of different materials can be used to simulate various blade designs
in real-life situations, thereby reducing waste of real materials and finding the most suitable blade
materials and designs. This design process not only reduces the production cost of wind turbine blades,
but also greatly improves design efficiency and blade usage effectiveness. Appadurai et al. used finite
element analysis to design wind turbine blades using carbon epoxy values, graphite epoxy resin, and
steel materials for stress simulation [3]. After a comprehensive analysis of quality and blade strength,
they ultimately found that carbon epoxy composite material is the most suitable material for
manufacturing wind turbine blades.
Computer modeling and simulation can also be applied to the study of load and dynamic
characteristics of propeller blades. Numerical simulation can be used to analyze the performance of wind
turbines under different environmental factors, such as wind speed, force conditions of propeller blades,
vibration response, etc. This kind of simulation modeling can optimize the design, improve the
efficiency and stability of wind turbines. Jucuan Dai et al. established a real-time dynamic simulation
model of wind turbine blade load in the Simulink environment, and ultimately found that the flap torque
is mainly affected by aerodynamic torque, with gravity having a greater impact on the edgewise torque
of wind turbine blades [4].

2.2. Macro Simulation and Digital Twin in Wind Farm

In addition to modeling and simulation of wind turbine blades, at the macro scale, modeling and
simulation of the entire wind farm also have a high guiding role in optimizing the establishment of wind
farm structure and improving wind energy utilization efficiency. The power generation of wind farms
largely depends on their layout, scale, and degree of direct matching with the power grid. Due to
differences in wind direction and speed at different locations within the same power plant, wind farms
often experience power fluctuations. Without appropriate reactive power compensation equipment,
voltage fluctuations in wind farms can be quite significant [5]. Establishing a computer model during
the construction and research of wind farms can help simulate major issues such as power generation,
operation status, and system response. This allows for a quantitative description of power fluctuations
in wind farms from the beginning and optimization of power fluctuations through simulation. It can also
provide a solution for selecting the site of each unit during wind farm construction. The model used to
describe the wake of wind power generation has been developing since the 1980s [6].
Nowadays, various fluid dynamics models can run well on computers and produce conclusions for
optimizing wind farms. People have found that computer simulation has a natural and significant

193
 Proceedings of the 2nd International Conference on Functional Materials and Civil Engineering
DOI: 10.54254/2755-2721/66/20240949

advantage in evaluating the macro level effects of the wake effect and other interactions between wind
turbines in wind farms. Through computer models, simulation can be used to coordinate the direction,
distance, and position of wind turbines, in order to reduce the impact of wake effects on wind power
efficiency and maximize the utilization of local wind potential. Just like F Gonz á lez Longatt et al. used
MATLAB to establish a wake model for wind power generation, and analyzed in detail the influence of
wake effects on steady-state and dynamic behavior by changing factors such as wind direction, wind
speed, and distance between generator units [7]. In the case where all details of the wind farm are
unknown, the team’s conclusion can serve as an empirical rule for general evaluation.
The current digital models for wind farms are mainly divided into three types: low-fidelity, medium-
fidelity, and high-fidelity wind farm models [8]. Among them, the low-fidelity model is often a static
model used to reflect the state of a wind power average state; the medium-fidelity model is already a
dynamic model that can reflect more flow field characteristics and details but often uses simplified
physical equations; and the high-fidelity model provides a high-precision wind farm model at the cost
of computational complexity.
With the continuous development of computer technology, the computing power of computer CPUs
and GPUs has increased exponentially. A real-time digital model that reflects the real situation of things
has emerged in people’s vision, which is the digital twin model. In the field of wind power, digital twins
can also provide solutions for wind farm monitoring, control, and optimization. Due to the fact that the
physical quantities established through simulation in the digital twin model can correspond one-to-one
with the various physical quantities of the real world wind power field, it can achieve the control and
optimization goals of the generator unit entity through real-time feedback and model interaction. The
future application of computer simulation and modeling in wind farms should pay more attention to the
real-time and punctuality of physical models. However, at the same time, we must also see the real
physical situation, such as the impact of complex fluid dynamics, which brings challenges to the
establishment of high-fidelity real-time digital twin models for wind farms.

3. Application of Machine Learning in Wind Power

3.1. Overview
Machine learning is a technology in the field of artificial intelligence. With the release of artificial
intelligence models such as ChatGPT in recent years, the application of artificial intelligence in
industries, medicine, and other fields has gradually been accepted by the public. Establishing a digital
twin model for an industrial system can monitor and control the entire system in real-time, but the
patterns or patterns of certain physical characteristics of the entire system still cannot be well grasped
by people, resulting in cost waste or low production efficiency. Machine learning technology, on the
other hand, can provide effective solutions for such problems. By learning and training from industrial
system data, it can grasp the patterns of the data and make predictions or decisions. In the field of energy,
the characteristics of machine learning can be well adapted to multiple factors in wind power generation.
Wind power is greatly influenced by natural factors, and both wind speed and direction will affect the
power generation of wind turbines. Therefore, machine learning the laws of local environmental wind
will help to quickly and accurately adjust the orientation and blade angle of wind turbines, and improve
wind energy utilization. At the same time, the power fluctuations generated by wind farms will also
exert significant pressure on the power grid. The use of machine learning technology to predict wind
power can also provide early warning and decision support for peak shaving of the power grid. The
application of machine learning technology in wind power mainly revolves around the two aspects
mentioned above. This paper elaborates on the application of machine learning technology in
environmental factors and power generation.

3.2. Machine Learning in Wind Forecasting

When machine learning technology was not yet mature, weather forecasting was the main source of
information for wind farms to predict wind energy for the next few days or months. However, weather

194
 Proceedings of the 2nd International Conference on Functional Materials and Civil Engineering
DOI: 10.54254/2755-2721/66/20240949

forecasting often involves macroscopic predictions of a larger area, while it is difficult to achieve
accurate predictions of more microscopic physical characteristics such as wind speed and direction in
the local wind farm. Machine learning technology can generate learning models based on on-site
monitoring data from wind farms over the past few years, allowing for more accurate predictions of
future wind speeds and directions at the micro level. On a time scale, current environmental wind
forecasting is mainly divided into four types: very short term, short term, medium term, and long term
[9]. Machine learning technology is mainly applied in short-term and medium-term forecasting. Very
short term prediction focuses on predictions ranging from a few seconds to a minute, mainly predicting
wind speeds affected by turbulence and local meteorological conditions, while short-term and medium-
term predictions are influenced by variations in thermal properties between the atmosphere and the
ground, as well as changes in local weather patterns.
After decades of development, today’s machine learning has incorporated artificial neural networks
(ANNs) and may also couple with other mathematical and physical models to improve the accuracy of
wind speed prediction. In [10], Chang et al. used multiple different neural network-based models,
IRBFFNN-EF, for short-term wind speed prediction and proposed a new prediction method with error
feedback, proving that the model helps to maximize wind energy capture. And M.M. Ardehali et al.
integrated the artificial neural network with Markov Chain and used it for short-term wind speed
prediction, while the Mc method was used for mid-term wind speed pattern analysis and prediction [11].
Also, LSTM(long short-term memory neural network), SVRM, and EO three computer-based
algorithms were coupled by the investigators [12]. The resulting EnsemLSTM model can provide more
accurate wind speed forecasting. Therefore, from existing research, coupling with mathematical or
physical models can enable machine learning models to have higher prediction accuracy or time span,
thus enabling better application in wind power prediction. However, traditional machine learning models
based solely on neural networks are constantly being eliminated and banned. In future research, more
attention should be paid to the integration of major data and artificial intelligence, as well as the
integration of artificial intelligence and physical models in the real world, in order to make new energy
generation more stable and predictable.

3.3. Machine Learning in Wind Power Prediction

The power of wind power generation fluctuates due to fluctuations in wind energy, and today, AC power
grids around the world operate at a fixed frequency. This type of power grid requires each power
generation equipment connected to the grid to have a fixed phase. Therefore, it is necessary to equip
rectification and power stabilization facilities at the wind power generation end to reduce the impact of
wind power instability on the power grid. At this point, predicting wind power will have multiple
benefits for the operation of the wind power system. Firstly, it can actively improve the stability and
reliability of the power grid, enabling the power system to better cope with fluctuations and changes and
reducing the supply-demand imbalance caused by an unstable power supply. Secondly, predicting wind
power can help optimize the operation and scheduling of the power system, improve power generation
efficiency, and reduce energy production costs. Meanwhile, accurate power prediction helps power
companies better plan future electricity demand and corresponding investments, promoting the
sustainable development of renewable energy. Similar to wind prediction, machine learning a dataset of
the power generation of a wind farm over a period of time can enable computer models to predict power
levels for a future period of time based on existing data. Many machine learning algorithms used for
environmental wind prediction can also be applied to wind power prediction.
Sina Ibne Ahmed er al. used two machine learning algorithms, Gradient Boosting Machine (GBM)
and Support Vector Machine (SVM) to predict the medium- and long-term wind power generation, and
they will incorporate neural networks in future work to better handle the randomness of data [13].
Researchers have also proposed a novel hybrid data driven model based on the concepts of deep
learning-based convolutional-long short term memory (CLSTM), mutual information, evolutionary
algorithm, neural architectural search procedure, and ensemble-based deep reinforcement learning (RL)
strategies [14]. Compared with various advanced wind power prediction models in the world at that time,

195
 Proceedings of the 2nd International Conference on Functional Materials and Civil Engineering
DOI: 10.54254/2755-2721/66/20240949

this prediction algorithm has excellent performance. It can be seen that constantly updated neural
networks and machine learning algorithms will be closely related to wind power generation in the future.
For the entire wind power system, wind power prediction should be combined with digital twin systems
in the future, so that wind power utilization can be accurately controlled from the generation end to the
grid end.

4. Conclusion
This paper mainly discusses the modern utilization of computer modeling and simulation, machine
learning in wind power. These methods of utilizing computer technology greatly increase the power
generation efficiency of individual wind turbines and wind farms, and reduce the volatility and
predictability of their power generation. With the continuous development of emerging computer
technologies such as artificial intelligence in recent years, the impact of wind power instability on the
power grid will be further reduced, thereby accelerating the transition from traditional energy to
renewable energy. In the future, in terms of computer modeling and simulation, wind farm models
should develop towards high-fidelity real-time and data-driven digital twin models so as to make the
control of wind power systems more precise and efficient. And wind power prediction should not be
confined to the traditional neural network prediction mode, but introduce physical and mathematical
models into machine learning, allowing computers to achieve higher prediction accuracy under
supervised learning. Meanwhile, the above two technologies should be integrated and used to achieve
higher system control efficiency and accuracy. Due to the limitations of computer models and the
complexity of the real physical world, the challenges for the future application of computer technology
in wind power generation will lie in the development of these two aspects.

References
[1] bp. (2023) bp-energy-outlook-2023. bp Energy Outlook 2023.
[2] Peter J. Schubel, Richard J. Crossley. (2012) Wind Turbine Blade Design. Energies 2012, 5, 3425-
3449.
[3] M. Appadurai et al. (2021) Finite Element Analysis of Composite Wind Turbine Blades. 7th
International Conference on Electrical Energy Systems (ICEES).
[4] Juchuan Dai, Wei Hu, Xiangbin Shen. (2017) Load and dynamic characteristic analysis of wind
turbine flexible blades. Journal of Mechanical Science and Technology 31 (4) (2017)
1569~1580.
[5] Z. Chen, F. Blaabjerg. (2009) Wind farm—A power source in future power systems. Renewable
and Sustainable Energy Reviews 13 (2009) 1288–1300.
[6] Ainslie JF. (1988) Calculating the flow field in the wake of wind turbines. Journal of Wind
Engineering and Industrial Aerodynamics 1988;27:213e24
[7] F. González-Longatt*, P. Wall, V. Terzija. (2012) Wake effect in wind farm performance: Steady-
state and dynamic behavior. Renewable Energy 39 (2012) 329e338.
[8] Hongyang Dong, Jingjie Xie, Xiaowei Zhao. (2022) Wind farm control technologies: from
classical control to reinforcement learning. Prog. Energy 4 (2022) 032006.
[9] Jannik Schütz Roungkvist, Peter Enevoldsen. (2020) Timescale classification in wind forecasting:
A review of the state-of-the-art. https://doi.org/10.1002/for.2657.
[10] G.W. Chang a, H.J. Lu a, Y.R. Chang b, Y.D. Lee. (2017) An improved neural network-based
approach for short-term wind speed and power forecast. Renewable Energy 105 (2017)
301e311.
[11] S.A. Pourmousavi Kani, M.M. Ardehali. (2011) Very short-term wind speed prediction: A new
artificial neural network–Markov chain model. Energy Conversion and Management 52
(2011) 738–745.
[12] Jie Chena, Guo-Qiang Zengb, Wuneng Zhoua, Wei Dua, Kang-Di Lua. (2018) Wind speed
forecasting using nonlinear-learning ensemble of deep learning time series prediction and
extremal optimization. Energy Conversion and Management 165 (2018) 681–695.

196
 Proceedings of the 2nd International Conference on Functional Materials and Civil Engineering
DOI: 10.54254/2755-2721/66/20240949

[13] Sina Ibne Ahmed et al. (2021) Forecasting of Mid- and Long-Term Wind Power Using Machine
Learning and Regression Models. 2021 IEEE Kansas Power and Energy Conference (KPEC)
| 978-1-6654-4119-3/21/$31.00 ©2021 IEEE | DOI: 10.1109/KPEC51835.2021.9446250.
[14] Seyed Mohammad Jafar Jalali et al. (2022) New Hybrid Deep Neural Architectural Search-Based
Ensemble Reinforcement Learning Strategy for Wind Power Forecasting. IEEE
TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 58, NO. 1,
JANUARY/FEBRUARY 2022.

197