Climate change has become one of the most pressing global challenges the world faces today. It has significant implications for the environment, societies, and economies around the world [1,2]. This has led to an evolution of the global energy landscape, driven by the urgent need to reduce greenhouse gas emissions and secure a sustainable future. There is now widespread adoption of renewable energy generation (REG) sources and related technologies, and the power generation capabilities are growing faster than the overall power demand [3].
One of the leading REG sources is wind energy [4, 5, 6, 7] (Fig. 1). Energy is produced by converting the kinetic energy of moving air into rotational and eventually electrical energy, through the use of wind turbines. Over the years it
has evolved from a niche power source to a major source of renewable energy globally. Wind energy generation (WEG) has been identified as a clean and renewable source that does not contribute towards greenhouse gas emissions or air pollutants. It is an abundantly available energy source that enables countries to diversify their power generation portfolio and increase their energy independence. The recent technological advancements in WEG technology allow for the development of highly efficient and reliable WEG systems [8]. These advancements, along with the economies of scale, have significantly reduced the capital and operational costs of both offshore and onshore wind energy generation [9]. In 2020, the UK generated more than 75 terawatt hours (TWh) of electricity from WEG projects alone [10]. Due to the emerging trend of WEG, there are various policies and regulations in place to promote their development. A total of 150 countries had adopted REG targets by 2017, and a total of 126 had implemented dedicated regulations and policies for them [11, 12, 13]. Thus, with the growing demand for WEG across the world, it is important to identify, assess and address the challenges faced in this domain. The following blog discusses various challenges faced in this domain and the applicability of AI to address those challenges.
This blog provides a description of the challenges faced in the WEG domain. It describes why AI should be used to address these challenges, and how it can be implemented. The wider ethical implications of the use of AI in this domain and a conclusion is provided towards the end.
What are the challenges in this domain?
As is the case with most REG systems, WEG comes with a complex set of challenges. WEG is highly dependent on wind availability, speeds, and power. Sudden changes in wind speeds can lead to fluctuations in the power output. Furthermore, based on how the Wind Farm (WF) is structured, the wake-loss experienced due to the positioning of the Wind Turbines (WTs) and the direction of the wind, can also have an impact on the amount of energy produced. However, due to the complex nature of wind flow in WFs, it is difficult to forecast wind characteristics using conventional statistical methods. Additional variations are added by seasonal changes in wind movement and other meteorological phenomena. In order for WEG to be used successfully and efficiently to meet grid demands, it is important to be able to forecast the amount of energy generation.
Due to the high CapEx costs of developing WFs, it is important to evaluate the location and layout of WFs before construction. A location with strong and constant wind is the ideal location for WF development but determining such locations is tricky. Ideally, long-term wind speed and direction data can be collected accurately using existing meteorological technologies. However, in reality, it is not possible to set up these technologies at the scale required for such data collection due to the large associated costs [14, 15]. It is difficult to forecast wind speeds, intensities, and direction even for short time durations. When developing WFs, these conditions need to be forecasted for the lifetime of the WTs. Terrain elements such as hills and valleys for onshore, and waves and tides in offshore WFs can also have an effect on the wind. Furthermore, the layout of the WTs itself needs to be so that wake-loss effects and turbulence can be mitigated for maximum WEG production. The environmental effects of WF development need to be accounted for as well, as it is likely that the development of such infrastructure can have an impact on the wildlife in the surrounding areas.
A common approach to tackling the reliability issue is using hybrid energy systems, i.e., combining WEG with other energy generation and storage systems. A common one is solar energy, due to their complementary patterns of availability [16]. Solar energy also tends to be inherently intermittent [17]. Thus fluctuations in power output from wind and solar energy can lead to instabilities in voltage and frequency control. Often these infrastructures employ Energy Storage systems (ESSs), which are capable of enabling a more reliable energy output as well as managing fluctuations in voltages. Integrating WEG with ESSs is another challenge as the lifespan of ESSs is sensitive to the amount of energy stored in them. Thus, in order to maximize the lifespan of ESSs, it is important to optimize energy distribution and storage amongst these systems.
WTs are made up of complex components, such as gearboxes, bearings, generators, blades, and various control systems, that require constant maintenance and careful configurations. Resolving failures and maintenance can result in long downtimes which affect power outputs, especially considering the accessibility of WTs. Due to the high maintenance and repair costs of WFs [18], it is important to monitor the performance and health of WTs to ensure faultless operation and quick, predictable repairs. Faults can occur in any of these systems during the varying stress conditions faced in the dynamic environment. Thus, comprehensive monitoring and diagnostic methodologies are required for the accurate detection and prediction of faults in these systems.
Why should AI be adopted in wind energy systems?
As mentioned in the previous section, WEG is a complex process that poses considerable challenges. It is important to first understand what types of challenges need to be addressed, as this creates a better platform for finding solutions. Most of these challenges can be grouped into the following categories.
- Forecasting. One of the main forecasting requirements for WFs is predicting wind patterns. Being able to accurately do so allows for a better understanding of what future power generation trends will look like. Predicting wind speeds and intensities over a long term for an area can also allow for determining optimal locations for WFs. Additionally, being able to predict faults and failures in existing WT hardware can enable quick and smooth repairs with little to no operational downtimes.
- Optimization. Various aspects of the WEG process can be improved with effective optimized strategies. Models can be used to determine optimal layouts of WTs which can maximize power output and minimize costs. Models can also be used in areas such as grid integration, ESS management, and maintenance scheduling to increase the reliability and stability of WEG output.
A combination of these approaches along with others can offer significant benefits for the WEG domain. Various AI models can be used to address these issues.
- Energy Generation Forecasting. Due to the high dependence on varying weather conditions, wind energy forecasting proves to be an important yet tricky challenge to tackle. Data of different granularities and types is required for forecasting short-term versus long-term wind energy production [19, 20]. Energy generation can be optimized by analyzing wake loss, weather data/patterns, and other factors in order to adjust energy output for wind turbines [21]. Considering how sensitive wind speeds and wind power are to the terrain, topology, atmospheric pressure, and dynamic weather conditions, accounting for these parameters in predictive models can lead to more realistic results. AI Models are capable of addressing such complex problems with reasonable accuracy, and can thus enhance the reliability and quality of WEG.
- Size, Site, and Layout Selection. Long-term wind characteristics, environmental considerations (such as ecological factors), topographies, and community engagements are some of the parameters that need to be considered for optimal WF location and layout. AI algorithms are capable of analyzing a large amount of data and parameters to solve multiobjective optimization problems. Various forecasting and predictive models can be used to predict wind speed and intensities to determine the best locations for WF development.
- Grid Integration and Configuration. AI can be used in numerous ways to optimize the grid integration process for WEG. Models can be used to enhance the stability of the grid by predicting fluctuations in WT power out-
puts. The process of resource distribution and the integration of multiple sources of energy can be optimized for maximum power generation and minimum capital and operational costs using various AI models. - Energy Storage System Management. Various AI models can be used to optimize the configuration and distribution of ESS infrastructure to increase their lifespan and balance voltage fluctuations. The reliability and predictability of wind power output can be optimized using efficient ESS control strategies [22].
- Monitoring and Fault Detection. WT monitoring can produce a large amount of data about the different components inside it. This data can indicate their stress levels, physical attributes (such as temperatures), and performance. The analysis of such comprehensive data can allow AI models to accurately detect and predict faults in WT components. This can allow effective mitigation measures to be put in place before the fault occurs, in order to reduce the operational downtime of WTs.
- Maintenance Scheduling. AI models can be used for optimally scheduling the maintenance of wind turbines during their two-decade lifespan allowing for maximizing the power generation output and minimizing the operational and maintenance costs. These models can learn from historical WT data (such as performance metrics, maintenance logs, and sensor readings), resource availability, weather conditions, and other maintenance requirements (such as regulated inspections) to determine optimal strategies for performing WT maintenance that minimizes energy production downtime.
How can AI be adopted in wind energy systems?
n the last few years, AI publications targeted towards the energy sector have experienced a massive increase. Various AI models such as artificial neural networks (ANNs), genetic algorithms (GAs), particle swarm optimization (PSO), and fuzzy logic controllers (FLCs) can be used to address the challenges in the WEG domain. The following section discusses how these models can be used for this purpose.
Relevant AI models
ANNs are computational models inspired by the structure and functioning of biological neural networks, such as the ones in the human brain. ANNs are extremely powerful and versatile models which are widely used for pattern recognition, data processing, and decision-making tasks. They have gained widespread popularity due to their ability to handle complex, non-linear relationships in data, making them suitable for a wide range of applications. ANNs are also used for clustering tasks by finding data similarities, function approximation where a theoretical model cannot be defined, forecasting future behavior by using time series data, and optimizing functions based on given constraints. Based on the internal structure of ANN and the functions of different layers, different types of ANNs can be created. Some of these include Convolutional NN (CNN), Feed-Forward NN (FFNN), Back Propagation NN (BPNN), Recurrent NN (RRN), and Radial Bias Function NN (RBFNN).
Similar to ANNs, GAs are a type of optimization algorithm that is inspired by biology, specifically genetic evolution. They operate on the survival of the fittest principle to find the most optimal solution for a given problem. GAs iteratively generate numerous potential solutions for the problem and evaluate their fitness in order to determine the most optimal solution. Due to the inherent flexibility and problemspace independence of GAs, they prove to be extremely useful in solving complex problems.
PSO is another commonly used algorithm. It is a population-based algorithm inspired by the behavior of social organisms. It can be compared to the flocking of birds or the schooling of fish. The algorithm simulates the movement of these particles in the provided search space in order to find optimal solutions based on the observed particle interactions. Similar to GAs, the PSO algorithm offers flexibility which makes it appropriate for complex domains.
These models are often paired with Fuzzy Logic (FL) algorithms, which enable the representation of an uncertain and unpredictable domain. Fuzzy sets allow for modeling a domain with a partial functional relationship between its members. This makes these algorithms robust to noise in uncertain data. Thus, FL algorithms tend to be used in decision support systems and for approximate reasoning, where the raw data might be uncertain, incomplete, or imprecise.
Energy generation forecasting and optimization
It is found that physical models, like numerical weather prediction (NWP), tend to be better suited for long-term wind speed forecasting, whereas statistical models [23], are more efficient for short-term predictions. Very short-term predictions (in the range of seconds) are useful for turbine control. Due to the quick response times needed for such predictions, the efficiency of the algorithm and the computational cost are important factors. There are numerous studies in this field [24, 25], amongst which is a wavelet-based network integrated with PSO for very-short term wind speed forecasting [26]. The non-linear dynamic behavior of wind is modeled using a wavelet network. PSO is used to learn the training problem as an optimization problem in order to improve the efficiency of ANNs used for forecasting.
For short-term wind speed forecasting, research has shown that FFNNs are capable of accurately predicting the wind speeds for coastal regions with complex topologies [27] (Fig. 2). This technology exploits the ability of ANNs to
perform spatial predictions and incorporate unstable characteristics of the wind, to accurately estimate hourly wind speeds. Other techniques for short-term forecasting utilize models such as BPNN [28, 29], RBFNN [29, 30], RNN [31], and other hybrid models. The evaluation demonstrates that these ANN-based models are capable of forecasting wind speeds more accurately and efficiently than traditional methods.
Wavelet-based ANN models are also used for medium-term (up to 24 hours) wind speed forecasting [32, 33]. These methods allow realistic modeling of wind characteristics and offer scope for training optimizations. Due to the complicated
nature of these problems, hybrid models, such as a combination of multi-layered perceptrons (MLP) with multi-objective GAs [34], and integrating elman RNNs with other machine learning techniques [35], prove to be the most effective at tackling such issues.
Long-term (more than 24 hours) predictions rely on NWP methods, which utilize meteorological and topological information. This allows the models to learn from realistic data but also increases their complexity and inaccuracy. There are limited models for long-term wind speed forecasting [36, 37,38, 39]. Existing models are only capable of providing moderately accurate forecasts with low granularity [15].
Size, site, and layout selection
The determination of a suitable site, layout, and size for a WF is a crucial step before the development of a WF can begin. According to the wake decay model [40], the wind speed in the install location of the turbine directly affects the power generated by the turbine. Due to the difficulty of obtaining long-term wind speed data, it is usually collected through means such as satellite imaging, simulation models, and sparse meteorological data [41].
Various AI models can be implemented to determine optimal size locations, and turbine layouts for WFs, by considering factors such as available wind energy and speed [42, 43]. The goal is to maximize energy production while minimizing costs. It is also important to take into account that the logistical requirements such as the internal electrical infrastructure and transportation are auxiliary parameters that might not help increase the revenue but are essential for the operation of WFs.
[44] uses GAs to optimize the layout of a WF by considering the economic model (economies of scale and overlapping wakes). Over the years, many improvements have been made to this process [45]. The economic [46,47,48] and wake-effect models have also become more realistic. Factors such as electrical and civil infrastructures can also be modeled. [49] proposes a GA optimization model that divides an offshore WF into smaller areas based on residual wake effects and then optimizes each area using an economic model to maximize the net present value (NPV), by minimizing the CapEx costs and power losses and maximizing the generation of energy and revenue (Fig. 3).
[50] compares the effectiveness of ANN-based models to existing statistical and mathematical models. This model was able to take into account the wind speed and direction along with several other factors such as terrain morphology, type of turbines, and associated costs, in order to calculate the optimal number of WTs in a WF.
It is also important to consider the various regulatory, environmental, and community engagement-related criteria for WF site selection [51, 52, 53, 54, 55]. These can be addressed by parameterizing such criteria in the model. For instance, onshore projects in New York can take advantage of the ”Important Bird Areas” dataset [56] (Fig. 5) to help protect bird habitats from the noise and turbulence caused by WFs [57]. Other studies take factors such as fish habitats, commercial fishery efforts, navigable waterways, proximity to powerlines, and noise pollution into account as well [55, 58, 59, 60].
Grid integration and configuration
Various AI models can be used to optimize WEG output stability, reliability, and performance to match grid demand expectations [61]. These include ANNs, PSO, GA, and independent component analysis (ICA), amongst others. It is common for WEG to be combined with other renewable or conventional energy sources along with arrays of battery storage systems in order to develop a hybrid architecture that is capable of supplying a constant reliable source of energy. Most commonly, onshore wind energy grid configurations are integrated with solar energy (with photovoltaic) generation. Solar energy generation is generally negatively correlated with WEG and thus offers a complementary solution that promotes reliable and stable grid output [16]. In this case, in addition to optimizing the number, type, and capacity of the WTs, the number, type, capacity, and tilt angle of the photovoltaic cells also need to be optimized.
GA and PSO are generally used for microgrid size and configuration while using other AI techniques to improve their efficiency and reduce the computational burden. In order to produce these configurations, various economic, operational, and environmental variables are considered. GA and PSO are often employed with cost-based objective functions.
Consider the case stated in [62] which was implemented in Indonesia; GAs were successfully used to optimize a standalone hybrid power generation system (Fig. 6) to supply the load demand with high reliability under varying weather conditions. This study aimed to combine the power generated by micro-hydro and WT units, which are both susceptible to weather conditions. In order to ensure reliable power without the use of oversized batteries (which increase the cost of the system), it was important to design an optimal system with minimal annual costs. Thus, GAs were used with the objective to minimize the cost of the system over its 20 years of operation, subject to providing a reliable supply of energy. Similarly, [63] evaluates the effectiveness of algorithms such as GA, PSO, and ICA in order to minimize total costs against varying selling and buying electricity tariffs, and [64] discusses approaches to optimize an exhaustive set of costs (including capital, replacement, operational, maintenance, and fuel) while accounting for economical parameters like interest rates, capital recovery factor, sinking fund factor, etc, using GAs.
PSOs (and variations) are also used in numerous microgrid optimization papers [65, 66, 67].
Energy storage system mangement
It is common for WEG to be coupled with ESSs. As WEG is inherently intermittent, ESSs enable a stable and continuous power supply by time-shifting the energy output. When excess energy is produced due to high wind speeds, the surplus energy after grid demands are satisfied can be stored in ESSs. Similarly, when there are low wind speed periods and thus low power generation from the turbines, the stored energy from these systems can be used to supply power to the grid [68,69]. These systems are also capable of managing the voltages being supplied to the grid or other systems such as generators [70].
The lifespan of ESSs (especially Battery Energy Storage Systems (BESSs)) is highly sensitive to the upper and lower limits of the state of charge within the system. Depending on the requirements of storage, these systems need to be capable of quickly and efficiently storing energy in the battery for durations that could range from hours to days [71], and even to months in some cases [72]. Thus, ensuring the reliability and effectiveness of ESSs complements efficient power generation from wind energy systems. This can be achieved by optimizing the ESS system configuration and the energy control strategy [61].
PSOs can be used to optimize ESS system configurations for multiple objectives such as maximizing the efficiency and lifetime of the ESS while minimizing the operational costs [73, 74]. PSOs can be combined with other models such as Teaching Learning ( [75] aims to optimize ESS scheduling over a period of time to optimize emissions and costs), and Bee Colony Algorithm ( [76] aims to optimize the ESS system distribution and sizing to minimize total power loss, electrical energy cost, and emissions produced while maximizing its voltage stability index).
ANNs in combination with dynamic programming have been used to optimize the power flow balance in ESS by minimizing power loss and battery current volumes to increase its lifespan [77]. Research suggests that by using ANNs for optimizing the control strategy of ESSs the reliability and predictability of wind power plants can be significantly improved while minimizing overall costs [22]. PSO and GA, paired with FLCs, can also be used for optimizing the energy control strategy. [78] employs an FLC to optimize the hybrid ESS to minimize energy costs over a given operational window while using PSO to optimize electricity price and load, wind speed, solar irradiation, and ambient temperature data. This study demonstrated that this optimization could reduce operational costs by 18.72% and improve the state of charge by 16.89%. [79] proposed an FLC optimized with GA with the objective to smooth and match the fluctuations of intermittent wind power output to the demand of the grid, and to keep the ESS devices operational in an appropriate state of charge. This optimization was able to reduce the voltage fluctuations in the ESS by 43.46%.
Monitoring and fault detection
A host of algorithms can be used to improve monitoring and fault detection methodologies [42,80], these include Bayesian networks, principal component analysis (PCA), clustering algorithms, and ANNs. Data-driven models can be used to estimate the load and stress on WTs, to determine a reliable fatigue estimator for its entire operational life. Load monitoring and fault detection are some of the most common applications of ANNs in this space. Fault detection is carried out most commonly for the gearbox, blades, and bearings of a WT. Out of these components, the gearboxes are most commonly studied due to the important role they play in rotating machinery and their likelihood of failure [81].
The two main data sources used for fault detection are Supervisory Control and Data Acquisition (SCADA) [83,84,85] data and vibrational signals. Typical SCADA parameters relevant for monitoring and fault detection include wind data (like speed and deviations), WT performance data (like power output, blade pitch angle, rotor speed), vibrational data (like the accelerations of the drive train and tower), and temperature data (like temperatures of the bearing and gearbox). This technology has the massive benefit of not requiring the installation of new components on WFs. Additionally, in most cases, SCADA data collected from a turbine can be used for training models for other turbines as well (given that they are the same types of turbines) [86].
SCADA-based systems aim to protect WTs from an excess operational load, which can result in the degradation of the turbine’s performance and lifetime. These systems are capable of monitoring a huge number of components in the drive train, nacelles, rotors, support structures, and various other electrical modules. Fault detection using ANNs and SCADA data was tested in Portugal [87] in a WF with thirteen 2MW turbines and proved to be an effective method of early fault prediction. The model was able to predict oil leaks in the gearboxes 4 months in advance.
Faults in bearings usually come in the form of cracks and have the potential for causing long downtimes. [88] utilizes SCADA data with ANNs to predict faults in bearings approximately 90 minutes in advance with an accuracy of 97%. Similarly, [86] uses SCADA data such as power output, wind speeds, and component temperatures and vibrations, to train a BPNN model that can predict faults in the bearings as early as 10 days in advance.
WT blades can be monitored by analyzing ultrasonic signals and visual sensors [89]. CNNs are widely employed for visual sensor data analysis, which involves measuring the rate of change of blade pitch angles measured by sensors in each blade.
To improve the reliability and reduce the false alarm rates of monitoring and fault detections, studies have employed methods to denoise SCADA data and reduce fluctuations in prediction error using ANNs [90, 91] and Adaptive Neuro-Fuzzy Inference Systems (ANFIS) models [92].
Maintenance scheduling
Various existing linear [93] and non-linear [94, 95] integer programming approaches are widely adopted for this purpose. In these cases, objectives are set to maximize the revenue or asset reliability or to minimize the costs of the WF. How- ever, in order to structure the problem within these models, certain relaxations need to be made to real-world data. GAs are employed for large-scale maintenance scheduling [96, 97, 98], which are capable of taking into account parameters like CapEx and OpEx costs, while also factoring elements such as legal authorizations, surveys, engineering, installation, insurance, clearance, etc. ANNs are also used for maintenance optimization by taking factors such as the remaining lifetime of WTs [99, 100], the inspection time window, and various costs.
In addition to the requirements for maintenance, the op- portunity of maintenance needs to be accounted for as well. If relevant parties are not able to access the WTs due to meteorological phenomena, then even a well-crafted maintenance plan cannot be followed. Thus, studies account for factors such as wind speed for onshore wind energy systems [101, 102], and wind speeds and wave heights for offshore wind energy systems [103, 104]. These studies evaluate the type, in- tensity, and duration of such phenomenon against safe work- ing rules and regulations for maintenance activities such as crane usage.
What are the ethical-oriented implications of using AI?
Despite the significant cost and performance benefits that AI models offer, they are still lacking in terms of their explainability, interpretability, transparency, and trustworthiness. Many AI models such as ANNs, GAs, and PSOs operate as black boxes and provide their results without providing any explicit underlying reasoning. For instance, if a model is unable to forecast extreme weather events and schedules maintenance activities during these events, it can lead to un-
precedented safety hazards and ethical concerns relating to the health and safety of workers. Similarly, the failure of predicting catastrophic faults and reporting the normal functioning of WT hardware can also lead to operational hazards and safety concerns. Thus, despite the performance benefits that AI offers, there is still a significant amount of human oversight required in such cases.
These algorithms are also susceptible to bias. If the data being used to train these models is biased or unrepresentative, the models can learn unexpected and undesirable patterns, which can lead to biased results. Again, due to the
lack of explainability and interpretability, it can be difficult to identify and understand such characteristics.
Data security is another concern as AI models require large amounts of data. As discussed in this blog, there are lots of parameters that need to be considered for forecasting and optimization in the WEG domain, such as meteorological data, ecological data, WT component configuration and maintenance logs, operational requirements, and community engagements. Thus, it is important to ensure sufficient security and privacy for sensitive data.
WEG projects can cause disruptions to the surrounding areas in the form of noise pollution, wind turbulence, and construction and operational activities. Thus, it is important to take the surrounding wildlife and local community engagement into consideration while developing WFs [52]. If a model disregards these factors, it can lead to ethical violations relating to community and wildlife rights and strained trustworthiness. Even though numerous datasets, biological assessments, and research can provide insights into the impact on wildlife [53, 54, 56, 105], the lack of explainability from AI models can still lead to undesirable results and detrimental ecological consequences. Even when such data is accounted for, it needs to be comprehensive, complete, and reliable in order for model outputs to be trustworthy.
AI models require high computing hardware and infratructure. The AI model development pipeline can produce large amounts of carbon emissions [106]. Even though there is a global attempt to reduce the environmental impact of AI models, currently it cannot go unnoticed. WEG projects need to consider the negative environmental impact of implementing AI models in WFs.
Adoption of AI in wind energy systems shows promise
AI models have demonstrated their ability to greatly enhance the performance and optimize various aspects of WEG systems. The flexibility and scalability of these models can improve WF performance, forecast WEG outputs, determine sites and layouts for WF development, optimize grid integration and ESS management, predict and detect faults, and optimize maintenance schedules. This provides meaningful value to WEG projects. Considering the emerging trend of REG and increasing demand for clean energy, the automation, and optimization of key processes can accelerate the growth of REG technologies.
However, it is important to acknowledge the ethical impact of implementing such models. The black-box nature of various AI models results in lacking explainability, interpretability, transparency, and accountability. Models pose another limitation in terms of the quantity and quality of data required for training. Without fair, unbiased, complete, and comprehensive datasets, it is possible for models to develop biases and disregard wider objectives that do not contribute towards maximizing WEG outputs. Thus, a constant collaboration between interdisciplinary experts is required to improve the quality of such models and to address these concerns.
AI models have the power to revolutionalize the wind energy industry. Both the usage of wind energy and the academic research of applications of AI in wind energy have experienced an exponential rise over the past few years. By harnessing the power of AI responsibly and ethically, we can unlock the full potential of wind energy and contribute to a cleaner and more sustainable energy ecosystem.
References
[1] Muhammad Shahbaz, Chandrashekar Raghutla, Krishna Reddy Chittedi, Zhilun Jiao, and Xuan Vinh Vo, “The effect of renewable energy consumption on economic growth: Evidence from the renewable energy country attractive index,” Energy, vol. 207, pp. 118162, 2020.
[2] Abdul Ghani Olabi, Tabbi Wilberforce, and Mohammad Ali Abdelkareem, “Fuel cell application in the automotive industry and future perspective,” Energy, vol. 214, pp. 118955, 2021.
[3] Global Renewables Outlook IRENA, “Energy transformation 2050,” International Renewable Energy Agency, Abu Dhabi, 2020.
[4] “Modern renewable energy generation by source — ourworldin-data.org,” https://ourworldindata.org/grapher/modern-renewable-prod, [Accessed 29-May-2023].
[5] “Per capita electricity generation from wind — ourworldin-data.org,” https://ourworldindata.org/grapher/wind-electricity-per-capita, [Accessed 30-May- 2023].
[6] “Wind power generation — ourworldindata.org,” https://ourworldindata.org/grapher/wind-generation, [Accessed 29-May-2023].
[7] “Investment in renewable energy, by technology — ourworldin- data.org,” https://ourworldindata.org/grapher/investment-in-renewable-energy-by-technology, [Accessed 29-May-2023].
[8] “Annual patents filed for renewable energy technologies —ourworldindata.org,” https://ourworldindata.org/grapher/patents-filed-for-renewables, [Accessed 29-May-2023].
[9] “Levelized cost of energy by technology — ourworldin-data.org,” https://ourworldindata.org/grapher/levelized-cost-of-energy, [Accessed 29-May-2023].
[10] “Wind energy in the UK – Office for National Statistics — ons.gov.uk,” https://www.ons.gov.uk/economy/environmentalaccounts/articles/windenergyintheuk/june2021#why-wind-energy-is-important, [Accessed 29-May-2023].
[11] REN21, “RENEWABLES 2017 GLOBAL STATUS REPORT — ren21.net,” https://www.ren21.net/gsr-2017/, [Accessed 29-May-2023].
[12] “Renewable Purchase Obligations (RPO) — Anert — anert.gov.in,” http://www.anert.gov.in/node/114, [Accessed 29-May-2023].
[13] “Renewable energy policies in a time of transition — irena.org,” https://www.irena.org/publications/2018/apr/renewable-energy-policies-in-a-time-of-transition, [Accessed 29-May-2023].
[14] Choong-Ki Kim, Seonju Jang, and Tae Yun Kim, “Site selection for offshore wind farms in the southwest coast of south korea,” Renewable energy, vol. 120, pp. 151–162, 2018.
[15] M Madhiarasan and S Deepa, “Application of ensemble neural networks for different time scale wind speed prediction,” neural networks, vol. 4, no. 5, 2016.
[16] Ramteen Sioshansi and Paul Denholm, “Benefits of colocating concentrating solar power and wind,” IEEE Transactions on Sustainable Energy, vol. 4, no. 4, pp. 877–885, 2013.
[17] Shakir D. Ahmed, Fahad S. M. Al-Ismail, Md Shafiullah, Fahad A. Al-Sulaiman, and Ibrahim M. El-Amin, “Grid integration challenges of wind energy: A review,” IEEE Access, vol. 8, pp. 10857–10878, 2020.
[18] TW Verbruggen, LWMM Rademakers, EJ Wiggelinkhuizen, SJ Watson, J Xiang, G Giebel, E Norton, MC Tipluica, AJ Christensen, and E Becker, “Conmow final report,” 2007.
[19] Yiling Cai and Francois-Marie Breon, “Wind power potential and intermittency issues in the context of climate change,” Energy Con- version and Management, vol. 240, pp. 114276, 2021.
[20] GB Ezhiljenekkha and M MarsalineBeno, “Review of power quality issues in solar and wind energy,” Materials Today: Proceedings, vol. 24, pp. 2137–2143, 2020.
[21] Yirui Wang, Yang Yu, Shuyang Cao, Xingyi Zhang, and Shangce Gao, “A review of applications of artificial intelligent algorithms in wind farms,” Artificial Intelligence Review, vol. 53, pp. 3447–3500, 2020.
[22] Ted KA Brekken, Alex Yokochi, Annette Von Jouanne, Zuan Z Yen, Hannes Max Hapke, and Douglas A Halamay, “Optimal energy storage sizing and control for wind power applications,” IEEE Transactions on Sustainable Energy, vol. 2, no. 1, pp. 69–77, 2010.
[23] Fausto Pedro Garcia Marquez and Isidro Pe ̃na Garc ́ıa-Pardo, “Principal component analysis applied to filtered signals for maintenance management,” Quality and Reliability Engineering International, vol. 26, no. 6, pp. 523–527, 2010.
[24] SA Pourmousavi Kani and MM Ardehali, “Very short-term wind speed prediction: A new artificial neural network–markov chain model,” Energy Conversion and Management, vol. 52, no. 1, pp. 738–745, 2011.
[25] Shuang Gao, Lei Dong, Xiaozhong Liao, and Yang Gao, “Very-short-term prediction of wind speed based on chaos phase space reconstruction and nwp,” in Proceedings of the 32nd Chinese Control Conference. IEEE, 2013, pp. 8863–8867.
[26] E Safavieh, A Jahanbani Ardakani, A Kashefi Laviani, SA Pour-mousavi, SH Hosseinian, and M Abedi, “A new integrated approach for very short-term wind speed prediction using wavelet networks and pso,” in Proceedings of the International Conference on Power Systems, 2007.
[27] Kostas Philippopoulos and Despina Deligiorgi, “Application of artificial neural networks for the spatial estimation of wind speed in a coastal region with complex topography,” Renewable Energy, vol. 38, no. 1, pp. 75–82, 2012.
[28] Jos ́e Carlos Palomares-Salas, Agust ́ın Ag ̈uera-P ́erez, Juan Jos ́e Gonz ́alez de la Rosa, and Antonio Moreno-Mu ̃noz, “A novel neural network method for wind speed forecasting using exogenous measurements from agriculture stations,” Measurement, vol. 55, pp. 295–304, 2014.
[29] Gong Li, Jing Shi, and Junyi Zhou, “Bayesian adaptive combination of short-term wind speed forecasts from neural network models,” Renewable Energy, vol. 36, no. 1, pp. 352–359, 2011.
[30] Younes Noorollahi, Mohammad Ali Jokar, and Ahmad Kalhor, “Using artificial neural networks for temporal and spatial wind speed forecasting in iran,” Energy Conversion and Management, vol. 115, pp. 17–25, 2016.
[31] Stefan Balluff, J ̈org Bendfeld, and Stefan Krauter, “Short term wind and energy prediction for offshore wind farms using neural networks,” in 2015 International Conference on Renewable Energy Research and Applications (ICRERA). IEEE, 2015, pp. 379–382.
[32] Anbo Meng, Jiafei Ge, Hao Yin, and Sizhe Chen, “Wind speed forecasting based on wavelet packet decomposition and artificial neural networks trained by crisscross optimization algorithm,” Energy Con- version and Management, vol. 114, pp. 75–88, 2016.
[33] Boubacar Doucoure, Kodjo Agbossou, and Alben Cardenas, “Time series prediction using artificial wavelet neural network and multi- resolution analysis: Application to wind speed data,” Renewable Energy, vol. 92, pp. 202–211, 2016.
[34] Ronay Ak, Olga Fink, and Enrico Zio, “Two machine learning approaches for short-term wind speed time-series prediction,” IEEE transactions on neural networks and learning systems, vol. 27, no. 8, pp. 1734–1747, 2015.
[35] Jianzhou Wang, Shanshan Qin, Qingping Zhou, and Haiyan Jiang, “Medium-term wind speeds forecasting utilizing hybrid models for three different sites in xinjiang, china,” Renewable Energy, vol. 76, pp. 91–101, 2015.
[36] Alberto Pliego Marug ́an, Fausto Pedro Garc ́ıa M ́arquez, Jesus Mar ́ıa Pinar Perez, and Diego Ruiz-Hern ́andez, “A survey of artificial neural network in wind energy systems,” Applied energy, vol. 228, pp. 1822–1836, 2018.
[37] MA Ansari, Nidhi Singh Pal, Hasmat Malik, et al., “Wind speed and power prediction of prominent wind power potential states in India using grnn,” in 2016 IEEE 1st International Conference on Power Electronics, Intelligent Control and Energy Systems (ICPE-ICES). IEEE, 2016, pp. 1–6.
[38] KP Moustris, D Zafirakis, DH Alamo, RJ Nebot Medina, and JK Kaldellis, “24-h ahead wind speed prediction for the optimum operation of hybrid power stations with the use of artificial neural networks,” in Perspectives on Atmospheric Sciences. Springer, 2017, pp. 409–414.
[39] Hanieh Borhan Azad, Saad Mekhilef, and Vellapa Gounder Ganapathy, “Long-term wind speed forecasting and general pattern recognition using neural networks,” IEEE Transactions on Sustainable Energy, vol. 5, no. 2, pp. 546–553, 2014.
[40] Niels Otto Jensen, A note on wind generator interaction, vol. 2411, Citeseer, 1983.
[41] Atsushi Yamaguchi and Takeshi Ishihara, “Assessment of offshore wind energy potential using mesoscale model and geographic information system,” Renewable Energy, vol. 69, pp. 506–515, 2014.
[42] Fausto Pedro Garc ́ıa M ́arquez and Alfredo Peinado Gonzalo, “A comprehensive review of artificial intelligence and wind energy,” Archives of Computational Methods in Engineering, pp. 1–24, 2021.
[43] Adel El Shahat, Rami J Haddad, and Youakim Kalaani, “An artificial neural network model for wind energy estimation,” in SoutheastCon 2015. IEEE, 2015, pp. 1–2.
[44] GPCDB Mosetti, Carlo Poloni, and Bruno Diviacco, “Optimization of wind turbine positioning in large windfarms by means of a genetic algorithm,” Journal of Wind Engineering and Industrial Aerodynamics, vol. 51, no. 1, pp. 105–116, 1994.
[45] SA Grady, MY Hussaini, and Makola M Abdullah, “Placement of wind turbines using genetic algorithms,” Renewable energy, vol. 30, no. 2, pp. 259–270, 2005.
[46] Javier Serrano Gonz ́alez, Angel G Gonzalez Rodriguez, Jos ́e Castro Mora, Jes ́us Riquelme Santos, and Manuel Burgos Payan, “Optimization of wind farm turbines layout using an evolutive algorithm,” Renewable energy, vol. 35, no. 8, pp. 1671–1681, 2010.
[47] Javier Serrano Gonzalez, Manuel Burgos Payan, and Jes ́us M Riquelme-Santos, “Optimization of wind farm turbine layout including decision making under risk,” IEEE Systems Journal, vol. 6, no. 1, pp. 94–102, 2011.
[48] J Serrano Gonz ́alez, ́AG Gonz ́alez Rodr ́ıguez, J Castro Mora, M Burgos Pay ́an, and J Riquelme Santos, “Overall design optimization of wind farms,” Renewable Energy, vol. 36, no. 7, pp. 1973–1982, 2011.
[49] Javier Serrano Gonz ́alez, Manuel Burgos Pay ́an, and Jesus Riquelme Santos, “A new and efficient method for optimal design of large offshore wind power plants,” IEEE Transactions on Power Systems, vol. 28, no. 3, pp. 3075–3084, 2013.
[50] Lambros Ekonomou, Stavros Lazarou, George E Chatzarakis, and Vasiliki Vita, “Estimation of wind turbines optimal number and produced power in a wind farm using an artificial neural network model,” Simulation Modelling Practice and Theory, vol. 21, no. 1, pp. 21–25, 2012.
[51] “New York Bight — Bureau of Ocean Energy Management — boem.gov,” https://www.boem.gov/renewable-energy/state-activities/new-york-bight, [Accessed 30-May-2023].
[52] “Environmental Consultations and Offshore Renewable Energy — Bureau of Ocean Energy Management — boem.gov,” https://www.boem.gov/environmental-consultations, [Accessed 29-May-2023].
[53] Robert E Blyth-Skyrme, “Options and opportunities for marine fisheries mitigation associated with windfarms,” Final report for Collaborative Offshore Wind Research into the Environment contract FISHMITIG09. COWRIE Ltd, London, 2010.
[54] Walter Musial and Bonnie Ram, “Large-scale offshore wind power in the united states: Assessment of opportunities and barriers,” Tech. Rep., National Renewable Energy Lab.(NREL), Golden, CO (United States), 2010.
[55] “Regulatory Framework and Guidelines — Bureau of Ocean Energy Management — boem.gov,” https://www.boem.gov/environment/regulatory-framework-and-guidelines, [Accessed 30-May-2023].
[56] “Important Bird Areas — ny.audubon.org,” https://ny.audubon.org/conservation/important-bird-areas, [Accessed 29-May-2023].
[57] Rob Van Haaren and Vasilis Fthenakis, “Gis-based wind farm site selection using spatial multi-criteria analysis (smca): Evaluating the case for new york state,” Renewable and sustainable energy reviews, vol. 15, no. 7, pp. 3332–3340, 2011.
[58] Addisu D Mekonnen and Pece V Gorsevski, “A web-based participatory gis (pgis) for offshore wind farm suitability within lake erie, ohio,” Renewable and Sustainable Energy Reviews, vol. 41, pp. 162–177, 2015.
[59] L Cradden, Christina Kalogeri, I Martinez Barrios, George Galanis, David Ingram, and George Kallos, “Multi-criteria site selection for offshore renewable energy platforms,” Renewable energy, vol. 87, pp. 791–806, 2016.
[60] G Benassai, Patrizio Mariani, Claus Stenberg, and Mads Christoffersen, “A sustainability index of potential co-location of offshore wind farms and open water aquaculture,” Ocean & coastal management, vol. 95, pp. 213–218, 2014.
[61] Ahmed N Abdalla, Muhammad Shahzad Nazir, Hai Tao, Suqun Cao, Rendong Ji, Mingxin Jiang, and Liu Yao, “Integration of energy storage system and renewable energy sources based on artificial intelligence: An overview,” Journal of Energy Storage, vol. 40, pp. 102811, 2021.
[62] Soedibyo Soedibyo, Heri Suryoatmojo, Imam Robandi, and Mochamad Ashari, “Optimal design of fuel-cell, wind and micro-hydro hybrid system using genetic algorithm,” TELKOMNIKA (Telecommunication Computing Electronics and Control), vol. 10, no. 4, pp. 695–702, 2012.
[63] Mohammad Reza Ranjbar and Sajjad Kouhi, “Sources’ response for supplying energy of a residential load in the form of on-grid hybrid systems,” International Journal of Electrical Power & Energy Systems, vol. 64, pp. 635–645, 2015.
[64] AH Shahirinia, SMM Tafreshi, A Hajizadeh Gastaj, and AR Moghaddomjoo, “Optimal sizing of hybrid power system using genetic algorithm,” in 2005 International Conference on Future Power Systems. IEEE, 2005, pp. 6–pp.
[65] Ajay Kumar Bansal, RA Gupta, and Rajesh Kumar, “Optimization of hybrid pv/wind energy system using meta particle swarm optimization (mpso),” in India International Conference on Power Electronics 2010 (IICPE2010). IEEE, 2011, pp. 1–7.
[66] Akbar Maleki, Hamed Hafeznia, Marc A Rosen, and Fathollah Pourfayaz, “Optimization of a grid-connected hybrid solar-wind-hydrogen chp system for residential applications by efficient metaheuristic approaches,” Applied Thermal Engineering, vol. 123, pp. 1263–1277, 2017.
[67] Ali Kashefi Kaviani, Hamid Reza Baghaee, and Gholam Hossein Riahy, “Optimal sizing of a stand-alone wind/photovoltaic generation unit using particle swarm optimization,” Simulation, vol. 85, no. 2, pp. 89–99, 2009.
[68] AG Olabi, C Onumaegbu, Tabbi Wilberforce, Mohamad Ramadan, Mohammad Ali Abdelkareem, and Abdul Hai Al-Alami, “Critical review of energy storage systems,” Energy, vol. 214, pp. 118987, 2021.
[69] Ling Ai Wong, Vigna K Ramachandaramurthy, Phil Taylor, JB Ekanayake, Sara L Walker, and Sanjeevikumar Padmanaban, “Review on the optimal placement, sizing and control of an energy storage system in the distribution network,” Journal of Energy Storage, vol. 21, pp. 489–504, 2019.
[70] Francisco D ́ıaz-Gonz ́alez, Andreas Sumper, Oriol Gomis-Bellmunt, and Roberto Villaf ́afila-Robles, “A review of energy storage technologies for wind power applications,” Renewable and sustainable energy reviews, vol. 16, no. 4, pp. 2154–2171, 2012.
[71] Haoran Zhao, Qiuwei Wu, Shuju Hu, Honghua Xu, and Claus Nygaard Rasmussen, “Review of energy storage system for wind power integration support,” Applied energy, vol. 137, pp. 545–553, 2015.
[72] AB Gallo, JR Sim ̃oes-Moreira, HKM Costa, MM Santos, and E Moutinho Dos Santos, “Energy storage in the energy transition context: A technology review,” Renewable and sustainable energy reviews, vol. 65, pp. 800–822, 2016.
[73] Pablo Garc ́ıa-Trivi ̃no, Luis M Fern ́andez-Ram ́ırez, Antonio J GilMena, Francisco Llorens-Iborra, Carlos Andr ́es Garc ́ıa-V ́azquez, and Francisco Jurado, “Optimized operation combining costs, efficiency and lifetime of a hybrid renewable energy system with energy storage by battery and hydrogen in grid-connected applications,” International Journal of Hydrogen Energy, vol. 41, no. 48, pp. 23132– 23144, 2016.
[74] Su Guo, Yi He, Huanjin Pei, and Shuyan Wu, “The multi-objective capacity optimization of wind-photovoltaic-thermal energy storage hybrid power system with electric heater,” Solar Energy, vol. 195, pp. 138–149, 2020.
[75] Tingli Cheng, Minyou Chen, Yingxiang Wang, Bo Li, Muhammad Arshad Shehzad Hassan, Tao Chen, and Ruilin Xu, “Adaptive robust method for dynamic economic emission dispatch incorporating renewable energy and energy storage,” Complexity, vol. 2018, 2018.
[76] H Nasiraghdam and SJSE Jadid, “Optimal hybrid pv/wt/fc sizing and distribution system reconfiguration using multi-objective artificial bee colony (moabc) algorithm,” Solar Energy, vol. 86, no. 10, pp. 3057–3071, 2012.
[77] Junyi Shen and Alireza Khaligh, “A supervisory energy management control strategy in a battery/ultracapacitor hybrid energy storage system,” IEEE Transactions on transportation electrification, vol. 1, no. 3, pp. 223–231, 2015.
[78] MH Athari and MM Ardehali, “Operational performance of energy storage as function of electricity prices for on-grid hybrid renewable energy system by optimized fuzzy logic controller,” Renewable Energy, vol. 85, pp. 890–902, 2016.
[79] Shu Wang, Yuejin Tang, Jing Shi, Kang Gong, Yang Liu, Li Ren, and Jingdong Li, “Design and advanced control strategies of a hybrid energy storage system for the grid integration of wind power generations,” IET Renewable Power Generation, vol. 9, no. 2, pp. 89–98, 2015.
[80] Ling Xiang, Xin Yang, Aijun Hu, Hao Su, and Penghe Wang, “Condition monitoring and anomaly detection of wind turbine based on cascaded and bidirectional deep learning networks,” Applied Energy, vol. 305, pp. 117925, 2022.
[81] Mahendra Singh Raghav and Ram Bihari Sharma, “A review on fault diagnosis and condition monitoring of gearboxes by using ae technique,” Archives of Computational Methods in Engineering, vol. 28, no. 4, pp. 2845–2859, 2021.
[82] “Wind turbine — technology — britannica.com,” https://www.britannica.com/technology/wind-turbine, [Accessed 30-May-2023].
[83] Yue Cui, Pramod Bangalore, and Lina Bertling Tjernberg, “An anomaly detection approach based on machine learning and scada data for condition monitoring of wind turbines,” in 2018 IEEE International Conference on Probabilistic Methods Applied to Power Systems (PMAPS). IEEE, 2018, pp. 1–6.
[84] RF Mesquita Brand ̃ao, JA Beleza Carvalho, and FP Maciel Barbosa, “Intelligent system for fault detection in wind turbines gearbox,” in 2015 IEEE Eindhoven PowerTech. IEEE, 2015, pp. 1–6.
[85] RF Mesquita Brand ̃ao, JA Beleza Carvalho, and FP MacIel Barbosa, “Forecast of faults in a wind turbine gearbox,” in 2012 ELEKTRO. IEEE, 2012, pp. 170–173.
[86] Zhen-You Zhang and Ke-Sheng Wang, “Wind turbine fault detection based on scada data analysis using ann,” Advances in Manufacturing, vol. 2, pp. 70–78, 2014.
[87] RF Mesquita Brand ̃ao, Assinatura Digital de Aerogeradores, Ph.D. thesis, PhD dissertation, Faculty of Engineering-University of Porto, Porto 2011 (in . . . , 2012.
[88] Andrew Kusiak and Anoop Verma, “Analyzing bearing faults in wind turbines: A data-mining approach,” Renewable Energy, vol. 48, pp. 110–116, 2012.
[89] Adel Aloraini and Moamar Sayed-Mouchaweh, “Graphical model based approach for fault diagnosis of wind turbines,” in 2014 13th International Conference on Machine Learning and Applications. IEEE, 2014, pp. 614–619.
[90] Pramod Bangalore and Lina Bertling Tjernberg, “An artificial neural network approach for early fault detection of gearbox bearings,” IEEE Transactions on Smart Grid, vol. 6, no. 2, pp. 980–987, 2015.
[91] Alberto Pliego Marug ́an and Fausto Pedro Garc ́ıa M ́arquez, “Scada and artificial neural networks for maintenance management,” in Proceedings of the Eleventh International Conference on Management Science and Engineering Management 11. Springer, 2018, pp. 912–919.
[92] Meik Schlechtingen, Ilmar Ferreira Santos, and Sofiane Achiche, “Wind turbine condition monitoring based on scada data using normal behavior models. part 1: System description,” Applied Soft Computing, vol. 13, no. 1, pp. 259–270, 2013.
[93] Chandra Ade Irawan, Djamila Ouelhadj, Dylan Jones, Magnus St ̊alhane, and Iver Bakken Sperstad, “Optimisation of maintenance routing and scheduling for offshore wind farms,” European Journal of Operational Research, vol. 256, no. 1, pp. 76–89, 2017.
[94] Shuya Zhong, Athanasios A Pantelous, Michael Beer, and Jian Zhou, “Constrained non-linear multi-objective optimisation of preventive maintenance scheduling for offshore wind farms,” Mechanical Systems and Signal Processing, vol. 104, pp. 347–369, 2018.
[95] Shuya Zhong, Athanasios A Pantelous, Mark Goh, and Jian Zhou, “A reliability-and-cost-based fuzzy approach to optimize preventive maintenance scheduling for offshore wind farms,” Mechanical Systems and Signal Processing, vol. 124, pp. 643–663, 2019.
[96] Libo Liu, Yifan Zhou, Bin Yan, and Jingjing Liu, “Large-scale maintenance scheduling of wind turbines,” in 2019 Prognostics and System Health Management Conference (PHM-Qingdao). IEEE, 2019, pp. 1–5.
[97] Anastasia Ioannou, Lin Wang, and Feargal Brennan, “Design implications towards inspection reduction of large scale structures,” Procedia CIRP, vol. 60, pp. 434–439, 2017.
[98] G Srinivasulu, B Subramanyam, and M Surya Kalavathi, “Market based transmission expansion planning for tamilnadu test system,” Int J Simul Syst Sci Technol, vol. 17, pp. 32, 2016.
[99] Juan Izquierdo, A Crespo M ́arquez, Jone Uribetxebarria, and Asier Erguido, “On the importance of assessing the operational context impact on maintenance management for life cycle cost of wind energy projects,” Renewable Energy, vol. 153, pp. 1100–1110, 2020.
[100] Yang Lu, Liping Sun, Xinyue Zhang, Feng Feng, Jichuan Kang, and Guoqiang Fu, “Condition based maintenance optimization for offshore wind turbine considering opportunities based on neural network approach,” Applied Ocean Research, vol. 74, pp. 69–79, 2018.
[101] Nurseda Y Y ̈ur ̈us ̧en, Paul N Rowley, Simon J Watson, and Julio J Melero, “Automated wind turbine maintenance scheduling,” Reliability Engineering & System Safety, vol. 200, pp. 106965, 2020.
[102] Sofia Carlos, Ana S ́anchez, Sebastian Martorell, and Isabel Mart ́on, “Onshore wind farms maintenance optimization using a stochastic model,” Mathematical and Computer Modelling, vol. 57, no. 7-8, pp. 1884–1890, 2013.
[103] Jichuan Kang, Jose Sobral, and C Guedes Soares, “Review of condition-based maintenance strategies for offshore wind energy,” Journal of Marine Science and Application, vol. 18, pp. 1–16, 2019.
[104] Tomas Gintautas and John Dalsgaard Sørensen, “Improved methodology of weather window prediction for offshore operations based on probabilities of operation failure,” Journal of Marine Science and Engineering, vol. 5, no. 2, pp. 20, 2017.
[105] “NMFS ESA Consultations — Bureau of Ocean Energy Management — boem.gov,” https://www.boem.gov/renewable-energy/state-activities/nmfs-esa-consultations, [Accessed 29-May-2023].
[106] Payal Dhar, “The carbon impact of artificial intelligence.,” Nat. Mach. Intell., vol. 2, no. 8, pp. 423–425, 2020.
Emerging Topics in Integrated Machine Learning Systems 