Next Article in Journal
Solvothermal Liquefaction of Blackcurrant Pomace in the Water-Monohydroxy Alcohol Solvent System
Previous Article in Journal
Introducing a Novel Rice Husk Combustion Technology for Maximizing Energy and Amorphous Silica Production Using a Prototype Hybrid Rice Husk Burner to Minimize Environmental Impacts and Health Risk
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Energies
Volume 16
Issue 3
10.3390/en16031125
energies-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Review

A Review of Fault Diagnosis, Status Prediction, and Evaluation Technology for Wind Turbines

by
Fanghong Zhang
1,
Mingsong Chen
1,
Yuze Zhu
1,
Kai Zhang
2,* and
Qingan Li
3,*
1
The National Center for Applied Mathematics in Chongqing, Chongqing Normal University, Chongqing 401331, China
2
CSIC Haizhuang Windpower Co., Ltd., Chongqing 401122, China
3
University of Chinese Academy of Sciences, Beijing 100049, China
*
Authors to whom correspondence should be addressed.
Energies 2023, 16(3), 1125; https://doi.org/10.3390/en16031125
Submission received: 25 November 2022 / Revised: 27 December 2022 / Accepted: 28 December 2022 / Published: 19 January 2023
Browse Figures
Versions Notes

Abstract

:
With the rapid development and increasing energy production capacity of high-power wind turbines, a corresponding increase in maintenance requirements has been observed. Reducing the failure rate of wind turbines is a critical objective, alongside decreasing affiliated operation and maintenance costs. This review focuses on the status monitoring, fault diagnosis, fault prediction, and status evaluation of wind turbines. The early fault diagnosis of wind turbines is explored with regard to existing condition monitoring technology. Moreover, the current mathematics-based fault diagnosis and smart fault diagnosis technologies are further explored. Through comprehensive investigation, this paper summarizes the research status of wind turbine fault prediction and complete machine status evaluation, conclusively presenting relevant research points and trends in the fault diagnosis, fault prediction, and status assessment of high-power wind turbines.

1. Preface

With the improvement in the living standard of human society, the demand for energy has also accordingly increased. According to the 2022 “BP World Energy Statistical Yearbook”, total global energy demand grew by 5.8% in 2021, recorded as the largest single-year increase observed in history. Consequently, the overexploitation of fossil fuel energy sources, such as coal, oil, and natural gas, has resulted in a universal exacerbation of energy crises, most prominently observed through the negative effects of air pollution [1]. Furthermore, as the primary source of power, the functional use of fossil fuel energy cannot be easily replaced in totality. However, the constant advancement in sustainable energy practices has provided several alternatives that merit consideration. China has proposed the ambitious goal of striving to achieve a carbon peak in 2030 and carbon neutrality in 2060 [2].
In view of these current circumstances, low-carbon clean energy sources such as wind, solar, and thermal energy have been widely instituted and are highly valued. Figure 1 illustrates the various renewable energy sources and their proportional contribution to electricity production. As shown, wind power accounts for a quarter of the total [3]. Wind energy has achieved sustained and rapid development in countries all over the world due to its short construction period, low environmental requirements, abundant energy sources, and high efficiency [4,5]. According to statistical data provided by the Global Wind Energy Council (GWEC) in 2021, the global wind market is set to add 557 GW of new capacity in the next five years, with more than 110 GW annually added by 2026, as shown in Figure 2. Wind power has ushered in a rapidly developing market opportunity, but it also brings with it several challenges. The constructive concerns of wind turbines are numerous, and practical difficulties are most commonly associated with design and geographical installation. Wind turbines are usually located in remote places with inconvenient transportation, and the infrastructure of the cabin is frequently located hundreds of meters above the ground, resulting in several maintenance difficulties. Furthermore, exposure to meteorological conditions is of significant consideration. Wind turbines are often exposed to wind, frost, snow, and heavy rain [6,7], as well as numerous other environmental factors. Moreover, offshore wind turbines usually face an even harsher operating environment than onshore wind turbines. Statistical evidence has demonstrated that maintenance costs of offshore wind turbines are at least twice that of their onshore counterparts, with operation and maintenance expenses totaling an estimated 30~35% of the gross income produced. An estimated 25~35% of recorded maintenance costs have been reported as regular maintenance costs, with the remaining 65~75% accounting for post-event maintenance expenses [8]. According to the General Electric Energy report for offshore wind turbines, a bearing replacement that usually only costs 5000 USD can easily turn into a project that costs 25,000 USD [9,10].
For these reasons, industry consensus is directed at facilitating reduced operating costs through actively avoiding excessive maintenance and corresponding downtime resultant from mechanical failure or fault, effectively improving economic benefits and financial gains [11,12,13,14,15,16]. Therefore, it is necessary to understand the current research status of fan units using fault diagnosis and fault prediction to summarize the research methods and achievements in this field, as well as to promote further research. The framework for this review is shown in Figure 3.

2. Fault Diagnosis of Early Condition Monitoring

The performance of large units is usually diminished after an extended period of time as a result of operating in harsh environments [17]. Traditional equipment maintenance strategies include post-event maintenance and regular maintenance services, to which end the combination of these two service approaches can potentially lead to either insufficient maintenance or over-maintenance [8]. Real-time status monitoring of wind turbines can help to effectively overcome these shortcomings. Condition monitoring technology for wind turbines refers to the monitoring of vibration, temperature, pressure, and electrical parameters. By comparing these results with predetermined optimal values, early mechanical and electrical failures of wind turbines can be readily detected. The obtained data is then collected, analyzed, and processed, in order to accurately determine the factors causing the equipment failure. For wind turbines, condition monitoring technology can be divided into the following categories [18,19,20,21].

2.1. Vibration Monitoring

Vibration analysis is the most widely used method for analyzing the various environmental effects observed on rotating machinery. Vibration analysis can be applied to observe accurate equipment inspection and fault diagnosis of a majority of mechanical equipment, such as rotor imbalance, shaft bending, bearing looseness, shaft misalignment, etc. In wind turbines, the vibration monitoring method is primarily used to monitor the gears and bearings of the gearbox, the bearings of the generator, the main bearing, the vibration of the nacelle, and, in some cases, the blades. Conventional vibration analysis methods include power spectral density analysis, frequency domain amplitude spectrum analysis, cepstrum analysis, etc. Zheng et al. [22] used cepstrum analysis of vibration signals to monitor the state of rotating machinery. Peeters et al. [23] applied the cepstrum editing technology of vibration signals to determine fault detection of bearings. Abboud et al. [24] compared the envelope spectrum of the vibration signals of rotating machinery under various wind speeds and used them for fault diagnosis of wind turbines.

2.2. Electric Signal Monitoring

Based on several statistical analyses assessing all the faults of generators, the fault rate of bearings is reported to be 40%, with that of the stator being 38%, and that of the rotor being 10%, with the remaining faults accounting for 12% [25]. According to the fault characteristics of the generator, motor operational characteristics are determined by the transformation of electrical signals, such as the stator and rotor current and power. However, when compared with vibrational signals, fault-related signals contained in electrical signals (such as current signals) are often relatively weak, and it is difficult to determine specific fault features. Dang et al. [26] used the impeller mass imbalance of wind turbines as an object, analyzed the transformation characteristics of the electrical signal of the generator under the fault mechanism, and conducted correlation analysis based on Hilbert transformation and the variational mode decomposition (CA-VMD) electrical signal feature extraction method. Furthermore, the fault frequency was successfully extracted and the data provided novel research avenues for investigating the early imbalance fault of the wind turbine blades.

2.3. Temperature Monitoring

Temperature monitoring is commonly used in the fault diagnostic process applied to various electronics and electronic components. In the case of equipment deterioration, temperature has been identified as a crucial factor that can significantly affect the operational state of the equipment. In wind turbines, temperature sensors are installed in gearboxes, generators, frequency converters, and other components. Huo et al. [27] used the temperature data of the generator under normal operating conditions, as well as related linear regression technology, to obtain the corrected temperature after eliminating the influence of ambient temperature and output power. Furthermore, the real-time reliability monitoring model of the generator operating temperature was established. This evidence maintains that various abnormal states of the generator can be determined using the example data of the unit.

2.4. Oil Monitoring

Oil monitoring technology is used to assess the performance of lubricating oil and hydraulic oil in the monitored equipment, while simultaneously allowing for the effects of abrasive particles in the oil to be detected, which enables comprehensive understanding regarding the effects of lubrication and wear of parts during equipment operation [20]. In wind turbine condition monitoring, the main purpose of oil analysis is to monitor oil quality and component quality. In a previous study [28], the oil in a gear box on a wind farm was analyzed, and an abnormal increase in iron elements and PQ in the gear oil was observed. It was then determined that there was evident pitting on the tooth surface of the main gearbox, whereby specific data, which was visually confirmed by an endoscope, revealed that the tooth surface was corroded and partially scratched.

3. Current Fault Diagnosis

3.1. Fault Diagnosis Based on the Mathematical Method

In the fault diagnosis of complex wind turbines, one faulty state may correspond to multiple fault causes, and one fault cause may also reflect multiple faulty states to varying degrees. Traditional fault diagnosis technology is no longer sufficient to meet the data requirements created by new technological developments. It is difficult to use mathematical models to describe fault causes and faulty states of the equipment, with the related rules of the fault diagnosis mechanism often being unclear. The most commonly used mathematical models applied to this type of fault analysis include fuzzy logic diagnosis, gray theory, fuzzy petri net, and several others.

3.1.1. Fuzzy Logic Diagnosis

Fuzzy logic diagnosis technology was developed based on the fuzzy set theory founded by the American scientist Professor L. A. Zadeh in the 1960s. It uses fuzzy logic to describe the fuzzy relationship between the cause of the fault and the fault phenomenon. It also aims to determine specific fault origins and detailed status identification by using the membership function and fuzzy relational equations [29,30,31]. The principle of fuzzy logic diagnosis is shown in Figure 4; wherein R represents the fuzzy diagnosis matrix that embodies diagnostic expert knowledge, X represents the symptom library vector, Y represents the fault library vector, and “o” represents the fuzzy logic operator. Practically, we solve the relationship matrix Y = X o R, obtain the fault vector Y of the state to be determined, and then draw a reasonable diagnostic conclusion.
Ren et al. [32] combined fuzzy logic theory with the expert system theory in order to successfully diagnose faults within the wind turbine, thus improving the speed, accuracy, and reliability of the wind turbine fault diagnosis expert system. Silvio et al. [33] used empirical mode decomposition to extract fault feature vectors from vibration signals, and used multi-class fuzzy support vector machine (FSVM) classifiers to diagnose the faults of wind turbines. The multi-classification fault diagnosis of wind turbines was determined through the application of various experimental processes, and a higher diagnostic accuracy was obtained. Liu et al. [34] applied the fuzzy logic theory to fault diagnosis of the wind turbine pitch control system using the fault tree method. This method can be used to quickly and accurately diagnose the fault of the pitch system. Li et al. [35] proposed an improved fuzzy comprehensive condition evaluation method, which could be applied to the online data monitoring of grid-connected wind turbines (WTGS). Compared with the traditional fuzzy evaluation method, this method can predict the change of operating conditions with better consistency.
It is worth noting that the fuzzy relationship matrix R in fuzzy logic theory is synthesized from the opinions of many experts. There is great room for continuous improvement in the determination and conditions of use for R, as well as in the accuracy of fuzzy logic diagnosis.

3.1.2. Gray Theory

Gray theory was founded by Professor Deng Julong in 1982. It reveals the invisible relationship between known and unknown information from a systematic point of view, which is then logically applied to determine unknown variables using known variables. This theory and approach has gained great popularity and is widely used in economics [36,37,38]. Taking wind turbine systems as a reference, the fault diagnosis process based on vibration theory uses limited known information, such as vibrational energy, vibrational frequency, etc., to diagnose, predict, and make decisions pertaining to systems that contain unknown information through information processing. It is evident that gray theory can be regarded as a more suitable tool for equipment fault diagnosis [39]. So far, the gray correlation degree analysis method in gray theory is the most widely used technique in the diagnosis and identification of mechanical faults. The degree of correlation is calculated based on each object to be tested, and the comprehensive evaluation result is obtained from the diagnosis.
For example, Du et al. [40] proposed an improved FMEA method based on both gray theory and fuzzy theory, which was practically applied to the fault diagnosis and risk assessment of offshore wind turbine gearboxes. Case studies have shown that this method exhibits higher reliability and simultaneously optimizes the allocation of maintenance resources. Jin et al. [41] applied the gray theory to the diagnosis and prediction of a rotating machinery rotor fault, and the result was consistent with other studies. Lin et al. [42] applied the gray theory to the time series prediction of wind turbine operating state parameters, established a gray prediction model with non-equal intervals, and proved the feasibility of the method by comparing the online monitoring data of wind turbines. This method was compared with the prediction results of BP neural networks and support vector machines, with the results demonstrating higher prediction accuracy using gray theory in instances where the operating parameters drastically change. Yang et al. [43] used dual-tree wavelet transformation to extract fault features, and then calculated the gray level co-occurrence matrix of fault features in order to realize the classification of fault types. The results showed that the method was robust and achieved higher accuracy than traditional methods.
Using the gray theory analysis method does not require a large number of samples, and similarly requires a relatively small amount of calculation. However, it is essential to select the characteristic parameters of the object to be diagnosed and effectively establish a reference state model. The use of this method began late in the application of rotating machinery fault diagnosis, but has subsequently achieved remarkable results.

3.1.3. Fuzzy Petri Net

Petri net is a widely used tool for modeling and analysis of discrete event systems, and can effectively model fault diagnosis systems with functional applicability in a wide range of problems [44,45,46]. However, in practice, uncertainties in the fault information are often present. For example, the faults of wind turbines are characterized by diversity and complexity. There are substantial differences in fault symptoms, which can have similar characteristics, and it is often the case that multiple faults occur at the same time. In order to provide more certainty with regard to these principals, Chen et al. [47] proposed a fuzzy reasoning algorithm based on the Petri net reachability graph. Fuzzy Petri nets demonstrate certain advantages when used to describe the asynchronous and concurrent activities of the system, and are therefore suitable for studying component-level and system-level faults.
Huang et al. [48] applied the fuzzy Petri net to the fault detection of the gearbox, and provided positive and negative inferences according to the uncertainty of the cause of the fault, which subsequently enables the fault diagnosis problem of the gearbox to be solved, explicitly under complex conditions. Liu et al. [49] used the fuzzy Petri theory to model the discrete state in the operation of the wind turbine hydraulic pitch system, which provides an important basis for system maintenance. Moreover, Guo et al. [50] combined the fuzzy Petri net with the expert system to develop a remote network fault diagnosis system. The accuracy and effectiveness of this method were verified by the active simulation and use of various examples, which illustrated the ability to solve the problems of ambiguity, insufficient information, and, ultimately, in-process diagnosis of gearbox faults.
However, Petri net lacks an adaptive learning rule in knowledge expression and reasoning. With the widespread use of Petri net theory in the field of fault diagnosis, a variety of Petri net models have emerged, and, therefore, a standard model is urgently needed to streamline the process.

3.2. Fault Diagnosis Based on the Smart Technology Method

With the rapid development of artificial intelligence technology (particularly in the fields of knowledge engineering, expert systems, and artificial neural networks), as well as its application to the field of fault diagnosis, there is an increasing need for in-depth research and systematic investigation [51,52]. Common intelligent fault diagnosis methods for large wind turbines include an artificial neural network (ANN), the expert system, support vector machines (SVM), and deep learning (DP).

3.2.1. Artificial Neural Network

The fault learning method based on ANN is used to execute internal learning and processing of the network layer on a large amount of fault information collected by the input layer, in order to establish the mapping relationship between fault symptoms and fault classification. This is finally used to obtain fault diagnosis results and processing methods at the output layer [53,54].
Long et al. [55] proposed a fault diagnosis method for wind turbine gearboxes based on a particle swarm optimization BP neural network, and extracted the power spectrum entropy and wavelet entropy of the gearbox as the fault eigenvalues. The results of this experiment showed that fault identification accuracy was high, which inferred that this specific experiment presents a new and improved method for the in-depth study of gearbox fault diagnosis. Xiao et al. [56] analyzed Supervisory Control and Data Acquisition (SCADA) data, extracted the relevant parameters, and predicted the fault of the pitch system. Compared with that of the BP neural network, the diagnostic accuracy of the wavelet BP neural network showed a 17% increase, and the affiliated reliability rate was increased by 18%. Moreover, the results expressed an increase of 15.4%, with a related 17% decrease in the diagnostic false alarm rate. Furthermore, Zhang et al. [57] proposed a long-short-term memory neural network model, using wavelet transformation to extract the fault feature vector of rolling bearings for model training. This provided the ability to realize the accurate diagnosis of three kinds of frequent rolling bearing faults. Mi et al. [58] adopted the improved BP neural network based on both the momentum method as well as learning rate self-adaptation, and verified that the used model effectively reduced the number of training times through examples and provided effective application value to the fault diagnosis of wind turbines.
When using the artificial neural network method to diagnose faults, one is limited to the identification of known fault types and, therefore, has no ability to identify new faults. Moreover, the training of artificial neural network models is also time-consuming, and the verification process is complicated. These are all relevant issues that merit further investigation in order to improve the outcome of the systematic information process.

3.2.2. Expert System

Expert system (ES) theory was first proposed in 1965, and is regarded as one of the smart diagnostic methods used for wind turbine fault diagnosis. The expert system gathers the knowledge of many experts and concurrently analyzes and compares the fault information in order to finally obtain an intelligent diagnosis [59,60,61].
Feng et al. [62] effectively combined the decision tree algorithm with the expert system to develop a fault diagnosis expert system based on the C language integrated production system (CLIPS). The diagnosis results revealed that the misdiagnosis rate was below 5%. This result was better than the previously obtained results. In addition, Zhang et al. [63] effectively summarized the expert knowledge of wind turbine fault diagnosis, and used the ASP.NET platform development and SQL Server 2008 database to build a more effective wind turbine fault diagnosis expert system which could achieve the online fault diagnosis of wind turbines. Zhou et al. [64] effectively combined the fuzzy set and fuzzy reasoning method with the expert system to develop a distributed power system fault diagnosis expert system based on fuzzy reasoning. After testing, it solved a complex problem in power system fault diagnosis which was then further ratified by user experience. Yang et al. [65] established a fault tree model according to the fault characteristics of the gearbox, and subsequently developed a web-oriented expert system on the .NET platform through C#, enabling the gearbox fault expert solution to be more effective and accurate.
The diagnostic performance of the expert system depends on the quality level of the expert knowledge base. However, even with a high-quality expert knowledge base, the acquisition of this knowledge is still relatively difficult, and is regarded as the primary bottleneck of system development. Furthermore, with the need for continuous improvement of system reliability and stability, the development trend of expert systems has changed from the widely employed offline detection method to real-time diagnostic monitoring processes.

3.2.3. Support Vector Machine

The support vector machine is a machine learning algorithm proposed by VAPNIK [66]. It is effective in solving small sample and nonlinear practical classification problems. It is easy to model and has good adaptability to the fault data collected on wind turbines, and has been widely used in the field of fault diagnosis [67,68].
Zhang et al. [69] used the empirical mode decomposition (EMD) method to decompose the gear vibration signal, and used the energy feature in the intrinsic mode function (IMF) component containing the main fault information as input to establish an SVM. The results show that this method can successfully identify the working state and fault type of the gear. Moreover, diagnostic accuracy is higher than the method of combining neural networks and EMD. Furthermore, Shi et al. [70] combined the spectral features, adaptively extracted by deep learning, and the time-domain features, extracted by mathematical statistics methods, to form a joint feature vector. The particle swarm support vector machine was then used to carry out gear fault diagnosis on the joint feature vector, thus realizing the medium-speed reliable identification of various types of faults of shaft gears. When compared with traditional fault diagnosis methods, this method is faster and achieves higher diagnostic accuracy. The improved hypersphere multi-class support vector machine has been applied to the multi-class fault diagnosis of rolling bearings and to accurately assess the degree of performance degradation [71]. Given the evidence that supports the efficacy of these methods, current investigations [72] have proposed a method combining EMD and SVM that can be effectively applied to the fault diagnosis of the rotor system.
The support vector machine was developed late in fault diagnosis research and requires further development to attain a significant level of performance. For example, there is room for improvement in the SVM method which applies to the corresponding reduction of the computational loads directly applied to the algorithm. Specifically, in fault diagnosis, multiple faults often occur at the same time, but SVM can only individually assess faults, and the proposed combination method displays several defects that need to be addressed.

3.2.4. Deep Learning

The concept of deep learning comes from artificial neural networks. A multi-layer perceptron with multiple hidden layers is a deep learning structure. Deep learning forms a more abstract high-level representation by combining low-level features to discover the distributed feature representation of data [73,74,75,76,77,78]. Common deep learning algorithms include: Deep Belief Network (DBN), Convolutional Neural Network (CNN), Stacked Autoencoder (SAE), and Recurrent Neural Network (RNN) [79,80].
A deep learning method based on the layer-by-layer encoding network of the SCADA condition monitoring data on the main bearing of the wind turbine has been proposed, whereby the effective detection of the main bearing fault of the wind turbine has been accurately determined [81]. Liu et al. [82] proposed a Deep Auto-encoder (DA) model, and the simulation proved that this method could effectively detect the state of the gearbox. Furthermore, the SAE network to deep-learn the relationships within the SCADA data of the generator has been developed, and the condition monitoring results of the generator have been subsequently extracted. Combined with the trend change of the adaptive threshold detection state monitoring results, the fault assessment was realized, and the abnormal change of the variable residual trend was further used to analyze the cause of the fault [83]. Zhou et al. [84] proposed a neural network model of residual convolutional autoencoders, which can be applied to the feature extraction of vibration signals in order to effectively achieve gearbox fault diagnosis.
With an appreciation of the powerful learning ability of the deep learning algorithm, the use of deep learning technology clearly exhibits great potential in the fault diagnosis process of complex wind turbines. However, the model training of deep learning is time-consuming, and the fault diagnosis method based on deep learning requires a large amount of training data. Due to the large amount of calculation involved in this method, future research should focus on optimizing model parameters and improving the operational speed of deep learning.

4. Prediction and Evaluation Technology for Determining Wind Turbine Status

4.1. Fault Prediction

Fault prediction begins with the assessment of the current use conditions of the equipment, and then analyzes the equipment’s condition monitoring data and fault model, effectively combining its structure, operating environment, and other parameters in order to successfully predict the probability of faulty components in the future, including determining which type of fault may occur. Observing the nature, category, and cause of the fault, as well as other factors, can reveal the development trend of the fault and predict the service life of the equipment. In some instances, failure prediction is also referred to as remaining life prediction [85,86,87,88,89]. Fault prediction methods can generally be divided into three categories: the first is based on the mechanism of physical failure (mechanical failure); the second is based on data-driven methods; and the third is comprised from the fusion of physical mechanisms and data-driven methods. The method of physical failure mechanism assessment mainly involves constructing physical models based on the theory of equipment damage, fatigue, cracks, etc., and predicts the remaining life of the equipment. Methods based on the fusion of the two are generally used to predict the remaining lifetime or failure of a single device [90]. In view of the objective characteristics of wind turbines, such as variable operating conditions and various failure mechanisms, it is often difficult to analyze the remaining life of wind turbines using prediction techniques that rely upon physical mechanisms. Therefore, data-driven failure prediction techniques should also be used. This newer technology differs from the relatively old fault diagnosis technology that has been developed for more than 30 years (although the research time of fault prediction technology is relatively short), and it has garnered significant attention in the industry.
For example, investigative models [91] employed the least squares support vector machine (LSSVM) to learn the normal operational behavior of the gearbox, and used statistical process control technology to analyze the residual error between the actual value and the estimated value of the gearbox oil temperature and bearing temperature that enables the effective prediction of an abnormal gearbox state. Alternatively, a linear regression analysis and prediction model was established based on the main bearing temperature under normal operating conditions through the collection of SCADA data on wind turbines, which was then subsequently used to introduce a deviation estimation function in order to predict the potential failure of the main bearing [92]. Moreover, the failure modes and categories of gearboxes were specifically considered, and the failure trend method based on the data and model was studied, which enabled the accurate estimation of the remaining life of gearboxes [93].
Fault prediction technology of wind turbines is still developing, and it still faces many difficulties. For example, fault prediction technology for physical failure requires more research both at home and abroad. To this end, the operating conditions of wind turbines are more complex today than they previously were, which can have a great impact on prediction results. In addition, the components of wind turbines are interconnected, and the current prediction models are mostly applied to single components. These factors contribute to a likelihood of undetected parameters which limit the outcomes and decrease the accuracy of predictions.

4.2. Status Assessment

Status evaluation refers to the use of comprehensive evaluation and analysis methods to divide the current status of equipment into different levels, such as excellent, good, medium, poor, and faulty, through the study of system data and manually recorded information [90]. In recent years, as the SCADA system has become widely used in the assessment of wind turbines, condition monitoring of important components of wind turbines has become more common [94,95]. Using the data collected by the SCADA system combined with evaluation and analysis methods can better reflect the operating status of the equipment, realize an accurate status assessment of the wind turbine, and improve the safety and reliability of the unit and its operation. At present, the most commonly used methods for the evaluation of the status of wind turbines include the probability statistics method, the intelligent method based on neural networks, and the fuzzy comprehensive evaluation method [96].
Huang et al. [96] proposed an evaluation model for the operating status of offshore direct-drive wind turbines that integrates multiple methods, such as the correlation coefficient method, the degradation degree analysis method, and fuzzy comprehensive evaluation to evaluate the deterioration phenomenon. Li Hui et al. [97] established an improved model of fuzzy comprehensive evaluation specifically for the online operation status of wind turbines based on SCADA data. Compared with conventional evaluation methods, this method can better reflect the real operation status of wind turbines. Liang Ying et al. [98] proposed an online evaluation scheme based on the combination of a regression prediction model and a SCADA alarm system, and established an SVR model to predict the active power of wind turbines. This method not only exhibited the function of a SCADA system, but could also bring about a greater understanding of the operating conditions of wind turbines. Zhou Yuan et al. [99] proposed a two-layer evaluation model method based on set pair analysis and evidence theory. Compared with traditional evaluation methods, this evaluation method was more accurate, and the trend analysis of status was also superior.

5. Comparison and Analysis

In the previous section of this article, the most common fault diagnosis methods for wind turbines were introduced. In this section, various fault diagnosis techniques are compared and analyzed through tables. A comparison of early state monitoring technologies is provided in Table 1, while Table 2 presents a comparison of current fault diagnosis technologies.

6. Conclusions and Outlook

With regard to the developmental needs of wind turbine fault diagnosis and fault prediction technology, this paper reviewed the early fault condition monitoring methods of wind turbines, with emphasis on the current techniques of fault diagnosis based on mathematical methods. Furthermore, smart technology methods of fault diagnosis, as well as failure prediction and condition assessment techniques for wind turbines, were also summarized in this work. Focusing on the existing problems observed in fault diagnosis, fault prediction, and state assessment of large wind turbines, the following research points and trends are proposed.
  • The wind turbine is a relatively complex interconnected and integrated electromechanical system. Electrical faults and mechanical faults can interact with each other and form more complex faults. Fault diagnosis based on electrical signals can introduce the possibility of online fault diagnosis without additional sensors, and has the advantages of low cost and reliable signals. The development of new ways to use this information in order to realize the accurate fault diagnosis of wind turbines is expected to be a key area for future research.
  • Fault prediction technology for wind turbines has just emerged, and there are many improvements that can be made. For example, there are many online monitoring parameters for wind turbines. The current research is mostly based on a single data source and serves to predict the fault of a single component of the wind turbine. Fault prediction technology combining various data points improves the reliability of fault prediction results.
  • The wind turbine is a very large and complex system. At present, the research based on the condition monitoring of the important components of the wind turbine is relatively old, but research on the conditional assessment of the whole wind turbine is still in its infancy. Although some good results have been achieved, due to the large, complex, and interconnected characteristics of wind turbines, a single evaluation method cannot fully deal with the uncertainty of wind turbine operation. Therefore, evaluation technology based on multi-method fusion will be an important direction for wind turbine status assessment in the future.
  • Intelligent fault diagnosis technology is the key to realizing online diagnosis of large wind turbines; remote fault diagnosis technology of wireless sensing and layered distributed networks will be an important development direction for the online monitoring of wind power technology.

Author Contributions

Writing—original draft preparation, M.C.; writing—review and editing, M.C. and F.Z.; supervision, Y.Z.; project administration, Q.L.; funding acquisition, K.Z. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by National Natural Science Foundation of China, grant number U2141245 and the Foundation Project of Chongqing Normal University, grant number 22XLB008.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Not applicable.

Acknowledgments

This research is supported by Chongqing Normal University.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Jiang, Z. Reflections on Energy Issues in China—ScienceDirect. Res. Energy Issues China 2010, 13, 7, 9–52. [Google Scholar]
  2. Hu, A. China’s goal of achieving carbon peak by 2030 and its main approaches. J. Beijing Univ. Technol. 2021, 21, 1–15. [Google Scholar]
  3. Rahman, A.; Farrok, O.; Haque, M.M. Environmental impact of renewable energy source based electrical power plants: Solar, wind, hydroelectric, biomass, geothermal, tidal, ocean, and osmotic. Renew. Sustain. Energy Rev. 2022, 161, 112279. [Google Scholar] [CrossRef]
  4. Li, J.; Wu, Z.; Tan, X.; Chen, B. Review of wind power generation and relative technology development. Electr. Power Constr. 2011, 32, 64–72. [Google Scholar]
  5. Zhang, B.-Q.; Yang, Y.-M. Status and trend of wind & photovoltaic power development. Electr. Power 2006, 39, 65–69. [Google Scholar]
  6. Huang, L.; Cao, J.; Zhang, K.; Fu, Y.; Xu, H. Status and prospects on operation and maintenance of offshore wind turbines. Proc. CSEE 2016, 36, 729–738. [Google Scholar]
  7. Jun, H.; Jianzhong, Z.; Ming, C.; Wei, W.; Ming, Z. An overview of condition monitoring and fault diagnostic for wind energy conversion system. Trans. China Electrotech. Soc. 2013, 28, 261–271. [Google Scholar]
  8. Li, G.; Qi, Y.; Li, Y.; Zhang, J.; Zhang, L. Research progress on fault diagnosis and state prediction of wind turbine. Autom. Electr. Power Syst. 2021, 45, 180–191. [Google Scholar]
  9. Kabir, M.; Oo, A.M.; Rabbani, M. A brief review on offshore wind turbine fault detection and recent development in condition monitoring based maintenance system. In Proceedings of the 2015 Australasian Universities Power Engineering Conference (AUPEC), Wollongong, NSW, Australia, 27–30 September 2015; pp. 1–7. [Google Scholar]
  10. Wilkinson, M.; Spianto, F.; Knowles, M. Towards the zero maintenance wind turbine. In Proceedings of the 41st International Universities Power Engineering Conference, Newcastle upon Tyne, UK, 6–8 September 2006; pp. 74–78. [Google Scholar]
  11. Faulstich, S.; Hahn, B.; Tavner, P.J. Wind turbine downtime and its importance for offshore deployment. Wind Energy 2011, 14, 327–337. [Google Scholar] [CrossRef]
  12. Lin, H. Overview of offshore wind power key technologies. J. Southeast Univ. Nat. Sci. Ed. 2011, 41, 882–888. [Google Scholar]
  13. Qiao, W.; Lu, D. A survey on wind turbine condition monitoring and fault diagnosis—Part I: Components and subsystems. IEEE Trans. Ind. Electron. 2015, 62, 6536–6545. [Google Scholar] [CrossRef]
  14. Amirat, Y.; Benbouzid, M.E.H.; Al-Ahmar, E.; Bensaker, B.; Turri, S. A brief status on condition monitoring and fault diagnosis in wind energy conversion systems. Renew. Sustain. Energy Rev. 2009, 13, 2629–2636. [Google Scholar] [CrossRef] [Green Version]
  15. Hau, E. Wind Turbines: Fundamentals, Technologies, Application, Economics; Springer Science & Business Media: Berlin/Heidelberg, Germany, 2013. [Google Scholar]
  16. Yang, W.; Tavner, P.J.; Crabtree, C.J.; Feng, Y.; Qiu, Y. Wind turbine condition monitoring: Technical and commercial challenges. Wind Energy 2014, 17, 673–693. [Google Scholar] [CrossRef] [Green Version]
  17. Jin, X.; Sun, Y.; Shan, J.; Wu, G. Fault diagnosis and prognosis for wind turbines: An overview. Chin. J. Sci. Instrum. 2017, 38, 1041–1053. [Google Scholar]
  18. Jiang, D.X.; Hong, L.Y.; Huang, Q.; Ding, Y.S. Condition Monitoring and Fault Diagnostic Techniques for Wind Turbine. Power Syst. Clean Energy 2008, 24, 40–44. [Google Scholar]
  19. Quan, L.; Liu, Y.; Yang, Y. Applications of Condition Monitoring and Fault Diagnosis to Wind Turbines. Mod. Electr. Power 2008, 25, 55–59. [Google Scholar]
  20. Na, W.U.; Sun, L.L.; Yang, P. Research on Wind Turbine Condition Monitoring and Fault Diagnosis. J. North China Inst. Water Conserv. Hydroelectr. Power 2012, 33, 86–90. [Google Scholar]
  21. Nie, M.; Wang, L. Review of condition monitoring and fault diagnosis technologies for wind turbine gearbox. Procedia CIRP 2013, 11, 287–290. [Google Scholar] [CrossRef] [Green Version]
  22. Zheng, G.; Wang, W. A new cepstral analysis procedure of recovering excitations for transient components of vibration signals and applications to rotating machinery condition monitoring. J. Vib. Acoust. 2001, 123, 222–229. [Google Scholar] [CrossRef]
  23. Peeters, C.; Guillaume, P.; Helsen, J. Vibration-based bearing fault detection for operations and maintenance cost reduction in wind energy. Renew. Energy 2018, 116, 74–87. [Google Scholar] [CrossRef]
  24. Abboud, D.; Antoni, J.; Sieg-Zieba, S.; Eltabach, M. Envelope analysis of rotating machine vibrations in variable speed conditions: A comprehensive treatment. Mech. Syst. Signal Process. 2017, 84, 200–226. [Google Scholar] [CrossRef]
  25. Popa, L.M.; Jensen, B.-B.; Ritchie, E.; Boldea, I. Condition monitoring of wind generators. In Proceedings of the 38th IAS Annual Meeting on Conference Record of the Industry Applications Conference, Salt Lake City, UT, USA, 12–16 October 2003; pp. 1839–1846. [Google Scholar]
  26. Jian, D.; Jinyuan, W.; Rong, J.; Ji, L. Feature Extraction Method for Unbalanced Fault Electrical Signal of Wind Turbine Impeller Based on Hilbert and CA-VMD. Power Syst. Clean Energy 2021, 37, 112–118+126. [Google Scholar]
  27. Huo, J.; Tang, G.; Liu, D.; Zhang, M.; Energy, S.O. A novel monitoring method of wind turbine generator real-time reliability based on tempeture signals. Renew. Energy Resour. 2016, 34, 408–412. [Google Scholar]
  28. Liang, P.; Sun, Y.; Chen, T.; Zhou, X.; Yue, J. Typical Cases of Oil-monitoring for Gearbox of Wind Turbine. Lubr. Eng. 2016, 41, 141–145. [Google Scholar]
  29. Zhang, S.; Asakura, T.; Xu, X.; Xu, B. Fault diagnosis system for rotary machine based on fuzzy neural networks. JSME Int. J. Ser. C Mech. Syst. Mach. Elem. Manuf. 2003, 46, 1035–1041. [Google Scholar] [CrossRef]
  30. Wang, H.; Chen, P. Fuzzy diagnosis method for rotating machinery in variable rotating speed. IEEE Sens. J. 2010, 11, 23–34. [Google Scholar] [CrossRef]
  31. He, Y. Application of Fuzzy Diagnosis Technique on Fault Diagnosis for Rotational Machines. J. China Text. Univ. 1998, 24, 41–45. [Google Scholar]
  32. Lachun, R.; Lida, Z. Construction of fault diagnosis expert system for wind turbine using fuzzy theory. Mech. Sci. Technol. 2007, 26, 581. [Google Scholar]
  33. Simani, S.; Farsoni, S.; Castaldi, P. Wind turbine simulator fault diagnosis via fuzzy modelling and identification techniques. Sustain. Energy Grids Netw. 2015, 1, 45–52. [Google Scholar]
  34. Liu, H.; Dong, X.-H.; Yang, Z.-L.; Zheng, K. The application of intelligent fuzzy inference to the fault diagnosis in pitch-controlled system. Energy Procedia 2012, 16, 1839–1844. [Google Scholar] [CrossRef] [Green Version]
  35. Li, H.; Hu, Y.; Yang, C.; Chen, Z.; Ji, H.; Zhao, B. An improved fuzzy synthetic condition assessment of a wind turbine generator system. Int. J. Electr. Power Energy Syst. 2013, 45, 468–476. [Google Scholar] [CrossRef]
  36. Liu, S. Emergence and development of grey system theory and its forward trends. J. Nanjing Univ. Aeronaut. Astronaut. 2004, 36, 267–272. [Google Scholar]
  37. Liu, S.; Forrest, J.Y.-L. Advances in Grey Systems Research; Springer: Berlin/Heidelberg, Germany, 2010. [Google Scholar]
  38. Lin, P.; Wu, Q. The grey relational degree analysis and its application. J. Salt Lake Res. 2001, 2, 48–50. [Google Scholar]
  39. Geng, L.; Pan, X.; Li, X. The applications of the grey system in fault diagnosis. J. Beijing Inst. Technol. Chin. Ed. 1997, 17, 194–197. [Google Scholar]
  40. Du, Z.; Zhu, H.; Tang, G. Grey theory-based analysis on failure mode of offshore wind turbine gearbox and its impact. Water Resour. Hydropower Eng. 2017, 48, 165–169+164. [Google Scholar]
  41. Jin, C.; Qiu, Y.; Duan, Z. Grey system applications in fault diagnosis and forecasting for rotating machinery. Chin. J. Appl. Mech. 2000, 17, 74–79. [Google Scholar]
  42. Li, H.; Li, X.; Hu, Y.; Yang, C.; Zhao, B. Non-equidistant grey prediction on running condition parameters of a wind turbine generator system. Autom. Electr. Power Syst. 2012, 36, 29–34. [Google Scholar]
  43. Yang, P.; Yang, G. Feature extraction using dual-tree complex wavelet transform and gray level co-occurrence matrix. Neurocomputing 2016, 197, 212–220. [Google Scholar]
  44. Peterson, J.L. Petri nets. ACM Comput. Surv. CSUR 1977, 9, 223–252. [Google Scholar] [CrossRef]
  45. Lin, Z.; Wen, F.; Chung, C.; Wong, K. A survey on the applications of Petri net theory in power systems. In Proceedings of the 2006 IEEE Power Engineering Society General Meeting, Montreal, QC, Canada, 18–22 June 2006; p. 7. [Google Scholar]
  46. David, R.; Alla, H. Petri nets for modeling of dynamic systems: A survey. Automatica 1994, 30, 175–202. [Google Scholar] [CrossRef]
  47. Manoj, T.; Leena, J.; Soney, R.B. Knowledge representation using fuzzy Petri nets-revisited. IEEE Trans. Knowl. Data Eng. 1998, 10, 666–667. [Google Scholar] [CrossRef]
  48. Huang, M. Fault diagnosis of wind power generator set based on fuzzy Petri net. Comput. Eng. Appl. 2012, 48, 206–210. [Google Scholar]
  49. Liu, W. Modeling and Fault Diagnosis of Hydraulic Variable Propeller Pitch System of Wind-Driven Generator; North China Electric Power University: Beijing, China, 2009. [Google Scholar]
  50. Guo, H.; Li, G.; Li, T. Analysis of gearbox fault diagnosis system of wind turbine based on fuzzy Petri nets. In Proceedings of the 2012 9th International Conference on Fuzzy Systems and Knowledge Discovery, Chongqing, China, 29–31 May 2012; pp. 540–544. [Google Scholar]
  51. Li, H.-W.; Yang, D.-S.; Sun, Y.-L.; Han, J. Study review and prospect of intelligent fault diagnosis technique. Comput. Eng. Des. 2013, 34, 632–637. [Google Scholar]
  52. Lei, Y.; He, Z.; Zi, Y. A new approach to intelligent fault diagnosis of rotating machinery. Expert Syst. Appl. 2008, 35, 1593–1600. [Google Scholar] [CrossRef]
  53. Yu, H.; Chen, C.; Zhang, S.; Zhou, J. Intelligent Diagnosis Based on Neural Networks; Metallurgy Industry Publication House: Beijing, China, 2000; pp. 1–294. [Google Scholar]
  54. Da-Qi, Z. The research progress and prospects of artificial neural networks. J. South. Yangtze Univ. 2004, 3, 103–111. [Google Scholar]
  55. Long, Q.; Liu, Y.; Yang, Y. Fault diagnosis method of wind turbine gearbox based on BP neural network trained by particle swarm optimization algorithm. Acta Energ. Sol. Sin. 2012, 33, 120–125. [Google Scholar]
  56. Cheng, X.; Liu, Z. Fault prediction of variable pitch system of wind turbine based on wavelet BP neural network. Renew. Energy Resour. 2017, 35, 893–899. [Google Scholar]
  57. Zhang, J.; Yu, S.; Gang, L.; Wang, C.; Jiao, Y. A Method of Fault Diagnosis for Rolling Bearing of Wind Turbines Based on Long Short-term Memory Neural Network. Comput. Meas. Control 2017, 25, 16–19. [Google Scholar]
  58. Mi, J.; Ji, G. Application of improved BP neural network in fault diagnosis of Fans. Noise Vib. Control 2011, 4, 94–98. [Google Scholar]
  59. Zhang, Y.; Wu, L.; Wang, S. Survey on development of expert system. Comput. Eng. Appl. 2010, 46, 43–47. [Google Scholar]
  60. Yang, X.; Zhu, D.Q.; Sang, Q.B. Research and Prospect of Expert System. Appl. Res. Comput. 2007, 24, 4–9. [Google Scholar]
  61. Mingqiang, W.; Hui, S.; Xiaohua, Z.; Kaiqing, X. Re-search and prospect of fault diagnosis expert system. Comput. Autom. Meas. Control 2005, 1301–1304. [Google Scholar] [CrossRef]
  62. Feng, Y.; Yang, W.; Jiang, D. Fault Recognition Method for Gear Case of Wind Power Generator Based on Decision Tree Classification Algorithm and Expert System. Guangdong Electr. Power 2013, 26, 17–21. [Google Scholar]
  63. Zhang, X.; Xu, B.; Wu, G. Research on fault diagnosis expert system of wind turbine. Manuf. Autom. 2014, 36, 21–25. [Google Scholar]
  64. Ming, Z.; Gengyin, L.; Jianwen, R.; Kaili, X.; Youjiang, S. A distributed power system fault diagnosis expert system based on fuzzy inference. In Proceedings of the 2000 International Conference on Advances in Power System Control, Operation and Management, APSCOM-00, Hong Kong, China, 30 October–1 November 2000. [Google Scholar]
  65. Zhi-Ling, Y.; Bin, W.; Xing-Hui, D.; Hao, L. Expert system of fault diagnosis for gear box in wind turbine. Syst. Eng. Procedia 2012, 4, 189–195. [Google Scholar] [CrossRef] [Green Version]
  66. Cortes, C.; Vapnik, V. Support-vector networks. Mach. Learn. 1995, 20, 273–297. [Google Scholar] [CrossRef]
  67. Yuan, S.; Chu, F. Support vector machines and its applications in machine fault diagnosis. J. Vib. Shock 2007, 26, 29–35. [Google Scholar]
  68. Widodo, A.; Yang, B.-S. Support vector machine in machine condition monitoring and fault diagnosis. Mech. Syst. Signal Process. 2007, 21, 2560–2574. [Google Scholar] [CrossRef]
  69. Zhang, C.; Chen, J.-J.; Guo, X. A gear fault diagnosis method based on EMD energy entropy and SVM. J. Vib. Shock 2010, 29, 216–220. [Google Scholar]
  70. Shi, P.K.N.S. Intelligent fault diagnosis for gears based on deep learning feature extraction and particle swarm optimization SVM state identification. China Mech. Eng. 2017, 28, 1056. [Google Scholar]
  71. Kang, S.; Wang, Y.; Yang, G.; Song, L.; Mikulovich, V. Rolling bearing fault diagnosis method using empirical mode decomposition and hypersphere multiclass support vector machine. Zhongguo Dianji Gongcheng Xuebao 2011, 31, 96–102. [Google Scholar]
  72. Yu, D.-J.; Chen, M.-F.; Cheng, J.-S.; Yang, Y. A fault diagnosis approach for rotor systems based on empirical mode decomposition method and supportvector machines. In Proceedings of the Zhongguo Dianji Gongcheng Xuebao; Chinese Society of Electrical Engineering: Shanghai, China, 2006; pp. 162–167. [Google Scholar]
  73. Du, X.; Cai, Y.; Wang, S.; Zhang, L. Overview of deep learning. In Proceedings of the 2016 31st Youth Academic Annual Conference of Chinese Association of Automation (YAC), Wuhan, China, 11–13 November 2016; pp. 159–164. [Google Scholar]
  74. Hornik, K.; Stinchcombe, M.; White, H. Multilayer feedforward networks are universal approximators. Neural Netw. 1989, 2, 359–366. [Google Scholar] [CrossRef]
  75. Chang, C.-H. Deep and shallow architecture of multilayer neural networks. IEEE Trans. Neural Netw. Learn. Syst. 2015, 26, 2477–2486. [Google Scholar] [CrossRef] [PubMed]
  76. Hinton, G.E.; Salakhutdinov, R.R. Reducing the dimensionality of data with neural networks. Science 2006, 313, 504–507. [Google Scholar] [CrossRef] [Green Version]
  77. LeCun, Y.; Bengio, Y.; Hinton, G. Deep learning. Nature 2015, 521, 436–444. [Google Scholar] [CrossRef]
  78. Bengio, Y.; Goodfellow, I.; Courville, A. Deep Learning; MIT Press: Cambridge, MA, USA, 2017; Volume 1. [Google Scholar]
  79. Ren, H.; Qu, J.; Chai, Y.; Tang, Q.; Ye, X. Deep learning for fault diagnosis: The state of the art and challenge. Control Decis. 2017, 32, 1345–1358. [Google Scholar]
  80. Chenglin, W.; Feiya, L. Review on deep learning based fault diagnosis. J. Electron. Inf. Technol. 2020, 42, 234–248. [Google Scholar]
  81. Zhao, H.; Liu, H. Fault detection of wind turbine main bear based on deep learning network. Acta Energ. Sol. Sin. 2018, 39, 588–595. [Google Scholar]
  82. Liu, H.; Zhao, X.; Zhao, H. Fault detection of wind turbine gearbox based on deep autoencoder network. Trans. China Electrotech. Soc. 2017, 32, 156–163. [Google Scholar]
  83. Zhao, H.; Liu, H.; Lin, Y. Condition Monitoring and Fault Diagnosis of Wind Turbine Generator Based on Stacked Autoencoder Network. Dianli Xitong Zidonghua/Autom. Electr. Power Syst. 2018, 42, 102–108. [Google Scholar]
  84. Zhou, X.; Yu, J. Gearbox fault diagnosis based on one-dimension residual convolutional auto-encoder. J. Mech. Eng. 2020, 56, 96–108. [Google Scholar]
  85. Jin, X.; Sun, Y.; Que, Z.; Wang, Y.; Chow, T.W. Anomaly detection and fault prognosis for bearings. IEEE Trans. Instrum. Meas. 2016, 65, 2046–2054. [Google Scholar] [CrossRef]
  86. Wang, Y.; Peng, Y.; Zi, Y.; Jin, X.; Tsui, K.-L. A two-stage data-driven-based prognostic approach for bearing degradation problem. IEEE Trans. Ind. Inform. 2016, 12, 924–932. [Google Scholar] [CrossRef]
  87. Peng, Y.; Wang, Y.; Zi, Y. Switching state-space degradation model with recursive filter/smoother for prognostics of remaining useful life. IEEE Trans. Ind. Inform. 2018, 15, 822–832. [Google Scholar] [CrossRef]
  88. Li, C.J.; Lee, H. Gear fatigue crack prognosis using embedded model, gear dynamic model and fracture mechanics. Mech. Syst. Signal Process. 2005, 19, 836–846. [Google Scholar] [CrossRef]
  89. Si, X.-S.; Wang, W.; Hu, C.-H.; Zhou, D.-H. Remaining useful life estimation—A review on the statistical data driven approaches. Eur. J. Oper. Res. 2011, 213, 1–14. [Google Scholar] [CrossRef]
  90. Jin, X.; Wang, Y.; Zhang, B. Industrial big data-driven fault prognostics and health management. Comput. Integr. Manuf. Syst. 2022, 28, 1314–1336. [Google Scholar]
  91. Zhao, H.-S.; Hu, Q.-C.; Li, Z. Failure prediction of wind turbine gearbox based on statistical process control. Power Syst. Prot. Control 2012, 40, 67–73. [Google Scholar]
  92. Zhang, X.; Yan, S.; Zhou, X.; Zhao, H. Early stage failure forecast method for main bearing of wind turbine based on state monitoring. Guangdong Electr. Power 2012, 25, 6–10. [Google Scholar]
  93. Ma, Y.F. Research on Fault Trend Prediction Method of Wind Turbine Gearbox; North China Electric Power University: Beijing, China, 2013. [Google Scholar]
  94. Hameed, Z.; Hong, Y.; Cho, Y.; Ahn, S.; Song, C. Condition monitoring and fault detection of wind turbines and related algorithms: A review. Renew. Sustain. Energy Rev. 2009, 13, 1–39. [Google Scholar] [CrossRef]
  95. Guo, P.; Xu, M.; Bai, N.; Ma, D. Wind turbine tower vibration modeling and monitoring driven by SCADA data. In Proceedings of the Zhongguo Dianji Gongcheng Xuebao; Chinese Society of Electrical Engineering: Shanghai, China, 2013; pp. 128–135. [Google Scholar]
  96. Biqing, H.; Yan, H.; Tingyan, W. Fuzzy synthetic evaluation of the operational status of offshore direct-drive wind turbines. Sci. Technol. 2015, 55, 543–549. [Google Scholar]
  97. Li, H.; Hu, Y.; Tang, X.; Liu, Z. Method for on-line operating conditions assessment for a grid-connected wind turbine generator system. Proc. CSEE 2010, 30, 103–109. [Google Scholar]
  98. Liang, Y.; Fang, R. An online wind turbine condition assessment method based on SCADA and support vector regression. Autom. Electr. Power Syst. 2013, 37, 7–12. [Google Scholar]
  99. Zhou, Y.; Xu, Q.P.; Li, J.; Wang, M.B.; Xiang, C.M. Operating state assessment based on set-pair analysis and evidential reasoning decision-making for wind turbine generator unit. Electr. Power Autom. Equip. 2017, 37, 38–45. [Google Scholar]
Energies 16 01125 g001 550
Figure 1. Electrical power generation by renewable energy source (RES)-based power plants.
Figure 1. Electrical power generation by renewable energy source (RES)-based power plants.
Energies 16 01125 g001
Energies 16 01125 g002 550
Figure 2. New wind power installations outlook for 2022–2026 (GW).
Figure 2. New wind power installations outlook for 2022–2026 (GW).
Energies 16 01125 g002
Energies 16 01125 g003 550
Figure 3. Framework of the present review.
Figure 3. Framework of the present review.
Energies 16 01125 g003
Energies 16 01125 g004 550
Figure 4. Schematic diagram of the fuzzy logic diagnosis method.
Figure 4. Schematic diagram of the fuzzy logic diagnosis method.
Energies 16 01125 g004
Table
Table 1. Comparison of commonly used signals in wind turbines.
Table 1. Comparison of commonly used signals in wind turbines.
Signals
Monitored		Signal Processing
Complexity		Installation
Complexity		ImplantOnline/Offline
VibrationMediumHighYesOnline
Electric SignalHigh/MediumLowNoOnline
TemperatureLowMediumYesOnline/Offline
OilLowMediumNoOnline
Table
Table 2. Comparison of current fault diagnosis methods for wind turbines.
Table 2. Comparison of current fault diagnosis methods for wind turbines.
Fault Diagnosis
Methods		AdvantagesDisadvantage
Fuzzy logic
diagnosis
➢
Fuzzy logic system has powerful processing ability for fuzzy information
➢
It is simple and practical, and has good application value
➢
The size of the fuzzy logic system exponentially increases with the number of failure modes in the wind turbine
➢
Rule setting requires a good understanding of wind turbine failure mechanisms
Gray theory
➢
Less sample size required
➢
It is better for short-term analysis of complex systems with uncertain factors
➢
Focuses less on absolute accuracy and more on relative ranking
➢
The medium- and long-term analysis of the system is less accurate
➢
It is often difficult to establish a typical fault library
Fuzzy Petri net
➢
It enables fast and accurate fault diagnosis of the system
➢
High diagnostic reasoning efficiency
➢
It can detect the quasi-fault state of the system
➢
It significantly increases the computational complexity of the system
➢
Lack of strong self-learning ability
Artificial neural network
➢
Able to use expert knowledge and experience to reason and judge
➢
It can continuously increase knowledge and revise existing knowledge
➢
It can solve its own reasoning process and answer the questions raised by users
➢
The size of the Expert system will exponentially grow with the number of failure modes
➢
Computational costs are high
➢
Unable to observe the learning process between, the output is difficult to interpret
Expert system
➢
Able to use expert knowledge and experience to reason and judge
➢
It can continuously increase knowledge and revise existing knowledge
➢
It can solve its own reasoning process and answer the questions raised by users
➢
The size of the Expert system will exponentially grow with the number of failure modes
➢
Computational costs are high
➢
Expert system does not achieve effective productivity
Support Vector Machine
➢
It can solve higher dimensional problems
➢
It does not have to rely on the entire data
➢
It can improve generalization ability
➢
Missing data sensitivity
➢
When there are many observation samples, the efficiency is not high
➢
There is no general solution to nonlinear problems
Deep Learning
➢
Powerful fitting ability
➢
It can extract valuable information from features for self-tuning and training
➢
It can also be combined with the probability model, so that the neural network has inference ability
➢
Model training is slow
➢
Interpretation of the middle layer is relatively poor
➢
Model optimization problems are more pronounced
➢
Visualization for deep learning fault diagnosis awaits further research
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Zhang, F.; Chen, M.; Zhu, Y.; Zhang, K.; Li, Q. A Review of Fault Diagnosis, Status Prediction, and Evaluation Technology for Wind Turbines. Energies 2023, 16, 1125. https://doi.org/10.3390/en16031125

AMA Style

Zhang F, Chen M, Zhu Y, Zhang K, Li Q. A Review of Fault Diagnosis, Status Prediction, and Evaluation Technology for Wind Turbines. Energies. 2023; 16(3):1125. https://doi.org/10.3390/en16031125

Chicago/Turabian Style

Zhang, Fanghong, Mingsong Chen, Yuze Zhu, Kai Zhang, and Qingan Li. 2023. "A Review of Fault Diagnosis, Status Prediction, and Evaluation Technology for Wind Turbines" Energies 16, no. 3: 1125. https://doi.org/10.3390/en16031125

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop