Fault classification in wind turbine based on deep belief network optimized by modified tuna swarm optimization algorithm

The high cost of operation and maintenance (O&M) management has become an important factor hindering the sustainable development of the wind power industry. Performing accurate condition assessment of wind turbine components to optimize the structural design and O&M strategy has become a research trend. However, the random and varying operating conditions of wind turbines make this problem difficult and challenging. A Supervisory Control and Data Acquisition (SCADA) system collects signals that contain a large amount of raw and useful information from critical wind turbine sub-assemblies. Extracting key information from the SCADA data is an economical and effective way for condition assessment. A real-time reliability assessment method of wind turbine components using a Back-Propagation Neural Network (BPNN) and SCADA data is presented in this paper. The normal behavior models are established with the processed SCADA data, and the real-time reliability of wind turbine components are assessed based on the prediction result. For verification, the BPNN-based reliability assessment method is applied to a gearbox with real SCADA data of a 1.5MW onshore wind turbine located along the southeast coast of China. The results show the capability of the proposed model in assessing the reliability of wind turbine components continuously and in real time.

Sensor fault detection of wind turbines plays an important role in improving the reliability and stable operation of turbines. The supervisory control and data acquisition (SCADA) system of a wind turbine provides promising insights into sensor fault detection due to the accessibility of the data and the abundance of sensor information. However, SCADA data are essentially multivariate time series with inherent spatio-temporal correlation characteristics, which has not been well considered in the existing wind turbine fault detection research. This paper proposes a novel classification-based fault detection method for wind turbine sensors. To better capture the spatio-temporal characteristics hidden in SCADA data, a multiscale spatio-temporal convolutional deep belief network (MSTCDBN) was developed to perform feature learning and classification to fulfill the sensor fault detection. A major superiority of the proposed method is that it can not only learn the spatial correlation information between several different variables but also capture the temporal characteristics of each variable. Furthermore, this method with multiscale learning capability can excavate interactive characteristics between variables at different scales of filters. A generic wind turbine benchmark model was used to evaluate the proposed approach. The comparative results demonstrate that the proposed method can significantly enhance the fault detection performance.

A de-ambiguous condition monitoring scheme for wind turbines using least squares generative adversarial networks

Blade icing is one of the main problems of wind turbines installed in cold climate regions, resulting in increasing power generation loss and maintenance costs. Traditional blade icing detection methods greatly rely on dedicated sensors, such as vibration and acoustic emission sensors, which require additional installation costs and even reduce reliability due to the degradation and failures of these sensors. To deal with this challenge, this paper aims to develop a cost-effective detection system based on the existing operation data collected from the supervisory control and data acquisition (SCADA) systems which are already equipped in large-scale wind turbines. Considering that SCADA data is essentially a multivariate time series with inherent non-stationary and multiscale temporal characteristics, a new wavelet-based multiscale long short-term memory network (WaveletLSTM) approach is proposed for wind turbine blade icing detection. The proposed method incorporates wavelet-based multiscale learning into the traditional LSTM architecture and can simultaneously learn global and local temporal features of multivariate SCADA signals, which improves fault detection ability. A real case study has shown that our proposed WaveletLSTM method achieved better detection performance than the existing methods.

Effective early warning of wind turbine failures is of great significance to reduce the operation and maintenance costs of wind farms and improve power generation efficiency. At present, most wind farms are installed with supervisory control and data acquisition (SCADA) system, and SCADA data contains a lot of hidden information, which can be used for fault early warning. This paper uses the generator temperature and gearbox oil temperature in the SCADA data as the entry point for fault warning. Firstly, the eXtreme gradient boosting (XGBoost) algorithm is used to establish the normal temperature regression prediction model of wind turbine components. Then, the residual between the predicted value and the actual value is calculated, and the change trend of the residual is monitored by the principle of exponentially weighted moving-average (EWMA) control chart. Finally, by setting an appropriate threshold, the variation trend of the residual is judged to determine the occurrence and development of the fault. This paper uses two fault detection methods: fixed threshold and dynamic threshold based on adaptive algorithm, and compares the advantages and disadvantages of the two methods. Based on the SCADA data of a wind farm in Inner Mongolia (China), this paper designs the fault early warning test of the wind turbine generator and gearbox. The experimental results show that for the generator, the fixed fault threshold method can give the fault alarm 3 hours in advance, while the dynamic fault threshold determination method can give fault alarm 4.25 hours in advance. For gearbox, the fixed fault threshold method can give the fault alarm 2 hours in advance, while the dynamic threshold fault diagnosis method can send out the fault alarm 2.75 hours in advance.

Fault diagnosis technology is key to the safe and stable operation of wind turbines. An effective fault diagnosis technology for wind turbines can quickly identify fault types to reduce the operation and maintenance costs of wind farms and improve power generation efficiency. Currently, most wind farms obtain operation and maintenance data via supervisory control and data acquisition (SCADA) systems, which contain rich information related to the operation characteristics of wind turbines. However, few SCADA systems provide fault diagnosis functionality. Support vector machines (SVMs) are a popular intelligence method in the fault diagnosis of wind turbines. SVM parameter selection is key for accurate model classification. The sparrow search algorithm (SSA) is a novel and highly efficient optimization method used to optimize the penalty factor and kernel function parameter of SVM in this paper and to construct the SSA-SVM wind turbine fault diagnosis model. Data are acquired from a wind farm SCADA system and form a faulting set after preprocessing and feature selection. Experiments show that the SSA-SVM diagnostic model effectively improves the accuracy of wind turbine fault diagnosis compared with the GS-SVM, GA-SVM and PSO-SVM models and has fast convergence speed and strong optimization ability. Moreover, the SSA-SVM diagnostic model can be used to diagnose faults in practical engineering applications.

The Supervisory Control and Data Acquisition (SCADA) system can provide significant information about the wind turbine health monitoring. However, the SCADA data contains enormous abnormal state data due to wind curtailment, equipment or sensor malfunction. In addition, complicated and constant changing operating conditions pose great challenges to effective and reliable fault detection. In this work, an automatic data cleaning and operating conditions classification approach is proposed. First, fundamental characteristics of different operating condition are analyzed according to control strategy, and data cleaning rules are constructed to remove abnormal state data. Then, change-point method is adopted to further clean residual abnormal state data that is stacked below the normal generation state data in the generator rotor speed-power curve. Second, the operating conditions classification criteria are constructed based on fundamental characteristics of different operating stages, and the classification parameter values are obtained adaptively according to data characteristics of different operating stages. Finally, the data cleaning and operating conditions classification approach is evaluated on the real-world SCADA dataset that was collected from a wind farm in East China. The results demonstrate that the proposed method can effectively distinguish normal power generation state data from abnormal power state data and realize operating conditions classification for different wind turbines automatically.

In the context of the rapid development of wind energy as a renewable resource, ensuring effective monitoring and maintenance of wind turbines is increasingly important. Despite the large volume of data collected by Supervisory Control and Data Acquisition (SCADA) systems, extracting key feature information to construct successful condition monitoring models remains a complex and challenging task. This study introduces a deep learning approach, called Convolutional Neural Network-based Support Vector Data Description (CNN-SVDD), aiming at effectively monitoring the long-term health status of wind turbines and providing timely warning for potential faults. The method employs Convolutional Neural Network (CNN) to deeply analyze time-series features and the correlations between sensors from SCADA data, capable of overcoming the challenges posed by environmental noise and changes in operational conditions. Subsequently, within the deep feature space learned by CNN, the SVDD classifier constructs a soft-boundary hypersphere. It allows for some samples to reside outside the hypersphere, thereby enhancing robustness against potential outliers in the SCADA data. Notably, the optimization processes of CNN and SVDD are linked, forming together an end-to-end process. The anomaly score generated by SVDD may act as an intelligent health indicator for assessing long-term performance degradation in wind turbines. Additionally, the radius of the hypersphere serves as a threshold for potential fault warnings. Extensive experiments on SCADA data acquired from multiple wind turbines over an 11-year period demonstrate the effectiveness of this method in early fault warning and long-term wind turbine performance degradation monitoring.

An operating condition recognition approach of wind turbine spindle is proposed based on supervisory control and data acquisition (SCADA) normal data drive. Firstly, the SCADA raw data of wind turbine under full working conditions are cleaned and feature extracted. Then the spindle speed is employed as the output parameter, and the single and combined normal behavior model of the wind turbine spindle is constructed sequentially with the pre-processed data, with the evaluation indexes selected as the optimal model. Finally, calculating the spindle operation status index according to the sliding window principle, ascertaining the threshold value for identifying the abnormal spindle operation status by the hypothesis of small probability event, analyzing the 2.5 MW wind turbine SCADA data from a domestic wind field as a sample, The results show that the fault warning time of the early warning model is 5.7 h ahead of the actual fault occurrence time, as well as the identification and early warning of abnormal wind turbine spindle operation without abnormal data or a priori knowledge of related faults.

Predicting the malfunction timing of wind turbines is essential for maintaining the high profitability of the wind power generation business. Machine learning methods have been studied using condition monitoring system data, such as vibration data, and supervisory control and data acquisition (SCADA) data, to detect and predict anomalies in wind turbines automatically. Autoencoder-based techniques have attracted significant interest in the detection or prediction of anomalies through unsupervised learning, in which the anomaly pattern is unknown. Although autoencoder-based techniques have been proven to detect anomalies effectively using relatively stable SCADA data, they perform poorly in the case of deteriorated SCADA data. In this letter, we propose a power-curve filtering method, which is a preprocessing technique used before the application of an autoencoder-based technique, to mitigate the dirtiness of SCADA data and improve the prediction performance of wind turbine degradation. We have evaluated its performance using SCADA data obtained from a real wind-farm.

As wind energy is widely used nowadays, it is more and more important to predict wind turbine downtime and organize maintenance just in time. Two types of data systems have been widely adopted for monitoring wind turbine condition: supervisory control and data acquisition (SCADA) and condition monitoring system (CMS). In this paper, a systematic framework is designed to combine CMS and SCADA data and assess drivetrain degradation. Advanced feature extraction techniques are applied to calculate health indicators. A pattern recognition algorithm is used to model baseline and actual condition, Self-organizing Map (SOM) and minimum quantization error (MQE) method is selected to do the degradation assessment. Eventually, the contribution of each component is calculated to achieve component-level fault localization. The approach is validated on a 2 MW wind turbine, where an incipient fault is detected before the wind turbine shuts down.

Gearbox oil temperature anomaly detection for wind turbine based on sparse Bayesian probability estimation

Short-term wind speed forecast for Urla wind power plant: A hybrid approach that couples weather research and forecasting model, weather patterns and SCADA data with comprehensive data preprocessing

The need for renewable energy has led to a fast increase of number of the wind turbines constructed each year. To monitor wind turbines farms, operating and maintenance managers need new effective and automatic tools compatible with a large number of wind turbines. This monitoring task is usually completed by Condition Monitoring System, but researches have been conducted on the utilization of SCADA (Supervisory Control And Data Acquisition) data for condition and predictive maintenance. This paper explains the difficulty of using this new source of information, and introduces the different techniques presented in the literature form 2001 up to 2014. Two classes of approaches can be identified: “internal” approaches using only data from one turbine, and “external” approaches relying on the comparison of one turbine to the other within the same farm. Both approaches have different pros and cons: internal approaches make use of the link between the components in the same turbine and so reduce the influence of operating conditions on fault indicators; external approaches make use of the correlation between SCADA variables measured on different turbines and can thus reduce the influence of wind conditions on the fault indicators. This paper sums up the latest available techniques and it shows that new areas of research can be explored with SCADA data. The obtained fault indicators still remain sensitive to the operating conditions and stochastic variations of the wind load. Combining advantages of the two approaches could reduce both influences.

As wind energy proliferates in onshore and offshore applications, it has become significantly important to predict wind turbine downtime and maintain operation uptime to ensure maximal yield. Two types of data systems have been widely adopted for monitoring turbine health condition: supervisory control and data acquisition (SCADA) and condition monitoring system (CMS). Provided that research and development have focused on advancing analytical techniques based on these systems independently, an intelligent model that associates information from both systems is necessary and beneficial. In this paper, a systematic framework is designed to integrate CMS and SCADA data and assess drivetrain degradation over itslifecycle. Information reference and advanced feature extraction techniques are employed to procure heterogeneous health indicators. A pattern recognition algorithm is used to model baseline behavior and measure deviation of current behavior, where a Self-organizing Map (SOM) and minimum quantization error (MQE) method is selected to achieve degradation assessment. Eventually, the computation and ranking of component contribution to the detected degradation offers component-level fault localization. When validated and automated by various applications, the approach is able to incorporate diverse data resources and output actionable information to advise predictive maintenance with precise fault information. Theapproach is validated on a 3 MW offshore turbine, where an incipient fault is detected well before existing system shuts down the unit. A radar chart is used to illustrate the fault localization result.