Blade Icing Research Articles

Numerical analysis of blade icing influence on the dynamic response of an integrated offshore wind turbine

When a wind turbine is working in a cold and humid environment, icing may occur which lead to its performance reduction or even blades fracture. In this paper, a CFD-WTIA (Wind Turbine Integrated Analysis) coupled method is established to analyze the blade icing process and its influence on the overall dynamic performance of an integrated jacket-support offshore wind turbine. Firstly, motions of the blades are calculated by the WTIA method and used as input into CFD. Then, dispersed multi-phase model and melting-solidification model are used to simulate the icing growth phenomenon of three-dimensional blades. The k-ε turbulence model is used to calculate the aerodynamic performance before and after icing. Finally, the aerodynamic results after blade icing are returned to WTIA for integrated dynamic response acquisition. At the same time, the dynamic response of the wind turbine under the combined influence of ice and sea ice is analyzed. Results show that the blade ice-accretion increases linearly along the blade span-wise direction and is mainly concentrated on the leading edge of the blade. Lift and drag coefficients are seen deceased and increased respectively after icing. Power production, generator torque, rotor speed, as well as blade vibration are quantitatively studied. The methodology and findings of this paper can provide a good reference for the safety performance evaluation of an icing offshore wind turbine.

A Novel Deep Class-Imbalanced Semisupervised Model for Wind Turbine Blade Icing Detection.

Wind energy is of great importance for future energy development. In order to fully exploit wind energy, wind farms are often located at high latitudes, a practice that is accompanied by a high risk of icing. Traditional blade icing detection methods are usually based on manual inspection or external sensors/tools, but these techniques are limited by human expertise and additional costs. Model-based methods are highly dependent on prior domain knowledge and prone to misinterpretation. Data-driven approaches can offer promising solutions but require a massive amount of labeled training data, which are not generally available. In addition, the data collected for icing detection tend to be imbalanced because, most of the time, wind turbines operate under normal conditions. To address these challenges, this article presents a novel deep class-imbalanced semisupervised (DCISS) model for estimating blade icing conditions. DCISS integrates class-imbalanced and semisupervised learning (SSL) using a prototypical network that can rebalance features and measure the similarities between labeled and unlabeled samples. In addition, a channel calibration attention module is proposed to improve the ability to extract features from raw data. The proposed model has been evaluated using the blade icing datasets of three wind turbines. Compared to the classical anomaly detection and state-of-the-art SSL algorithms, DCISS shows significant advantages in terms of accuracy. Compared to five different class-imbalanced loss functions, the proposed DCISS is competitive. The generalization and practicability of the proposed model are further verified in the use case of online estimation.

Temporal Attention Convolutional Neural Network for Estimation of Icing Probability on Wind Turbine Blades

Wind farms are usually located in high-latitude areas, which bring a high risk of icing. Traditional methods of anti-blade-icing are limited by extra costs and potential damages to the original mechanical structure. Model-based methods are heavily dependent on mathematical models of the blade icing, which are prone to produce erroneous estimation. As data-driven models are better able to achieve competitive performances for the blade icing estimation, this article proposes a temporal attention-based convolutional neural network (TACNN). This novel data-driven model introduces a temporal attention module into a convolutional neural network, with the goal of determining the importance of sensors and timesteps and automatically identifying discriminative features from raw sensor data. Benchmark experiments on ten public datasets of multivariate time-series classification show competitive performance against the state-of-the-art methods. Compared with ten baseline networks and three widely used attention mechanisms, the TACNN shows significant advantages applying to three real-world datasets. These datasets are logged by the supervisory control and data acquisition system and contain operational and environmental measurements such as power and temperature. The ablation study and sensitivity study demonstrate the effectiveness of the key components of the TACNN. The practicability of the TACNN is further verified through online estimation testing.

Flexible nanocomposite electrothermal films based on carbon nanotubes and waterborne polyurethane with high reliability, stretchability and low-temperature performance for wind turbine blade deicing

Since the “net-zero” goal was proposed, wind power has been developed rapidly in recent years. Due to its large-scale installations in cold regions, the icing of wind turbine blades is currently one of the main limitations for efficient operation. In response to this issue, a highly reliable, flexible, stretchable, and lightweight in-situ deicing heating system based on the CNT/WPU nanocomposite films was developed. The CNT/WPU thin films were fabricated by the roll-to-roll gap-coating and hot-pressing process leading to as-prepared CNT/WPU film heaters have so outstanding reliability that they can withstand 100,000 times 240° bending with the resistance fluctuation of 3.17 %, and the repeated stretching of 2.0 % strain in the length direction without any impact on their temperature distribution, and even the weak impact with 10 % strain. The films may have promising potential applications in various curved surface heating with strict requirements on reliability in the future.

Icing diagnosis model for wind turbine blade based on feature optimization and 1D-convolutional neural network

Wind turbines located in high humidity and high altitude areas are often accompanied by blade icing, which adverses the operating efficiency and even causes safety accidents. Early identification of blade icing will help improve the operating efficiency of the wind turbine. This paper proposes an icing diagnosis method for wind turbine blades based on feature optimization and the one-dimensional convolutional neural network (1D-CNN). First, feature optimization is achieved by feature selection and feature reconstruction. The XGBoost algorithm is used to calculate the importance of each feature and select the features comprehensively that reflect blade icing. Second, the important features related to blade icing are reconstructed by using the deviation principle to extract the deviation information of features accurately when blades ice. Finally, the features screened by XGBoost and the reconstructed features are combined into the final feature set as the input of the 1D-CNN, which takes the temporal and spatial characteristics of data into account, to diagnose the icing state of blades. The method is validated on the data set collected from a real wind farm. The experimental results show that the proposed icing diagnosis method for wind turbine blades is superior to the traditional deep learning methods. It is favorable to improve the efficiency of wind turbine operation and maintenance.

Detecting Wind Turbine Blade Icing with a Multiscale Long Short-Term Memory Network

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.

Industrial process data visualization based on a deep enhanced t-distributed stochastic neighbor embedding neural network

PurposeThe purpose of this paper is to propose a approach for data visualization and industrial process monitoring.Design/methodology/approachA deep enhanced t-distributed stochastic neighbor embedding (DESNE) neural network is proposed for data visualization and process monitoring. The DESNE is composed of two deep neural networks: stacked variant auto-encoder (SVAE) and a deep label-guided t-stochastic neighbor embedding (DLSNE) neural network. In the DESNE network, SVAE extracts informative features of the raw data set, and then DLSNE projects the extracted features to a two dimensional graph.FindingsThe proposed DESNE is verified on the Tennessee Eastman process and a real data set of blade icing of wind turbines. The results indicate that DESNE outperforms some visualization methods in process monitoring.Originality/valueThis paper has significant originality. A stacked variant auto-encoder is proposed for feature extraction. The stacked variant auto-encoder can improve the separation among classes. A deep label-guided t-SNE is proposed for visualization. A novel visualization-based process monitoring method is proposed.

Adaptive envelope protection control of wind turbines under varying operational conditions

This study introduces a new Envelope Protection System (EPS) algorithm for wind turbines. The algorithm is adaptive to turbine-changing operational conditions and can effectively reduce turbine excessive/ultimate loads. Through an adaptive neural network, the proposed algorithm continuously monitors instantaneous wind and turbine states. Simultaneously, it predicts the near future response of the turbine load and detects its future crossing with a predefined safe envelope limit by comparing the actual wind speed to a theoretically estimated wind speed. When required, a protection action is applied based on the comparison to keep the turbine load response within the safe limit. In this paper, the thrust force is used as the critical load and is chosen as the limit parameter. Simulations are carried out using the MS (Mustafa Sahin) Bladed Wind Turbine Simulation Model for the National Renewable Energy Laboratory (NREL) 5 MW turbine under normal turbulent winds with different mean values. Simulations show that the EPS algorithm adapts to varying operational conditions such as changes in turbine operating point in the below rated, transition, and above rated regions, as well as rotor blade icing and successfully reduces the excessive thrust forces. Performance analyses indicate that, for keeping the thrust force within the limit, the proposed EPS algorithm reduces the thrust force by 98.89%, 98.43%, 99.26% relative to standard baseline controls in the aforementioned regions, respectively and by 99.61% under blade icing. Also, the mean value and the fluctuations of thrust force are reduced up to 5.52% and 68.7%, respectively. Depending on the operating region, the mean power decreases up to 2.07% or increases up to 1.21%, while power fluctuations decrease up to 30.97%.

Blade icing detection of wind turbine based on multi-feature and multi-classifier fusion

Wind farms are usually located in high altitude areas with a high probability of ice occurrence. Blade icing has the potential to result in unexpected mechanical failures and downtimes. In order to avoid these problems, the priority we need to do is to detect blade icing accurately. For this purpose, a novel icing detection method based on multi-feature and multi-classifier fusion is proposed in this paper. Firstly, multi-feature composed of basic features and statistical features are extracted from the operational data. Significant features are then extracted by utilizing Light Gradient Boosting Machine. Secondly, a multi-classifier fusion approach is employed to build an fusion model, which aims to obtain a much more accurate estimation for the icing state. Overall, the proposed method in this paper can achieve more accurate detection on blade icing, compared with other models. This will minimize false alarms, helping wind farms manage the operations more efficiently.

Wind turbine ice detection using hyperspectral imaging

Wind energy has been playing a pivot role in replacing the traditional energy sources. This emerging paradigm has proved itself as a good candidate among all renewable energy sources. Although the exponential growth of the wind industry, wind turbines still suffer from blade icing especially in cold regions. Blade icing disturbs aerodynamic performance and results in power losses, safety risks, mechanical and electrical breakdowns, measurement, and control faults. Anti-icing and de-icing techniques mitigate these adverse effects. It is mandatory to rigorously evaluate the meteorological operating conditions during the assessment phase to determine the need and advantages of installing an anti-icing or a de-icing system. Moreover, this diagnostic is also essential during the operation to detect icing, prevent failure, and enhance production. Different ice detection methods, such as double anemometry, vane, relative humidity, and dew point, are used. These techniques have few drawbacks that can be overcome using hyperspectral imaging. This paper offers an overview of icing detection technologies and explores spectroscopy imaging applications for detecting ice accretion in wind farms. This study describes the application of this non-destructive and fast monitoring technique in remote sensing of icing incident on a wind turbine blade. This paper outlines the experimental approach conducted on a blade sample with an ice-covered portion. The icing model, on which this detection method is based, is designed, simulated, and confirmed to acquire enhanced blade icing knowledge. The hyperspectral imaging validation results for icing occurrence detection in their initial development phases are satisfactory. The experimental findings of this technique reveal that the accuracy and precision of blade icing detection are considerably enhanced.

Prediction of wind turbine blade icing fault based on selective deep ensemble model

In advance accurate prediction of wind turbine blade icing fault is of fundamental importance. Deep learning is the mainstream prediction technique while existing research about the prediction of wind turbine blade icing fault has primarily derived a single deep learning model. This study introduces the group method of data handling (GMDH) technique and proposes a GMDH-based selective deep ensemble (GSDE) model. First, the model combines the convolution neural network (CNN) with recurrent neural network (RNN), long short-term memory (LSTM), and gated recurrent unit (GRU) to construct CNN–RNN, CNN–LSTM, and CNN–GRU, respectively. Together with CNN, four cost-sensitive deep neural networks based on focal loss are formed and used as the basic prediction models. Second, a series of training sets are constructed by the Chi-square test. Four different basic prediction models are trained on each training set, and the prediction results of all base predictors are obtained. Third, the GMDH technique is applied to the cost-sensitive selective deep ensemble for final prediction results. Experiments are conducted to deeply verify the prediction performance of the GSDE model on two wind turbine datasets collected by the supervisory control and data acquisition (SCADA) system. Results show that the proposed model outperforms five existing ensemble models and five single deep learning models.

Wind Turbine Blade Icing Diagnosis Using Convolutional LSTM-GRU With Improved African Vultures Optimization

Wind farms are usually located in plateau mountains and northern coastal areas, bringing a high probability of blade icing. Blade icing even leads to blade cracks and turbine collapse. Traditional methods of blade icing diagnosis increase operating costs and have the potential risk of damaging the original mechanical structure. A data-driven model based on a novel convolutional recurrent neural network is proposed in this paper. The method can effectively extract hidden features for accurate icing diagnosis. The hyperparameters of the proposed model are optimized by the improved African Vultures optimization algorithm (IAVOA). To alleviate the critical data imbalance, the Adaptive Synthetic (ADASYN) is used to oversample the minority classes of icing status. In comparison to the state-of-art classification methods, the proposed method illustrates the outstanding effectiveness in blade icing diagnosis using the sensor data from Supervisory Control and Data Acquisition (SCADA) systems. The effectiveness analysis of variables, ablation study, and sensitivity analysis validate the performance of the proposed method.

A Novel Ellipsoidal Semisupervised Extreme Learning Machine Algorithm and Its Application in Wind Turbine Blade Icing Fault Detection

The conventional semi-supervised extreme learning machine (SS-ELM) algorithm can provide a solution to the lack of labeled samples in wind turbine blade icing fault detection, but its performance is limited by the irrationality of the spherical nearest neighbor graph (SNNG) calculation strategy. To solve this problem, a novel ellipsoidal semi-supervised extreme learning machine (ESS-ELM) algorithm is proposed in this paper and applied to wind turbine blade icing fault detection. In this study, we creatively propose a novel ellipsoidal nearest neighbor graph (ENNG) calculation strategy that considers the distribution information of the labeled samples to construct the ESS-ELM algorithm. Different from the conventional SNNG, the ENNG can adaptively assign corresponding calculation weights to each feature in the data space according to the Fisher criterion, which allows it to better match the smoothness assumption of the manifold regularization learning framework. In addition, the two adjustable parameters of the enhancement factor and the degradation factor are ingeniously introduced into the ENNG, which further ensure the applicability and security of the proposed ESS-ELM. The superiority of the ESS-ELM algorithm is verified by extensive benchmark industrial fault datasets and the real-world icing dataset of two wind turbines. It is demonstrated that the proposed ESS-ELM algorithm achieves better performance than the ELM, its existing variants, and some state-of-the-art wind turbine blade icing detection methods.

Wind Turbine Blade Icing Fault Prediction Based on SCADA Data by XGBoost

To deal with the icing problem of wind turbine blades, the traditional classification methods are introduced firstly in this paper, and the XGBoost model based on monitoring and data acquisition (SCADA) system is introduced to estimate the icing conditions of blades. Meanwhile, the generation process of the XGBoost model is introduced in detail. Finally, the superiority of the XGBoost model is verified by experiments. The results show that XGBoost has higher precision and efficiency than other methods.

Wind Turbine Blade Icing Detection Based on Random Forest

In this paper, a multimodal random forest method is proposed to solve the problem that the traditional linear method has low accuracy in predicting the icing of generator blade. This paper uses the random forest algorithm to model and analyze the icing event of wind turbine blades, and describes in detail the process of using the C4.5 algorithm to generate a decision tree, and then randomly selecting samples and sample features to train to generate a random forest. The effectiveness of the method in this paper is experimentally verified by SCADA data, and the results show that the multi-modal random forest algorithm reduces the error rate to 1.97% in wind turbine blade icing prediction, which is more accurate than the traditional method.

Enhancing the generalizability of predictive models with synergy of data and physics

Wind farms require predictive models for predictive maintenance. There is a need to predict values of non-observable parameters beyond ranges reflected in available data. A predictive model developed for one machine many not perform well for another similar machine. This is usually due to a lack of generalizability of data-driven models. To increase the generalizability of predictive models, this research integrates data mining with first-principle knowledge. Physics-based principles are combined with machine learning algorithms through feature engineering, strong rules and divide-and-conquer. The proposed synergy concept is illustrated with a wind turbine blade icing prediction and achieves significant predictive accuracy across different turbines. The proposed process should be widely accepted by wind energy predictive maintenance practitioners because of its simplicity and efficiency. Furthermore, the testing scores of KNN, CART and DNN algorithms are increased by 44.78%, 32.72% and 9.13%, respectively, with our proposed process. We demonstrate the importance of embedding physical principles within the machine learning process, and also highlight an important point that the need for more complex machine learning algorithms in industrial big data mining is often much less than it is in other applications, making it essential to incorporate physics and follow the ‘less is more’ philosophy.

Simulation and Experimental Study on the Ultrasonic Micro-Vibration De-Icing Method for Wind Turbine Blades

In cold and humid regions, ice accretion sometimes develops on the blades of wind turbines. Blade icing reduces the power generation of the wind turbine and affects the safe operation of the wind farm. For this paper, ultrasonic micro-vibration was researched as an effective de-icing method to remove ice from the wind turbine blade surface and improve the efficiency of wind turbine power generation. A blade segment with NACA0018 airfoil and the hollow structure at the leading edge was designed. The modal analysis of the blade was simulated by ANSYS, and the de-icing vibration mode was selected. Based on the simulation results, the blade segment sample with PZT patches was machined, and its natural frequencies were measured with an impedance analyzer. A return-flow icing wind tunnel system, and a device used to measure the adhesive strength of ice covering the airfoil blade, were designed and manufactured. The experiments on the adhesive strength of the ice were carried out under the excitation of the ultrasonic vibration. The experimental results show that the adhesive strength of the ice, which was generated under the dynamic flow field condition, was lower than the ice generated by water under the static flow field condition. Under the excitation of the ultrasonic vibration, the adhesive strength of the ice decreased. When the excitation frequency was 21.228 kHz, the adhesive strength was the lowest, which was 0.084 MPa. These research findings lay the theoretical and experimental foundations for researching in-depth the application of the ultrasonic de-icing technology to wind turbines.

Rapid warning of wind turbine blade icing based on MIV-tSNE-RNN

Rapid warning of wind turbine blade icing based on MIV-tSNE-RNN

Unsupervised transfer learning for condition monitoring of a fleet of wind turbines

The condition monitoring and health status prediction of a fleet of wind turbines are essential for the safety of wind turbines. At present, the Supervisory Control And Data Acquisition (SCADA) system has been widely used in wind turbines, which can monitor and collect various physical information and sensor information of wind turbines in real-time. Due to the fact that the amount of data obtained by SCADA systems is extremely large, developing an intelligent decision-making system based on deep learning is a very valuable research. Therefore, this paper is committed to exploring a health status prediction algorithm of wind turbines based on deep learning and SCADA systems. However, yet in actual industrial applications, it is very time-consuming and expensive to obtain a large amount of labeled data. In addition, as failures rarely occur, there is a serious sample imbalance problem in the datasets. More importantly, due to the difference in working environment and physical parameters, there are significant differences in the feature distribution of different wind turbines data, which lead to a significant drop in the performance of the deep learning model on unknown wind turbines.
 Therefore, we propose an unsupervised transfer learning algorithm based on Generative Adversarial Networks for wind turbine health status prediction (WT-GAN). WT-GAN can not only remove the domain shift between wind turbines, but also it is an unsupervised learning method. This practically means that only the unlabeled data for the target domain is required, which solves the problem of labeling data. In order to evaluate the effectiveness of WT-GAN on the condition monitoring of a fleet of wind turbines, we apply this method to one dataset about blade icing detection of wind turbines. The experimental results prove that the proposed method can predict the health of the wind turbine well. In addition, it can significantly reduce the domain shift among different wind turbines, thereby achieving excellent performance on unknown wind turbines.

A PCC-Ensemble-TCN model for wind turbine icing detection using class-imbalanced and label-missing SCADA data

Blade icing problems are ubiquitous for wind turbines located in cold climate zones. Data-driven indirect icing detection methods based on supervisory control and data acquisition system have shown strong potential recently. However, the supervisory control and data acquisition data is annotated through manual observation, which will cause the data between normal condition and icing condition to be unlabeled. In addition, the amount of normal data is far more than icing data. The above two issues restrict the performance of most current data-driven models. In order to solve the label missing problem, this article proposes a Pearson correlation coefficient–based algorithm for measuring the degree of blade icing, which calculates the similarity between the unlabeled data and the icing data as its label. Aiming at the class-imbalance problem, this article constructs multiple class-balanced subsets from the original dataset by under-sampling the normal data. Temporal convolutional networks are trained to extract features and make predictions on each subset. The final prediction result is obtained by ensembling the prediction results of all temporal convolutional network models. The proposed model is validated using the actual supervisory control and data acquisition data collected from a wind farm in northern China, and the results indicate that ensuring the consecutiveness and class-balance of the data are quite advantageous for improving the detection accuracy.