Blade Icing Research Articles

Deep fisher autoencoder combined with self-organizing map for visual industrial process monitoring

Feature extraction plays an important role in industrial process monitoring. Autoencoder and its deep framework, deep autoencoder, are used to extract latent features from complex data. However, the latent features extracted by autoencoder through unsupervised learning may not be useful for discriminative tasks. Fisher discriminant analysis (FDA) is another widely used supervised feature extraction technique that take full advantage of the Fisher criterion to enable the extracted discriminative features to maximize inter-class distance while minimizing intra-class distance. Drawing on FDA and autoencoder, this study proposes Fisher autoencoder (FAE) to extract discriminative features. FAE uses the Fisher criterion to guide the autoencoder in minimizing the reconstruction error while enabling the extracted features by the hidden layer to increase the separation between classes. We stack FAE to derive deep FAE (DFAE) for feature extraction, then we combine DFAE with self-organizing map (DFAE-SOM), which is a tool typically used in visualization for visual process monitoring. Tennessee Eastman process and an actual dataset of the blade icing of wind turbine are applied to test the performance of DFAE-SOM. The experiment demonstrates that DFAE increases the separation between classes more than DAE and other standard techniques. Therefore, DFAE is conducive to visualization and improves the accuracy of process monitoring.

Isogeometric analysis of ice accretion on wind turbine blades

For wind turbines operating in cold weather conditions, ice accretion is an established issue that remains an obstacle in effective turbine operation. While the aerodynamic performance of wind turbine blades with ice accretion has received considerable research attention, few studies have investigated the structural impact of blade ice accretion. This work proposes an adaptable projection-based method to superimpose complex ice configurations onto a baseline structure. The proposed approach provides an efficient methodology to include ice accretion in the high-fidelity isogeometric shell analysis of a realistic wind turbine blade. Linear vibration and nonlinear deflection analyses of the blade are performed for various ice configurations to demonstrate the impact of different ice accretion distributions on structural performance. These analyses indicate decreases in the blade natural frequencies and deflection under icing conditions. Such ice-induced changes clearly reveal the need for structural design consideration for turbines operating under icing conditions.

Diagnosis of Blade Icing Using Multiple Intelligent Algorithms

The icing problem of wind turbine blades in northern China has a serious impact on the normal and safe operation of the unit. In order to effectively predict the icing conditions of wind turbine blades, a deep fully connected neural network optimized by machine learning (ML) algorithms based on big data from the wind farm is proposed to diagnose the icing conditions of wind turbine blades. This study first uses the random forest model to reduce the features of the supervisory control and data acquisition (SCADA) data that affect blade icing, and then uses the K-nearest neighbor (KNN) algorithm to enhance the active power feature. The features after the random forest reduction and the active power mean square error (MSE) feature enhanced by the KNN algorithm are combined and used as the input of the fully connected neural network (FCNN) to perform and an empirical analysis for the diagnosis of blade icing. The simulation results show that the proposed model has better diagnostic accuracy than the ordinary back propagation (BP) neural network and other methods.

A Method to Detect Remote Alarm Based on Edge Computing in Wind Energy

Wind parks are usually located in places where few people live and is far from control center. Now, wind turbines often are embedded in so many sensors that operators can obtain their performance. But there are some important alarms difficultly to be monitored directly by sensors, such as blade icing which wind turbines in cold regions often encounter. These alarms involve many problems, such as energy losses, mechanical drawbacks and security. These complex alarms usually attribute to combination of many factors such as low temperature, wind speed, blade accelerator. Many scientists and researchers have proposed different methods or created new sensors to detect complex alarms. But these methods or facilities have different drawbacks that can cause low prediction accuracy, even high failure or error rate etc. This paper, taking blade icing for example, provides a practical method to detect remote alarm using monitoring data of wind turbine by machine learning algorithm without any new equipment. This method consists of four processes: preprocessing monitoring data, main features extraction, remote warning model construction, and deploying every sub-model as edge computing node. Firstly, getting massive historic monitoring data of wind turbine in wind park, and classifying; secondly, extracting main trait factors associating with blade icing; and then constructing detection model for each type wind turbine separately via machine learning according to massive historic monitoring data; finally, building edge computing nodes for every wind turbine to predict blade icing with model. This method has many advantages, such as low cost, high accuracy, clearly targeting, and early icing warning. Furtherly, with the help of edge computing, the generalization of machine learning model is improved. This study was completed with practical monitoring data of wind turbines in some park. Results show that the way to detect blade icing via machine learning is feasible and accuracy.

Load Calculation and Performance Analysis of Mega-MW Wind Turbine Based on Cloud Edge Computing

With the continuous development of megawatt-scale wind turbines, the impact of bladevibration on the dynamic stability and vibration of wind turbines cannot be ignored. However,there is currently no method capable of effectively analyzing the load and performance ofwind turbines. In order to improve the load calculation accuracy and performance of windturbines, this paper selects small base stations (SBS) as the research object of cloud edgecomputing, and explores the load analysis and performance analysis of megawatt-scale windturbines. In this paper, wind shear model, turbulence model, blade icing model andaerodynamic model of blades under icing conditions are established, and the effect of icing onthe dynamic load of blades of megawatt wind turbines is explored. It was found that the icinghad the greatest effect on the load in the direction of the blade array, followed by the load inthe swing direction, and the least in the torsional direction. In this paper, a load analysis isperformed on the blades of a megawatt wind turbine. When the blade speed increases, thefirst-order modal frequency deviation coefficient changes from -1.28% to 52.09%, and thethird-order modal frequency deviation coefficient changes from -0.77% to 48.67%. The speedof the wind wheel has a great influence on the first and third-order vibration frequencies of theblades of the wind turbine, but it does not have a great influence on the vibration frequenciesof the other steps.

Research on Effect of Icing Degree on Performance of NACA4412 Airfoil Wind Turbine

In order to study the effect of icing on the wind turbine blade tip speed ratio and wind energy utilization coefficient under working conditions, it is important to better understand the growth characteristics of wind turbine blade icing under natural conditions. In this paper, the icing test of the NACA4412 airfoil wind turbine was carried out using the natural low temperature wind turbine icing test system. An evaluation model of icing degree was established, and the influence of wind speed and icing degree on the performance parameters of wind turbines was compared and analyzed. It is shown that icing is mainly concentrated on the leading edge of the blade and the windward side. The icing on the leeward surface is concentrated near the leading edge of the blade. The amount of icing varies significantly along the direction of blade expansion, with the least amount of ice accumulation at 0.2R, followed by 0.6R, and most at 0.95R. At the beginning of the experiment, the icy thickness at the 0.2R, 0.6R and 0.95R locations increased at a rate of 12.03%, 8.87% and 6.00% respectively, and then the growth rate gradually decreased and stabilized. At the same wind speed, the blade tip speed ratio and wind energy utilization coefficient are inversely related to the change trend of the icing degree. When the blade icing degree is 0.82%, 1.49% and 2.73%, the maximum tip speed ratio is reduced by 7.1%, 14.1% and 21.2% and the maximum wind energy utilization coefficient is reduced by 36.3%, 51.2% and 61.6% respectively.

Prediction of Blade Icing Shape of Horizontal Axis Wind Turbine in Cold Climate

Prediction of Blade Icing Shape of Horizontal Axis Wind Turbine in Cold Climate

Intelligent wind turbine blade icing detection using supervisory control and data acquisition data and ensemble deep learning

AbstractIce accretion on wind turbine blades is one of the major faults affecting the operational safety and power generation efficiency of wind turbines. Current icing detection methods are based on either meteorological observing system or extra condition monitoring system. Compared with current methods, icing detection using the intrinsic supervisory control and data acquisition (SCADA) data of wind turbines has plenty of potential advantages, such as low cost, high stability, and early icing detection ability. However, there have not been deep investigations in this field at present. In this paper, a novel intelligent wind turbine blade icing detection method based on the wind turbine SCADA data is proposed. This method consists of three processes: SCADA data preprocessing, automatic feature extraction, and ensemble icing detection model construction. Specifically, deep autoencoders network is employed to learn multilevel fault features from the complex SCADA data adaptively. And the ensemble technique is utilized to make full use of all the extracted features from different hidden layers of the deep autoencoders network to build the ensemble icing detection model. The effectiveness of the proposed method is validated using the data collected from actual wind farms. The experimental results reveal that the proposed method is able to not only adaptively extract valuable fault features from the complex SCADA data, but also obtains higher detection accuracy and generalization capability compared with conventional machine learning models and individual deep learning model.

A Hybrid Model Based on Logistic Regression Algorithm and Extraction Algorithm Using Reward Extremum to Real-Time Detect Blade Icing Alarm

Generally, the blades of wind turbine installed in cold regions often encounter the situation of icing on their surface in winter, which affects the performance of wind turbine greatly and reduces power generation. Blade icing of wind turbine is a typical alarm related to multivariates, such as environment temperature, humidity, blade rotation and so on. To successfully achieve a supervised classification, it is necessary to extract the correlated features with blade icing. This paper proposes an extraction algorithm using reward extremum which can extract a group of variables correlated to blade icing from many monitoring variables of wind turbine, and they affect each other. This algorithm, using reward function which can measure the reward or value of alarm problem, can effectively distinguish the variables correlated to blade icing from the uncorrelated ones without knowing the mean of variables and the relations between them. Whether the blade of wind turbine is iced or not is a binary logic, so logistic regression algorithm is better choice to detect blade icing. Hence, this paper proposes a hybrid model using logistic regression algorithm and extraction algorithm using reward extremum, which can be built by machine learning with historic monitoring data. This model can real-time detect blade icing by real-time monitoring data. A correlation variable set can be obtained by the application of extraction algorithm using reward extremum on the historic monitoring data; and then the weights of the correlation variables in the set can be gained from the historic data by using logistic regression algorithm, so a blade icing detection pattern can be found using logistic regression algorithm; finally, whether the blade of wind turbine is iced in delay time or not can be real-time detected with the help of the hybrid model by real-time monitoring data. It is proved by wind turbine data that this hybrid model can work well not only to get correlation variable set but also to improve the generalization performance of logistic regression and achieve better prediction results. The performance of the model on blade icing alarm of wind turbine provides a new way to find out multivariate correlation to alarm from the mass historic monitoring data.

Uncertainty Quantification of Aerodynamic Icing Losses in Wind Turbine With Polynomial Chaos Expansion

Icing of wind turbine blades poses a challenge for the wind power industry in cold climate wind farms. It can lead to production losses of more than 10% of the annual energy production. Knowledge of how the production is affected by icing is of importance. Complicating this reality is the fact that even a small amount of uncertainty in the shape of the accreted ice may result in a large amount of uncertainty in the aerodynamic performance metrics. This paper presents a numerical approach using the technique of polynomial chaos expansion (PCE) to quantify icing uncertainty faster than traditional methods. Time-dependent bi-dimensional Reynolds-averaged Navier–Stokes computational fluid dynamics (RANS-CFD) simulations are considered to evaluate the aerodynamic characteristics at the chosen sample points. The boundary conditions are based on three-dimensional simulations of the rotor. This approach is applied to the NREL 5 MW reference wind turbine allowing to estimate the power loss range due to the leading-edge glaze ice, considering a radial section near the tip. The probability distribution function of the power loss is also assessed. The results of the section are nondimensionalized and assumed valid for the other radial sections. A correlation is found allowing to model the load loss with respect to the glaze ice horn height, as well as the corresponding probability distribution. Considering an equal chance for any of the ice profiles, load loss is estimated to be lower than 6.5% for the entire blade in half of the icing cases, while it could be roughly 4–6 times in the most severe icings.

A hybrid strategy combining minimized leading-edge electric-heating and superhydro-/ice-phobic surface coating for wind turbine icing mitigation

A hybrid anti-icing strategy that combines minimized electro-heating at the blade leading edge and a superhydro-/ice-phobic coating to cover the blade surface was explored for wind turbine icing mitigation. The experimental study was conducted in an Icing Research Tunnel available at Iowa State University (ISU-IRT) with a turbine blade model with DU91-W2-250 airfoil in the model cross-section exposed under different icing conditions. While a superhydro-/ice-phobic surface coating was used to cover the entire blade surface, a strip of electric heating film was used to wrap around the leading edge of the blade model. By using the superhydro-/ice-phobic coating to cover the entire blade surface, the hybrid strategy with the electric heating element covering only 5%–10% of the blade front surface would be able to keep the entire blade surface ice free under both rime and glaze icing conditions. In comparison to the conventional strategy to brutally heating the entire hydrophilic blade surface to keep the blade ice free, the hybrid strategy was found to be able to achieve the same anti-/de-icing performance with substantially less power consumption (i.e., up to ∼90% saving in the required power consumption), making it a very promising strategy for wind turbine icing mitigation.

Super-hydrophobicity effects on performance of a dynamic wind turbine blade element under yaw loads

Flow characteristics over a dynamic section of wind turbine blade in the presence of super-hydrophobic coatings were studied numerically using computational fluid dynamics method. Assuming a super-hydrophobic coating, blade icing can potentially be prevented. Fluids on super-hydrophobic surfaces are moving; so slip velocity exists on their walls. The dynamic motion was in sinusoidal mode that caused deep dynamic stall (DS) phenomenon. The occurrence of DS results in enlarged vortices with high strength of vorticity. These vortices cause overshoot in aerodynamic loads. By applying a super-hydrophobic surface on different locations of the airfoil, the DS vortex generation can be definitely affected. Investigation of these effects, is the aim of this paper. There is no study in the literature that analyzed DS in the presence of slip velocities at different locations. Thus, an SD7037 airfoil was simulated at Re≈4×104 using Transition-SST model. Results demonstrated that as the coating covered the leading edge, DS and vortex formation were postponed and the lift peak reduced about 10.6%; while for airfoil entirely covered by coating, the maximum lift value was augmented 14.6% and DS delayed. Applying slip boundary conditions at trailing edge as well as the pressure side had not considerable changes in cyclic loads.

Optimization design and simulationanalysis of composite material anti-/deicing component for wind turbine blade

Wind turbine blades are prone to icing under extreme working conditions, which will reduce the efficiency of wind power generation and seriously affect the service life of wind power system. Therefore, the anti-icing problem of wind turbine blades has become one of the hot topics in wind power technology research. It is of great practical significance to develop anti-icing/deicing technology for wind turbine blades for the safe and effective operation of wind power. In this paper, the numerical simulation optimization design for composite material wind turbine blades anti-/deicing component was studied, and the orthogonal optimization method was employed to optimize the parameters of the wind turbine blade to prevent ice formation. The optimization group parameter group is obtained. Through the wind turbine blade temperature distribution simulation, the correctness of the optimization results of the composite material anti-/deicing component was analyzed. The purpose of this paper is to optimize the heat transfer of anti-/deicing component for composite material wind turbine blade, and lay a theoretical and technical foundation for effectively solving the problem of wind turbine blade icing, which proposing a new composite wind turbine blade anti-/deicing scheme.

An accurate detection method for turbine icing issues using LSTM network

As one of the most widely applied clean energy resources, wind power generation is helpful for energy structure adjustment and sustainable development, whereas blade-icing failure affects the stability and benefits of wind farm. In consideration of efficiency improvement and safety hazard control, this paper proposed a failure detection model based on long short-term memory network. With the combination of turbine operation mechanism and data-driven method, the early warning of blade icing was implemented, which made it possible to take timely measures such as turning on the de-icing system at early moments of icing issues. The operational data collected by SCADA system were used as input for model, and progressive trend as well as hidden messages were learned from historical correlation at both temporal and characteristic scales. Consequently, the model achieved the fault detection and performed well on test dataset.

Effect on Blade Icing on Load of Horizontal Axis Wind Turbine in Cold Climate

Effect on Blade Icing on Load of Horizontal Axis Wind Turbine in Cold Climate

Evaluate the influence of icing on the rotor blades of a wind turbine with a horizontal axis to change its operating parameters

Currently the most commonly used wind turbines are three-blade turbines with a horizontal axis of rotation. The basic function of wind turbines is to convert the kinetic energy of the air into mechanical energy in the form of rotational motion of the rotor. This function is accomplished by rotor blades which, similarly to the wing of the aircraft, produce a lift force directed perpendicular to the movement of the air molecules flowing the blade profile, thus forcing the rotation of the rotor. The problem of operation of wind turbines is connected with high susceptibility to weather conditions. Under atmospheric conditions close to 0°C and less, and with high humidity, the wind turbine blades may be icing. Accumulation of ice on the wind turbine blades causes a violate air flow through the profile of the blades. This condition leads to a change in the wind speed range needed to generate the rotation of the wind turbine rotor and to reduce the efficiency of the turbine. To evaluate the effect of icing on turbine rotor blades, performed a numerical analysis of the model blade profile. The results of the numerical analysis allowed to evaluate the influence of icing of rotor blades on the change of operational parameters of wind turbine. The conclusions of the numerical simulation are presented in this article.

Effect of liquid water content on blade icing shape of horizontal axis wind turbine by numerical simulation

To research the law of the icing accretes on near the tip part of rotating blade of large-scale horizontal axis wind turbine (HAWT) influenced by liquid water content (LWC), the icing distribution on a HAWT rotor with rated power of 1.5 MW was simulated based on a Quasi-3D computation method. About 30% part length of blade from tip along span wise to blade root which are the most serious icing area was selected to research. Eight sections of this 30% part were decided and the ice distribution on each sections were simulated. Five kinds of LWC from 0.2 g/m3 to 1.4 g/m3 and two kinds of temperatures including ?6?C and ?18?C were selected. The medium volume droplet is 30 mm. Three kinds of icing time were selected to analyze the effects of icing time on ice accretion. The icing shape evaluate method was applied to quantitatively analyze the icing shape obtatined under different conditions. The results show that the icing shapes are all horn icing shape under the different LWC when the temperature is ?6?C. The icing shapes change from horn icing shape to streamline icing shape with LWC increasing under the temperature of ?18?C. The icing accretes on blade surface layer by layer with icing time increasing. The closer the section blade tip, the more icing accretes. This study can be as reference for the research on anti-icing and de-icing technologies for large-scale HAWT.

Quantitative detection method for icing of horizontal‐axis wind turbines

AbstractIcing seriously endangers the operational safety of wind turbines, and there has been a lack of research on the quantitative detection and early warning of the icing distribution on a blade. In order to address the present state and specific engineering problems of wind turbines, a method based on external sensor installation is proposed for quantitatively monitoring the icing distribution on a blade through numerical simulation and sensor detection technology. Field tests were performed on 1.5 and 2.0 MW wind turbines in icy weather. The detected distribution and thickness of the blade icing showed good agreement with the field results, which verifies the effectiveness of the detection method. The error between the detected and real ice thicknesses was approximately 20%, which breaks through the limitations of previous qualitative monitoring, and the error for specific ice formations on the blade was within the acceptable range. This new monitoring method can provide a reference for wind turbine icing detection and technical support for the efficient design and effective operation of icing protection facilities.

Wind turbine blade ice accretion: A correlation with nacelle ice accretion

This paper focuses on the meteorological icing phase of typical icing events to determine the factors that affect the ice accretion rate on a wind turbine blade. The linear mass density of ice on the nacelle of a 2 MW horizontal-axis wind turbine is estimated with a camera and is compared to the linear mass density of ice on a blade evaluated by a commercial blade ice detector (e.g. a fos4X ice detector). An automated image analysis software measures the dimension of the ice on a cylindrical tube located on the nacelle and calculates the linear mass density of the ice. The fos4X ice detector estimates the total ice load on the blade by measuring variations in the vibration frequencies of the blade. The total load is then divided by the blade length to obtain the linear density. On-site results were compared to simulations for validation. The ISO 12494 method for theoretical ice accretion on a cylindrical tube was adapted to simulate ice accretion on a rotating wind turbine blade. The main results show that wind speed and turbine rotor RPM directly influence the blade ice accretion rate. A correlation is made between blade and nacelle icing, with on-site results matching the simulation. In summary, for an operating turbine, as wind speed decreases, the rate of ice accretion on the blade increases relative to the ice accretion rate on the nacelle.

Linear and nonlinear features and machine learning for wind turbine blade ice detection and diagnosis

Abstract The mass of ice on wind turbines blades is one of the main problems that energy companies have in cold climates. This paper presents a novel approach to detect and classify ice thickness based on pattern recognition through guided ultrasonic waves and Machine Learning. To successfully achieve a supervised classification, it is necessary to employ a method that allows the correct extraction and selection of features of the ultrasonic signal. The main novelty in this work is that the approach considers four feature extraction methods to validate the results, grouped by linear (AutoRegressive (AR) and Principal Component Analysis) and nonlinear (nonlinear-AR eXogenous and Hierarchical Non-Linear Principal Component Analysis), and feature selection is done by Neighbourhood Component Analysis. A supervised classification was performed through Machine Learning with twenty classifiers such as Decision tree, Discriminant Analysis, Support Vector Machines, K-Nearest Neighbours, and Ensemble Classifiers. Finally, an evaluation of the classifiers was done in single frequency and multi-frequency modes, obtaining accurate results.