Search

Condition monitoring systems (CMS) are essential to reduce costs in the wind energy sector. This paper proposes a method based on Support Vector Machine (SVM) to detect rotor mass imbalance for a multi-class imbalance problem, using the estimated speed as an input variable, obtained through a combination of electrical quantities (currents and voltages). Moreover, it is sought to obtain the magnitude of the rotor mass imbalance. With the aid of statistical tools, intermediate classes can be estimated, other than the ones proposed for the SVM. Besides, if the azimuth position is provided, the angular position of the mass imbalance can be also obtained. A 1.5 MW three-bladed wind turbine model with a permanent magnet synchronous generator, was considered, and a database was built numerically using the software Turbsim, FAST, and Simulink. From the database, the Power Spectral Density (PSD) technique was used to transform the input data from the time to the frequency domain. Then, the SVM algorithm and statistical analysis were used to classify the magnitude and the angular position of the imbalance. Different scenarios of mass imbalance were tested under different wind speeds and turbulence intensities. The results demonstrate the satisfactory performance of the proposed method. [Display omitted] • This paper proposes a method to detect rotor mass imbalance in wind turbines. • Support Vector Machine (SVM) is used to classify imbalance from electrical signals. • The method locates the equivalent angular position of the rotor mass imbalance. • The method estimates intermediate imbalance classes from the SVM classes. • The impact of the wind turbulence intensity on the SVM performance was investigated. [ABSTRACT FROM AUTHOR]

In this paper, three kinds of models, including support vector machine-firefly algorithm (SVM-FFA), copula-base nonlinear quantile regression (CNQR) and empirical models were developed for daily diffuse radiation (H d) estimation. The meteorological data during 1981–2000 and 2001–2010 of Lhasa, Urumqi, Beijing and Wuhan in China were used for model training and validation, respectively. Five combinations of meteorological data: (a) clearness index (K t); (b) sunshine ratio (S); (c) K t and S ; (d) K t , S and average temperature (T a); (e) K t , S , T a and average relative humidity, were considered for simulation. The results showed that for the training phases, SVM-FFA outperformed the corresponding models while empirical models performed slightly better than corresponding CNQR models. For validation phases, CNQR and SVM-FFA models performed much better than empirical models. Compared CNQR and SVM-FFA, SVM-FFA performed slightly better than CNQR models with average MABE decreased by 0.67% (0.01 MJm−2d−1) and average R2 increased by 0.43% (0.004). For the training time, SVM-FFA (1.68 s) showed less computational costs than CNQR (6.68 s); but the parameter optimization time of SVM-FFA (4.9 × 105) were 105 times as much as CNQR. Thus, the overall computational costs of SVM-FFA during training phases were much higher than CNQR. Considering the trade-off between accuracy and computational costs, CNQR were highly recommended for the daily H d estimation. • CNQR, SVM and empirical models were developed for daily diffuse radiation. • The models were trained and compared using data from China as a case study. • SVM and CNQR models showed comparable estimation accuracy. • CNQR showed much more efficient than SVM models. [ABSTRACT FROM AUTHOR]

Abstract Effective condition monitoring and fault diagnosis of wind turbines are crucial for avoiding serious damages to wind turbines. The supervisory control and data acquisition (SCADA) system of a wind turbine provides valuable insights into turbine performance. In order to make full use of such valuable information, this paper investigates fault diagnosis of wind turbines by using Gaussian process classifiers (GPC) to the operational data collected from the SCADA system. Both real-time and predictive fault diagnosis were considered. As an alternative to the support vector machine (SVM) technique, the GPC possesses the capability of providing probabilistic information about the fault types, which is valuable for making maintenance plan in real practice. The comparison results show that the GPC method is able to provide more accurate fault diagnosis results than the SVM technique on average. Highlights • Suggest a new model based on Gaussian process classification for wind turbine diagnosis. • The suggested model is computationally more efficient than the support vector machine. • The suggested model can take the operational information into account. • The comparison results favor the new model. [ABSTRACT FROM AUTHOR]

This paper proposes a novel and hybrid intelligent algorithms to directly modelling prediction intervals (PIs), as an accurate, optimum, reliable and high efficient wind power generation prediction intervals (PIs) are developed by using extreme learning machines (ELM) and self-adaptive evolutionary extreme learning machines (SAEELM). Given significant of uncertainties existed in the wind power generation, SAEELM is the state-of-the-art technology to estimate and quantify the potential uncertainties that may result in risk facing the power system planning, economical operation, and control. In SAEELM, a single hidden layer extreme learning machine is constructed, where the output weight matrix is optimised by using the self-adaptive differential evolution (DE) optimisation method. Also, selecting and adjusting the control parameters and generation strategies involved in differential evolution algorithm to minimises the developed objective cost function. Different case studies using Australian real wind farms have been conducted and analysed. By comparing the statistical analysis and results to other models and methods, e.g. artificial neural networks (ANN), support vector machines (SVM), and Bootstrap, therefore, the proposed approach is an efficient, accurate, robust, and reliable for dealing with uncertainties involved in the integrated power systems, and generation of high-quality PIs. Moreover, the proposed SAEELM based algorithm has a better generalisation than other methods and has a high potential for practical applications. [ABSTRACT FROM AUTHOR]

This paper evaluates the accuracy of Support Vector Machine (SVM), Artificial Neural Network (ANN) and empirical solar radiation models with different combination of input parameters. The parameters include month, latitude, longitude, bright sunshine hours, day length, relative humidity, maximum and minimum temperature. The models are evaluated based on statistical measures. Four new empirical models are introduced and validated with experimental data. This work is focused on the prediction of monthly mean daily global solar radiation (GSR) for different cities in India with most influencing input parameters identified using Waikato Environment for Knowledge Analysis (WEKA) software. WEKA identifies month, latitude, maximum temperature and bright sunshine hours as the most influencing and relative humidity as the least influencing input parameter. SVM model with most influencing input parameter performs better than ANN and Empirical models. Exclusion of relative humidity does not affect the prediction accuracy. Therefore this work reduces the dimensionality of the data and improves the prediction accuracy. This work also attempts in assessing the solar energy potential of smart cities of Tamil Nadu, India using the SVM model. The predicted annual GSR varies from 17 to 22 MJ/m2/day which is precise enough for a wide range of solar applications. [ABSTRACT FROM AUTHOR]

Abstract: Solar radiation is the principal and fundamental energy for many physical, chemical and biological processes. However, it is measured at a very limited number of meteorological stations in the world. This paper presented the methods of monthly mean daily solar radiation estimation using support vector machines (SVMs), which is a relatively new machine learning algorithm based on the statistical learning theory. The main objective of this paper was to examine the feasibility of SVMs in estimating monthly solar radiation using air temperatures. Measured long-term monthly air temperatures including maximum and minimum temperatures (T max and T min, respectively) were gathered and analyzed at Chongqing meteorological station, China. Seven combinations of air temperatures, namely, (1) T max, (2) T min, (3) T max − T min, (4) T max and T min, (5) T max and T max − T min, (6) T min and T max − T min, and (7) T max, T min, and T max − T min, were served as input features for SVM models. Three equations including linear, polynomial, and radial basis function were used as kernel functions. The performances were evaluated using root mean square error (RMSE), relative root mean square error (RRMSE), Nash-Sutcliffe (NSE), and determination coefficient (R 2). The developed SVM models were also compared with several empirical temperature-based models. Comparison analyses showed that the newly developed SVM model using T max and T min with polynomial kernel function performed better than other SVM models and empirical methods with highest NSE of 0.999, R 2 of 0.969, lowest RMSE of 0.833 MJ m−2 and RRMSE of 9.00%. The results showed that the SVM methodology may be a promising alternative to the traditional approaches for predicting solar radiation where the records of air temperatures are available. [Copyright &y& Elsevier]