Health assessment of wind turbine based on laplacian eigenmaps

Power curve modelling for wind turbines and wind power prediction based on mixed Richards model

“…However, as shown in the right partial enlargement figure (1), the fitting results of the ninth-order polynomial fluctuate in a certain power range, some fitting curves of the 5PLF and threefold mixed Richards models deviate from the sample points, while the fitting curves of the eighth-order polynomial and two-fold mixed Richards models are the most stable, and the fitting results of the eighthorder polynomial are most consistent with the sample points, followed by the double Richards model. But from the partial enlarged figure (2), it can be found that the fitting curves of the eighth-order and ninthorder polynomial models fluctuate to some extent, and the two-fold and three-fold mixed Richards models are most consistent with the data sample points, that is, the fitting effect is the best. To sum up, the same fitting results of different models based on the measured data of 1# wind farm can be obtained in the end, that is, the fitting effect of two-fold Richards model is the best, and the over-fitting phenomenon is effectively avoided.…”

“…However, from the partial enlarged figure (1) on the right side, it can be seen that the fitting curves of eighth-order and ninth-order polynomial models fluctuate greatly, while the fitting curves of two-fold Richards, three-fold Richards, 4PLF and 5PLF models are relatively stable, and the fluctuation of two-fold Richards model is the least. And from the partial enlargement figure (2), it can be found that when the wind speed exceeds the rated wind speed, the fitting curve of the high-order polynomial model will still fluctuate within a certain power range with the change of the sample points, and there is an overfitting, and the fitting curve of the 4PLF model partially deviates from the sample points. In conclusion, based on the measured data of 1# wind farm, the fitting effect of mixed Richards and Logistic function models are better than that of high-order polynomial model, and the fitting effect of two-fold Richards model is the best, and the phenomenon of overfitting is avoided.…”

“…WPC modelling of wind turbines is one of the important aspects of wind power prediction, and to improve the accuracy of power prediction, we need to continuously enhance the precision of wind power curve (WPC) modelling. [2][3][4][5] The standard WPC of wind turbines is a curve describing the relationship between wind speed and output power per unit time. Due to the influence of turbulence intensity, air density, wind shear and other environmental factors, the WPC under actual conditions is different from the standard power curve provided by the manufacturer.…”

Abnormal data cleaning of wind turbine power curve using Bayesian change point-quartile combined algorithm

“…To identify and delete abnormal data, based on the local outliers factor of Gaussian kernel density estimation, a local outlier detection method is used in ref. 11. Image recognition technique is also widely used, when using this method, wind speed-power data should be converted into a binary image firstly, then the characteristics of normal and abnormal data can be obtained by comparing their differences, and finally according to the characteristics of these data, the identification and deleting of abnormal values can be completed.…”