Kusiak.qxd_Outsourcing_book_temp.qxd 02/03/2010 16:07 Page 67
Short-term Prediction of Windfarm Power – A Data-driven Approach
Virtual Turbines non-controllable parameter considered in this section, and three past
Considering the fact that wind speeds and wind turbine performance states are used in virtual models. Two controllable parameters, blade
vary across different turbine locations at a windfarm, the question pitch angle and generator torque, and their two past states are
arises as to whether a generalised model (referred to in this article as a
virtual model) of a wind turbine could be developed. Such a virtual
model has been developed based on SCADA data collected at 30
The ability of a wind turbine
wind turbines.
to extract power comprises
As a wind turbine is a complex system, two aspects of a
three factors: the measured
wind turbine are reported: the power output and the rotor speed. wind speed, the power curve
However, the methodology presented here can be applied to modelling
of the turbine and the ability
many aspects of a wind turbine. Predicting the power output
demonstrates the ability of the virtual model to improve the of the machine to handle
performance of a wind turbine, while predicting the rotor speed points
wind fluctuations.
to the utility of the virtual model to improve the lifetime of turbine
components, e.g. the gearbox.
selected. The two past states of performance parameters, power
Controllable parameters and non-controllable parameters constitute output and rotor speed, are selected.
inputs to the virtual models, while the performance parameters create
the outputs predicted by the models. The impact of input parameters The performance of the models built by six different algorithms,
on the output (performance) parameter varies. Some insignificant specifically random forest, neural network, boosting tree, SVM,
parameters are easy to eliminate based on the domain knowledge. The generalised additive and the kNN, are reported in Table 1.
impact of input parameters on the performance parameter is ranked by
data-mining algorithms. Based on the data in Table 1, the neural network performed best
among the six algorithms tested. The neural network algorithm is used
Figure 3 shows the final feature selection for the interest aspects: to train a virtual model to predict power output and rotor speed. Here,
power output and rotor speed. Data for three kinds of input 30 different NN were trained, and the best model was selected to be a
parameters and their past states are included. Wind speed is the only virtual model.
a member of FUCHS PETROLUB AG
We make the wind win –
with special lubricants for
wind power plants
FUCHS WINDPOWER DIVISION is a comprehensive
service provider to the wind power industry:
Individual selection and evaluation of lubricants
Full documentation with gearbox-condition reports
Reliable gear oil reports with high-accuracy oil analyses
Get in contact with your nearest FUCHS subsidiary:
www.fuchs-oil.com / worldwide.html
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88 |
Page 89 |
Page 90 |
Page 91 |
Page 92 |
Page 93 |
Page 94 |
Page 95 |
Page 96 |
Page 97 |
Page 98 |
Page 99 |
Page 100 |
Page 101 |
Page 102 |
Page 103 |
Page 104 |
Page 105 |
Page 106 |
Page 107 |
Page 108