Convolutional neural network based unbalance detecting method of blades of wind driven generator

Abstract

The invention discloses a convolutional neural network based unbalance detecting method of blades of a wind driven generator. The method comprises the following steps: S1, acquiring generator torque signals, generator rotating speed signals, and vibrating acceleration signals of a tower frame of the wind driven generator in a balancing state and an unbalancing state; and converting into frequencysignals; S2, preprocessing the frequency signal data; and building a balancing state sample database and an unbalancing state sample database of the wind driven generator; S3, building a convolutionalneural network model; and processing the convolutional neural network model by training, parameter adjusting and verifying based on the classified sample database so as to obtain the trained convolutional neural network model; and S4, recognizing and determining the detecting data through the convolutional neural network model trained in step S3; and outputting balancing, pneumatic unbalancing and quality unbalancing results so as to finish unbalance detecting of the blades of the wind driven generator. According to the method, the fault properties of the blades can be self-adaptively caught,and thus the interference of environmental factors to the determining of the convolutional neural network model can be avoided.

Description

technical field [0001] The invention relates to the technical field of mechanical equipment fault monitoring, in particular to a convolutional neural network-based wind turbine blade imbalance detection method. Background technique [0002] In recent years, my country's wind power industry has developed rapidly, and the grid-connected capacity of wind turbines has continued to increase. In particular, the planning of offshore wind turbines has also been put on the agenda. Require. Wind turbines have been working in the field for a long time, and the operating environment is usually relatively harsh. This environment also causes a high failure rate of wind turbines, and the reasons for the failures are also various. The blades will be unbalanced due to wind and sand, icing and uneven mass distribution. In addition, the blades will also have cracks due to fatigue stress during long-term operation, resulting in aerodynamic imbalance. Especially the promotion of low wind speed ...

AI Technical Summary

Problems solved by technology

[0003] Commonly used wind turbine blade unbalance detection methods have the following difficulties to be overcome: 1. The operating conditions of wind turbine blades are complex, there is large noise interference, and it is difficult to extract robust fault features

Some wind turbine blade unbalance detection methods proposed in the early days mostly use certain techniques to extract fault features contained in vibration signals, generator currents or other variables, through Hilbert demodulation, dq coordinate transformation, order tracking analysis, etc. Various methods are used to analyze the stator current or rotor current and other signals in the frequency domain, which provides a reasonable method for wind turbine blade imbalance detection. However, the analyzed signal is very susceptible to noise interference, and the fault information is easily submerged, resulting in judgment errors. This method is quite difficult; 2. The imbalance that occurs during the operation of wind turbine blades includes mass imbalance and aerodynamic imbalance. How to analyze the two from the imbalance characteristics is a research problem. Analyzing the aerodynamic torque signal measured on the hub, using the sequential tracking method to extract the characteristic frequency of the blade angle fault, and using a certain method to distinguish each other with different harmonics on the rotor speed spectrum, which better explains the aerodynamic imbalance and rotor quality. However, the result is only verified under stable wind conditions, and cannot be applied to actual operating conditions with variable wind speeds; 3. Most of the different signal data collection during the operation of wind turbines is through sensors. Acquired, the layout and service life of the sensor and the accuracy of the collected signal are often the most important basic requirements. It has been proposed to install the patch-type optical fiber load sensor on the blade of the fan, and to carry out the signal data during the operation of the wind turbine. Collecting and using certain methods to monitor blade cracks provides a method for blade crack detection. However, this method ignores that the performance of the sensor itself is easily affected by environmental factors. Under the complex operating conditions of the wind turbine, the installation The SMD fiber optic load sensor on the blade is prone to damage, which affects the accuracy of the detection results and increases the detection cost

Images