TWI781850B - Intelligent networked wind power generation fault diagnosis and detection system and detection method - Google Patents

Abstract

A fault diagnosis and detection method for smart networked wind power generation, which uses a three-axis vibration sensor to sense the vibration state of the three-axis direction when the multi-blade wind turbine rotates, and then captures the three-axis signal through a high-speed data acquisition card , and then the three-axis signal is filtered through a signal processing unit, and finally the filtered three-axis signal is analyzed through a computer operation system to analyze the type of fault.

Description

Fault diagnosis and detection system and detection method for intelligent networked wind power generation

The present invention relates to a smart networked wind power generation fault diagnosis and detection system and its method, in particular to a smart networked wind power generation fault diagnosis and recognition based on different vibration signals caused by blade rotation angle abnormality, blade surface damage and blade breakage Diagnostic detection system and detection method thereof.

The wind turbine is exposed to the outdoors for a long time, and after a long time of operation, the probability of failures such as component aging and mechanical wear is relatively high. And according to the past data statistics of the Global Wind Energy Council (GWEC), under long-term operation of wind turbines, the surface of the blades is damaged and broken due to the harsh external weather environment, or the blade rotation angle system is mechanically damaged due to long-term operation. Fatigue caused the highest failure problems caused by abnormalities.

However, wind turbines are usually erected in open areas, so wind turbines face a greater probability of being struck by lightning, and according to the statistics of the Global Wind Energy Association, the main component of the wind turbine that is more likely to be struck by lightning is the blade, according to the statistical report. , the damage probability of the leading edge and the trailing edge of the blade is the highest.

Therefore, when the blade rotation angle system of the wind turbine is abnormal or the blade surface is damaged or broken, it will seriously damage the blade structure and surface materials, resulting in expensive maintenance costs, and it must be shut down for a long time to find the problem Repairs will result in a reduction in the efficiency of green energy power generation. Therefore, when there is an abnormality in the blade and rotation angle system of the wind turbine, how to find out the cause of the fault type as early as possible, so as to perform replacement and maintenance work early, so that the wind turbine can operate safely and provide stable power supply. This is an urgent need for the industry and academia. one of the problems to be solved.

The purpose of the present invention is to provide a smart networked wind power generation fault diagnosis and detection system and its detection method, which can diagnose and identify the types of blade faults by means of different vibration signals caused by blade rotation angle abnormality, blade surface damage and blade fracture.

In order to achieve the above purpose, a smart networked wind power generation fault diagnosis and detection system disclosed in an embodiment of the present invention at least includes: a multi-bladed wind turbine, at least having a motor to drive the multi-bladed wind turbine to rotate; a three-axis vibration sensor , electrically connected to the blade fan unit, to sense the vibration state of the three-axis direction when the multi-blade fan unit rotates, and generate a three-axis signal; a high-speed data acquisition card, electrically connected to the three-axis vibration sensor A detector for capturing the three-axis signal; a signal processing unit electrically connected to the high-speed data acquisition card, and filtering the three-axis signal captured by the high-speed data acquisition card; a computer computing system , electrically or network-connected to the signal processing unit for analyzing the type of the fault on the filtered triaxial signal.

Among them, the computer computing system has: a data visualization platform, which converts the filtered three-axis signal into multiple sets of signal feature maps of different types and quantities; a multi-layer convolutional neural network, which separates the multiple sets of signal feature maps Through the corresponding multi-layer convolutional neural network to select different network layers, convolution Kernel size, activation function or pooling method to identify multiple identification results; a resonance theory-like neural network, and then perform an edge suppression operation on the multiple identification results, and then select and decide from a wind power fault defect model Finally output the recognition result.

In order to achieve the above purpose, a fault diagnosis and detection method for smart networked wind power generation disclosed by an embodiment of the present invention includes the following steps: Step 1: Use a three-axis vibration sensor to sense the three-axis vibration of the multi-blade wind turbine when it rotates Direction of the vibration state, and generate a three-axis signal. Step 2: Acquire the three-axis signals through a high-speed data acquisition card. Step 3: filtering the three-axis signal through a signal processing unit. The fourth step: analyzing the type of the fault through a computer operation system through the filtered three-axis signal.

The computer computing system includes the following steps: first step: using a data visualization platform to convert the filtered three-axis signal into multiple sets of signal feature maps of different types and quantities. The second step: through the corresponding multi-layer convolutional neural network to select different network layers, convolution kernel size, activation function or pooling method to identify multiple sets of signal feature maps result. The third step: performing an edge suppression operation on the plurality of identification results through a resonance theory-like neural network, and finally selecting and determining the final output identification result from a fault model of wind power generation.

The above description of the content of the present invention and the following description of the implementation are used to demonstrate and explain the principle of the present invention, and provide further explanation of the patent application scope of the present invention.

1: Fault diagnosis and detection system for intelligent networked wind power generation

11: Blade fan unit

111: fan blade

112: The surface of the blade is damaged

113: blade break

12: Three-axis vibration sensor

13: High-speed data acquisition card

14: Signal processing unit

15:Computer computing system

151: Data visualization platform

152: Convolutional Neural Networks

1521: The first convolutional neural network

1522: The second convolutional neural network

1523: The third convolutional neural network

153: Resonance Theory Neural Networks

154:Wind Power Fault Model

S10~S13: Steps

S20~S22: Steps

Fig. 1 is the overall block schematic diagram of the fault diagnosis and detection system for intelligent networked wind power generation of the present invention;

Fig. 2 is a flow chart of the steps of the intelligent networked wind power generation fault diagnosis and detection method of the present invention;

Fig. 3 is the flow chart of the steps of the computer computing system of the present invention;

FIG. 4A is a signal characteristic diagram of various types of X-axis and Y-axis of the present invention;

FIG. 4B is a characteristic diagram of various types of signals of the X-axis and the Z-axis of the present invention;

FIG. 4C is a characteristic diagram of various types of signals of the Y axis and the Z axis of the present invention;

Fig. 5A is a schematic diagram of the normal state model of the fan blade in the power generation fault defect model of the present invention;

Fig. 5B is a schematic diagram of a defect model of an abnormal fan blade rotation angle in the power generation fault defect model of the present invention;

Fig. 5C is a schematic diagram of a flaw model of a damaged fan blade surface in the power generation fault flaw model of the present invention;

Fig. 5D is a schematic diagram of a defect model of fan blade fracture in the power generation fault defect model of the present invention;

Fig. 6A is a schematic diagram of the vibration signals of the three-axis original signal and the filtered signal caused by the fan blade normal state model in the power generation fault defect model of the present invention;

Fig. 6B is a schematic diagram of the vibration signals of the three-axis original signal and the filtered signal caused by the defect model of the fan blade rotation angle abnormality in the power generation fault defect model of the present invention;

Fig. 6C is a schematic diagram of the vibration signals of the three-axis original signal and the filtered signal caused by the defect model of the fan blade surface damage in the power generation fault defect model of the present invention;

6D is a schematic diagram of vibration signals of three-axis original signals and filtered signals caused by the fault model of fan blade fracture in the power generation fault fault model of the present invention.

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments It is a part of embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts fall within the protection scope of the present invention.

Please refer to FIG. 1 . FIG. 1 is an overall block diagram of a fault diagnosis and detection system for smart networked wind power generation according to the present invention. The fault diagnosis and detection system 1 for smart networked wind power generation according to the embodiment of the present invention at least includes a multi-bladed wind turbine 11, a three-axis vibration sensor 12, a high-speed data acquisition card 13, a signal processing unit 14 and a computer computing system 15. Wherein the blade fan unit 11 is electrically connected Connect the three-axis vibration sensor 12, the three-axis vibration sensor 12 is electrically connected to the high-speed data acquisition card 13, the high-speed data acquisition card 13 is electrically connected to the signal processing unit 14, and the signal processing unit 14 The computer operating system 15 is electrically connected or connected to the network.

The aforementioned multi-blade fan unit 11 includes at least one motor (not shown) to drive the multi-blade fan unit 11 to rotate; wherein the multi-blade fan unit 11 is three sets of blades in the embodiment of the present invention as an implementation description.

The aforementioned computer computing system 15 also includes a data visualization platform 151 , a multi-layer convolutional neural network 152 , a resonance theory (Adaptive Resonance Theory, ART) type neural network 153 and a fault defect model 154 of wind power generation. Wherein the data visualization platform 151 utilizes a graphical programming language of a Laboratory Virtual Instrumentation Engineering Workbench (abbreviated as LabVIEW) and a matrix laboratory (Matrix Laboratory, abbreviated as MATLAB) algorithm in the embodiment of the present invention Composed of development software. Wherein the multi-layer convolutional neural network (Convolutional Neural Networks, referred to as CNN) 152 includes a first convolutional neural network 1521, a second convolutional neural network 1522 and a third convolutional neural network in the embodiment of the present invention Network 1523. Wherein the resonance theory neural network 153 uses a maximum output network (MAXNET) capable of suppressing feedback results in the embodiment of the present invention. Wherein the fault model 154 of power generation failure has at least four defect models in the embodiment of the present invention: a fan blade normal state model type, a defect model of abnormal rotation angle of fan blades, a defect model of damaged blade surface, and a defect model of broken blades.

Above, what needs to be specially explained is that the intelligent networked wind power generation fault diagnosis and detection system 1 of the embodiment of the present invention is used to detect the blade and rotation angle state of the blade fan unit 11 to find the fault type, so a three-axis vibration can be used Sensor 12 senses the vibration state of the three-axis directions (such as X-axis, Y-axis and Z-axis) when the multi-blade on the blade fan unit 11 rotates, and generates three-axis signals (such as an X-axis signal, A Y-axis signal and a Z-axis signal), and then a high-speed data acquisition card 13 is used to capture the three-axis signal, and then the three-axis signal is filtered through a signal processing unit 14, and then the filtered The three-axis signal is analyzed by a computer computing system 15 to analyze the type of the fault. The computer computing system 15 first uses a data visualization platform 151 to convert the three-axis signals into multiple sets of signal feature maps of different types and quantities (for example, a first signal feature map IF1, a second signal feature map IF2, and a third signal feature map. signal feature map IF3), and then the multiple sets of signal feature maps are respectively passed through the corresponding multi-layer convolutional neural network 152 (such as a first convolutional neural network 1521, a second convolutional neural network 1522 and a The third convolutional neural network 1523) selects different network layers, convolution kernel sizes, activation functions or pooling methods to identify a plurality of recognition results (such as a first recognition result CNN1, a first recognition result CNN1, and a second recognition result respectively). Two identification results CNN2 and a third identification result CNN3), and then the multiple identification results are passed through a resonance theory The artificial neural network 153 performs edge suppression calculations, and finally selects from a wind power generation fault defect model 154 to determine the final output identification result.

Please refer to FIG. 2 . FIG. 2 is a flow chart of the steps of the method for fault diagnosis and detection of smart networked wind power generation according to the present invention. The intelligent networked wind power generation fault diagnosis and detection method of the embodiment of the present invention includes the following steps:

The first step S10: use a three-axis vibration sensor to sense the vibration state of the three-axis directions (such as X-axis, Y-axis and Z-axis) when the blade fan unit rotates, and generate a three-axis signal (such as an X Axis signal, a Y-axis signal and a Z-axis signal);

The second step S11: capturing the three-axis signal through a high-speed data acquisition card;

The third step S12: filtering the three-axis signal through a signal processing unit;

The fourth step S13: Finally, analyze the type of the fault through a computer operation system through the filtered three-axis signal.

Please refer to FIG. 3 . FIG. 3 is a flowchart of steps of the computer computing system of the present invention. The computer operation system 15 of the embodiment of the present invention comprises the following steps:

The first step S20: using a data visualization platform to convert the three-axis signals into multiple sets of signal feature maps of different types and quantities, as shown in FIG. 4A , FIG. 4B , and FIG. 4C . Among them, the number N of types of the signal feature map can be calculated by the following mathematical equations to calculate the vibration signals of two relative three axes (such as X axis and Y axis, X axis and Z axis, Y axis and Z axis):

Among them, N is the number of types that can generate signal feature maps, and W is the number of input data.

The second step S21: through the corresponding multi-layer convolutional neural network, select different network layers, convolution kernel sizes, activation functions or pooling methods to identify the multiple sets of signal feature maps. A plurality of identification results; third step S22 : performing an edge suppression operation on the plurality of identification results through a resonance-theoretic neural network, and finally selecting and outputting the identification results from a fault model of wind power generation. Wherein the mathematical equation of the edge suppression operation is:

Where W _ij is the MAXNET weight, ε is the weight between nodes, i and j are MAXNET nodes, M is the number of classes, and the weight of each node to itself is 1.

The fault defect model of wind power generation in the embodiment of the present invention is to establish at least 4 kinds of defect models:

The first type is the normal state model of the fan blade 111 , as shown in FIG. 5A . The three-axis vibration signal (X-axis, Y-axis, Z-axis) of the normal state model will still have a slight vibration, whether it is the original three-axis vibration signal or the filtered three-axis vibration signal, so the defect model A schematic diagram of the induced vibration signal is shown in FIG. 6A .

The second type is the defect model of the abnormal rotation angle of the fan blade 111, as shown in FIG. 5B. When the rotation angles of the three sets of blades of the fan are operated at different angles for a long time, it is easy to cause the imbalance of the fan mechanism, which may cause damage to the mechanism, or cause the fan to collapse. Therefore, in order to describe the technical characteristics of the present invention in detail in this embodiment, one of the three groups of fan blades is adjusted to 45° of rotation angle, and the other two groups of fan blades are 30° in order to simulate the change of the rotation angle of the fan. When paddling, the vibration signal caused by one group of rotation angles not being synchronized. From the detected vibration signals, it can be found that the vibration of the X axis is quite different from the vibration signals of the other two axes, and the highest measured vibration amplitude is about 0.15G, as shown in Figure 6B.

The third type is a defect model of damage 112 on the blade surface, as shown in FIG. 5C . When the three-axis vibration signal measured by the model of the damaged blade surface 112, it can be found that the vibration signal of the Z-axis is relatively different from the vibration signals of the other two axes, and the highest vibration amplitude measured is about 0.38G, as shown in Figure 6C shown.

The fourth type is the defect model of blade fracture 113, as shown in FIG. 5D. When the three-axis vibration signal measured by the model of the blade fracture 113, from which It can be found that the vibration signal of the Z axis is quite different from the vibration signals of the other two axes, and the highest measured vibration amplitude is about 0.48G, as shown in Figure 6D.

Although the present invention has been disclosed above with the foregoing embodiments, it is not intended to limit the present invention. Any person familiar with similar skills may make some changes and modifications without departing from the spirit and scope of the present invention. Therefore, this The scope of patent protection for inventions shall be defined by the scope of protection of the claims attached to this specification.

1: Fault diagnosis and detection system for intelligent networked wind power generation

11: Blade fan unit

12: Three-axis vibration sensor

13: High-speed data acquisition card

14: Signal processing unit

15:Computer computing system

151: Data visualization platform

152: Convolutional Neural Networks

1521: The first convolutional neural network

1522: The second convolutional neural network

1523: The third convolutional neural network

153: Resonance Theory Neural Networks

154:Wind Power Generation Fault Defect Model

Claims (10)

A fault diagnosis and detection system for intelligent networked wind power generation, at least including: a multi-bladed wind turbine, at least having a motor to drive the multi-bladed wind turbine to rotate; a three-axis vibration sensor, electrically connected to the bladed wind turbine, for sensing Measuring the vibration state of the three-axis direction when the multi-blade fan unit rotates, and generating a three-axis signal; a high-speed data acquisition card, electrically connected to the three-axis vibration sensor, for capturing the three-axis signal; A signal processing unit, electrically connected to the high-speed data acquisition card, and filtering the three-axis signal captured by the high-speed data acquisition card; a computer computing system, electrically connected or networked to the signal processing unit , which is used to analyze the type of fault on the filtered triaxial signal; wherein the computer operation system has: a data visualization platform, which converts the filtered triaxial signal into multiple sets of signal characteristics of different types and quantities Figure: Multi-layer convolutional neural network, the multiple sets of signal feature maps are respectively passed through the corresponding multi-layer convolutional neural network to select different network layers, convolution kernel size, activation function or pooling method to identify multiple identification results; a resonance theory-like neural network, and then perform an edge suppression operation on the multiple identification results, and then select and determine the final output identification result from a fault model of wind power generation. According to the intelligent networked wind power generation fault diagnosis and detection system described in Claim 1, the three-axis signals are an X-axis signal, a Y-axis signal and a Z-axis signal. According to claim 1, the fault diagnosis and detection system for intelligent networked wind power generation, wherein the data visualization platform is composed of a graphical programming language of a laboratory virtual instrument engineering platform and a matrix laboratory algorithm development software. The intelligent networked wind power generation fault diagnosis and detection system as described in claim item 1, wherein the equations of the different types and quantities of the multiple sets of signal feature maps are:

Among them, N is the number of types that can generate signal feature maps, and W is the number of input data. The fault diagnosis and detection system for intelligent networked wind power generation according to Claim 1, wherein the resonance theory-based neural network uses a maximum output network capable of suppressing feedback results. The fault diagnosis and detection system for intelligent networked wind power generation as described in Claim 1, wherein the equation of the edge suppression calculation is:

Among them, W _ij is the MAXNET weight, ε is the weight between nodes, i and j are MAXNET nodes, M is the number of classes, and the weight of each node to itself is 1. A method for fault diagnosis and detection of intelligent networked wind power generation, comprising the following steps: first step: using a three-axis vibration sensor to sense the vibration state of the three-axis direction when the multi-blade wind turbine rotates, and generating a three-axis signal; The second step: capture the three-axis signal through a high-speed data acquisition card; the third step: filter the three-axis signal through a signal processing unit; the fourth step: filter the three-axis signal The signal is analyzed by a computer computing system for the type of fault; The computer computing system includes the following steps: first step: using a data visualization platform to convert the filtered three-axis signal into multiple sets of signal feature maps of different types and quantities; second step: then the multiple sets of signals The feature maps use the corresponding multi-layer convolutional neural network to select different network layers, convolution kernel sizes, activation functions or pooling methods to identify multiple identification results; the third step: then the multiple An identification result is subjected to an edge suppression operation through a resonance theory-like neural network, and finally a fault model of a wind power generation is selected to determine the final output of the identification result. The method for fault diagnosis and detection of smart networked wind power generation as described in claim item 7, wherein the equations of the different types and quantities of the multiple sets of signal feature maps are:

Among them, N is the number of types that can generate signal feature maps, and W is the number of input data. The fault diagnosis and detection method for intelligent networked wind power generation as described in Claim 7, wherein the resonance theory neural network uses a maximum output network that can suppress feedback results. The fault diagnosis and detection method of intelligent networked wind power generation as described in claim item 7, wherein the equation of the edge suppression operation is:

Among them, W _ij is the MAXNET weight, ε is the weight between nodes, i and j are MAXNET nodes, M is the number of classes, and the weight of each node to itself is 1.

Images