Wind turbine generator fault diagnosis system and method

Abstract

The invention provides a wind turbine generator fault diagnosis system and method, and the system comprises a data generation module which is used for calculating corresponding simulation data according to the fault mechanism of equipment through employing an aerodynamic system model, a transmission system model and a variable pitch system model, learning a distribution mode reflecting fault categories through a data training generation model together with historical faults, and generating virtual fault data; and the fault diagnosis module is used for performing fault diagnosis on the input operation data of the wind turbine generator by using the trained probabilistic neural network. The method has the advantages that enough training data can be generated according to training requirements, meanwhile, the distribution range of the training data is wider than that of mechanism simulation data, the training data is closer to real operation data, and the precision of a fault diagnosis model is greatly improved; compared with a single probabilistic neural network algorithm, the fault diagnosis module has the advantages that the training process of the probabilistic neural network is optimized by utilizing the genetic algorithm, and the training speed is increased.

Description

Technical Field

[0001] The present application belongs to the field of new energy fault diagnosis, and specifically relates to a wind turbine fault diagnosis system and method. Background Art

[0002] With the rapid development of the new energy industry, the installed capacity of wind power equipment in my country has increased rapidly, and the scale and complexity of wind turbine structures have also continued to increase. The harsh operating environment of wind power systems and the frequent switching of operating conditions have increased the probability of wind turbine failures. Once a failure occurs, it will not only reduce the dynamic performance of the system, but may even cause safety accidents and cause immeasurable losses. However, the fault diagnosis speed of wind turbines lags significantly behind the development speed of wind turbines. One of the important reasons for this phenomenon is the lack of sufficient fault data support.

[0003] The existing technology generally uses existing operating data for model training, but in actual practice it is generally difficult to obtain sufficient fault data, and due to interference, calculation accuracy, model simplification and other reasons, there will be certain errors between the mechanism simulation data and the actual operating data, thus affecting the accuracy of the fault diagnosis model. Summary of the invention

[0004] The purpose of this application is to overcome the defect of insufficient accuracy of the fault diagnosis model in the prior art.

[0005] In order to achieve the above object, the present application proposes a wind turbine fault diagnosis system, the system comprising:

[0006] A data generation module is used to calculate corresponding simulation data based on the failure mechanism of the equipment using an aerodynamic system model, a transmission system model, and a variable pitch system model, and to generate virtual fault data by training a model with historical fault data to learn the distribution pattern of the response fault category; and

[0007] The fault diagnosis module is used to perform fault diagnosis on the input wind turbine operating data using the trained probabilistic neural network.

[0008] As an improvement of the above system, the data generation module includes:

[0009] A wind turbine mechanism submodule, for generating fault data of a specified type using an aerodynamic system model, a transmission system model, and a variable pitch system model according to the fault mechanism, and mixing the data with the fault data and normal data in the historical operation data to form mixed operation data; and

[0010] The data generation submodule is composed of a conditional adversarial network, which is used to mark the fault types of the mixed operation data generated by the wind turbine mechanism submodule, and submit it to the conditional adversarial network for learning and training, and use the trained network to generate different types of virtual fault data.

[0011] As an improvement of the above system, the aerodynamic system model is:

[0012] The wind energy power P captured by the wind turbine rotor from the wind energy is:

[0013]

[0014] Among them, λ is the tip speed ratio; β is the blade pitch angle; R is the rotor radius; V is the wind speed; ρ is the air density;

[0015] C p (λ,β) is the wind energy utilization coefficient:

[0016]

[0017] The parameter λ i The solution is:

[0018]

[0019] As an improvement of the above system, the transmission system model is:

[0020]

[0021] J g =J ω / n 2

[0022] Among them, T ω is the starting torque; T g is the generator torque; n is the gear speed ratio; is the generator speed; ω r is the rotor speed; J g is the moment of inertia of the high-speed shaft side; J ω is the wind wheel moment of inertia.

[0023] As an improvement of the above system, the variable pitch system model is:

[0024]

[0025] β min ≤β≤β max

[0026] Among them, β and β ref They are the actual value of the pitch angle and the reference value of the pitch angle respectively; is the error between the reference value and the actual value; τ is the time constant; β min , β max are the minimum and maximum values ​​of the pitch angle respectively.

[0027] As an improvement of the above system, the probabilistic neural network is trained using virtual data of different fault types generated by a data generation module, and during the training process, a genetic algorithm is used to optimize the weights of the probabilistic neural network;

[0028] The training process of the probabilistic neural network includes:

[0029] Step 1: Initialize the genetic algorithm population size and use a random function to generate an N×D matrix between 0 and 1; N represents the population size and D represents the number of genes in each chromosome. The number of genes is the same as the number of weights in the probabilistic neural network.

[0030] Step 2: Use the probabilistic neural network as the fitness evaluation function, the genes in each chromosome as the weight of the probabilistic neural network, the training error output by the probabilistic neural network as the fitness, and select the chromosome with the smallest error to retain all its information directly to the next generation;

[0031] Step 3: Use the roulette wheel method to select two chromosomes for crossover operation based on fitness, and select one chromosome for mutation operation based on mutation rate;

[0032] Step 4: Determine whether the termination condition is met. If so, output the trained probabilistic neural network model, otherwise execute step 2.

[0033] As an improvement of the above system, the system further comprises:

[0034] The data acquisition module is used to collect the operating data of the wind turbine system.

[0035] The present application also provides a wind turbine fault diagnosis method, which is implemented based on the above system, and the method includes:

[0036] Input the wind turbine operation data into the fault diagnosis module to obtain the fault diagnosis result;

[0037] The training data of the fault diagnosis module is generated by a data generation module.

[0038] Compared with the prior art, the advantages of this application are:

[0039] The present invention uses a small amount of actual operation fault data and mechanism simulation data to generate virtual fault data of different fault types through a generative adversarial network. The virtual data can generate sufficient training data according to training requirements and is also more widely distributed than mechanism simulation data and closer to real operation data, thereby improving the accuracy of fault diagnosis. Compared with the simple use of probabilistic neural network algorithm, the fault diagnosis module uses a genetic algorithm to optimize the training process of the probabilistic neural network and improve the training speed. At the same time, the system includes the entire training and use process of the fault diagnosis model, with high repeatability. Different wind turbines do not need to be remodeled, saving labor costs, while also reducing the skill requirements for users, making it more versatile. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 The figure shows the architecture diagram of the wind turbine fault diagnosis system;

[0041] Figure 2 Shown is a flow chart of the probabilistic neural network training process. DETAILED DESCRIPTION

[0042] The technical solution of the present application is described in detail below with reference to the accompanying drawings.

[0043] The present invention proposes a wind turbine fault diagnosis system and method, wherein the data generation part of the system uses a mechanism model and a small amount of fault data in actual production to generate sufficient virtual fault samples through a conditional adversarial network, which are then handed over to the fault diagnosis part for training and learning. The method proposed by the present invention effectively solves the problem of insufficient model accuracy caused by the small amount of fault data.

[0044] Example 1

[0045] The wind turbine fault diagnosis system consists of three modules: data generation module, data acquisition module and fault diagnosis module.

[0046] 1. Data generation module

[0047] The data generation module contains two models: the wind turbine mechanism submodule and the data training generation submodule. The wind turbine mechanism submodule contains three parts: aerodynamic system model, transmission system model, and variable pitch system model. The data training generation submodule is mainly composed of conditional adversarial generative network (CGAN). According to the failure mechanism of the equipment and the main parameters of the wind turbine itself (such as blade mass, nacelle mass, generator maximum torque, etc.), the wind turbine mechanism submodule is used to calculate the corresponding simulation data, and together with a small number of historical faults, the data training generation model learns the distribution pattern of the response fault category, and finally generates sufficient virtual fault data.

[0048] The wind turbine mechanism submodule uses the aerodynamic system model, transmission system model, and variable pitch system model according to the fault mechanism to generate fault data of a specified type, and mixes the data with a small amount of fault data and normal data in the historical operation data to form mixed operation data, which is finally transmitted to the data generation model.

[0049] The three models are divided into input data and output data. By using the mechanism model to input different variable values ​​as needed, normal production data and fault data can be obtained. However, since the model is simplified, there is a certain gap between the generated data and the real data. Therefore, it is necessary to use a small amount of real fault data and a generative adversarial network to generate some virtual data that is closer to the real fault data.

[0050] Aerodynamic system model:

[0051] The wind energy captured by the wind turbine rotor is:

[0052]

[0053] Where λ is the tip speed ratio, β is the blade pitch angle, C p (λ,β) is the wind energy utilization coefficient, R is the radius of the wind wheel, V is the wind speed, and ρ is the air density.

[0054]

[0055] Where λ i The solution is:

[0056]

[0057] Transmission system model:

[0058] Without taking into account the internal losses of the gearbox and the changes in the power electronic converter of the generator, the transmission system can be expressed as follows:

[0059]

[0060] Among them, T ω is the starting torque, T g is the generator torque, and n is the gear speed ratio.

[0061]

[0062] J g =J ω / n 2

[0063] is the generator speed, ω r is the rotor speed, J gis the moment of inertia of the high-speed shaft side, J ω is the wind wheel moment of inertia.

[0064] Pitch system model:

[0065] The variable pitch system is a device for adjusting the pitch angle of the wind turbine blades. When the wind speed is between the rated wind speed and the cut-off wind speed, constant power output can be achieved by adjusting the pitch angle. The simplified model of the hydraulic variable pitch system can be expressed as follows:

[0066]

[0067] β min ≤β≤β max

[0068] Among them, β and β ref are the actual value of the pitch angle and the reference value of the pitch angle, respectively. is the error between the reference value and the actual value, τ is the time constant, β min , β max are the minimum and maximum values ​​of the pitch angle respectively.

[0069] The data generation submodule is composed of a conditional adversarial network. The data generation submodule marks the fault type of the mixed operation data generated by the wind turbine mechanism submodule, and then submits it to the conditional adversarial network for learning and training. Finally, the trained network is used to generate different types of virtual fault data as required and submitted to the model learning part in the fault diagnosis module for sample training.

[0070] The conditional adversarial network consists of a generator and a discriminator. The training goal of the discriminator is to maximize the accuracy of classification, while the training goal of the generator is to minimize the accuracy of the discriminator. The function evaluations of the two are opposite, thus forming a zero-sum confrontation. The objective function can be expressed as:

[0071]

[0072] Among them, G is the generator, D is the discriminator, and P data Represents the initial given data distribution, P z Represents the randomly initialized noise distribution, G(z) indicates that the input is the random seed z and the output is the virtual fault data, D(f(x)) indicates that the input is the mixed operation data generated by the wind turbine mechanism model, and the output is the probability.

[0073] The specific composition of the conditional adversarial network is as follows:

[0074] The generator is a three-layer neural network consisting of an input layer, a hidden layer, and an output layer. It is actually a reverse discriminator, which is opposite to the discriminator structure. The generator uses a randomly generated seed as input to generate data of the same dimension as the mixed operation data generated by the wind turbine mechanism submodule, and calls the discriminator to update the generator using the comparison loss of the discriminator (the discriminator is not updated at this time).

[0075] The discriminator is a three-layer neural network consisting of an input layer, a hidden layer, and an output layer. The discriminator receives a pair of mixed operation data generated by the generator and the wind turbine mechanism submodule, and marks the data from the generator as 0 and the data from the mixed operation as 1. The discriminator is used to classify and compare it with the label, and the discriminator is updated through the loss.

[0076] 2. Data acquisition module

[0077] This module uses FreeOpcUa to implement data collection. FreeOpcUa is an open source OPC UA (object-oriented unified architecture for process control) library based on Python. This module collects corresponding operating data from the equipment in a cycle of 5 seconds as the fault diagnosis part of the refueling fault diagnosis module with diagnostic samples for diagnosis.

[0078] 3. Fault diagnosis module

[0079] The present invention uses a combined algorithm of a probabilistic neural network and a genetic algorithm as the core part of the fault diagnosis module. The probabilistic neural network uses virtual data of different fault types generated by the data generation module to train the probabilistic neural network model, and uses real historical production fault data and normal data to verify the model. The genetic algorithm optimizes the weight of the probabilistic neural network during the probabilistic neural network training process. After the training is completed, the data transmitted regularly from the data acquisition module is received to perform fault diagnosis and early warning related work.

[0080] The specific training steps of the probabilistic neural network are:

[0081] Step 1: Initialize the genetic algorithm population size and use a random function to generate an N×D matrix between 0 and 1, where N represents the population size (number of chromosomes) and D represents the number of genes in each chromosome. The number of genes is the same as the number of weights in the probabilistic neural network.

[0082] Step 2: Use the probabilistic neural network as the fitness evaluation function, the genes in each chromosome as the weight of the probabilistic neural network, the training error output by the probabilistic neural network as the fitness, and select the chromosome with the smallest error to retain all its information directly to the next generation;

[0083] Step 3: Use the roulette wheel method to select two chromosomes for crossover operation based on fitness. Select a chromosome for mutation operation based on mutation rate;

[0084] Step 4: Determine whether the termination condition is met. If so, output the trained probabilistic neural network model, otherwise execute Step 2.

[0085] The probabilistic neural network consists of an input layer, a pattern layer, a summation layer, and an output layer. The input layer receives virtual data of different fault types generated by the data generation module and passes the feature vector to the network. The number of neurons in the input layer is equal to the dimension of the virtual data vector of different fault types generated by the data generation module. The pattern layer calculates the matching relationship between the input feature vector and each pattern in the training set. The number of neurons in the pattern layer is equal to the sum of the number of training samples of each category. The output of each pattern unit in this layer is:

[0086]

[0087] Among them, W i is the weight of the connection from the input layer to the pattern layer, and δ is the smoothing factor.

[0088] The summation layer accumulates the probability of belonging to a certain category and calculates it according to the above formula to obtain the estimated probability density function of the fault mode. Finally, through the normalization processing of the output layer, the probability estimation of each category can be obtained.

[0089] Example 2

[0090] The present application also provides a wind turbine fault diagnosis method, which is implemented based on the above system, and the method includes:

[0091] Input the wind turbine operation data into the fault diagnosis module to obtain the fault diagnosis result;

[0092] The training data of the fault diagnosis module is generated by a data generation module.

[0093] The present application may also provide a computer device, comprising: at least one processor, a memory, at least one network interface and a user interface. The various components in the device are coupled together through a bus system. It is understood that the bus system is used to achieve connection and communication between these components. In addition to the data bus, the bus system also includes a power bus, a control bus and a status signal bus.

[0094] The user interface may include a display, a keyboard or a pointing device, such as a mouse, a trackball, a touch pad or a touch screen.

[0095] It is understood that the memory in the embodiments disclosed in the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus random access memory (DRRAM). The memories described herein are intended to include, but are not limited to, these and any other suitable types of memories.

[0096] In some embodiments, the memory stores the following elements, executable modules or data structures, or a subset thereof, or an extended set thereof: an operating system and applications.

[0097] The operating system includes various system programs, such as a framework layer, a core library layer, a driver layer, etc., which are used to implement various basic services and process hardware-based tasks. The application includes various application programs, such as a media player (Media Player), a browser (Browser), etc., which are used to implement various application services. The program for implementing the method of the embodiment of the present disclosure can be included in the application.

[0098] In the above embodiment, the processor may also call a program or instruction stored in the memory, specifically, a program or instruction stored in an application program, and is used to:

[0099] Execute the steps of the above method.

[0100] The above method can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor may be a general processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The above-disclosed methods, steps and logic block diagrams can be implemented or executed. The general processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the above-disclosed method can be directly embodied as a hardware decoding processor to execute, or the hardware and software modules in the decoding processor are combined to execute. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0101] It is understood that the embodiments described in the present application can be implemented by hardware, software, firmware, middleware, microcode or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in the present application or a combination thereof.

[0102] For software implementation, the technology of the present application can be implemented by executing the functional modules (such as procedures, functions, etc.) of the present application. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or outside the processor.

[0103] The present application may also provide a non-volatile storage medium for storing a computer program. When the computer program is executed by a processor, each step in the above method embodiment can be implemented.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application is described in detail with reference to the embodiments, a person skilled in the art should understand that any modification or equivalent replacement of the technical solution of the present application does not depart from the spirit and scope of the technical solution of the present application and should be included in the scope of the claims of the present application.

Claims

1. A wind turbine fault diagnosis system, characterized in that: The system comprises: A data generation module is used to calculate corresponding simulation data based on the failure mechanism of the equipment using an aerodynamic system model, a transmission system model, and a variable pitch system model, and to generate virtual fault data by training a model with historical fault data to learn the distribution pattern of the response fault category; and The fault diagnosis module is used to perform fault diagnosis on the input wind turbine operating data using the trained probabilistic neural network.

2. The wind turbine fault diagnosis system according to claim 1, characterized in that: The data generation module comprises: A wind turbine mechanism submodule, for generating fault data of a specified type using an aerodynamic system model, a transmission system model, and a variable pitch system model according to the fault mechanism, and mixing the data with the fault data and normal data in the historical operation data to form mixed operation data; and The data generation submodule is composed of a conditional adversarial network, which is used to mark the fault types of the mixed operation data generated by the wind turbine mechanism submodule, and submit it to the conditional adversarial network for learning and training, and use the trained network to generate different types of virtual fault data.

3. The wind turbine fault diagnosis system according to claim 1, characterized in that: The aerodynamic system model is: The wind energy power P captured by the wind turbine rotor from the wind energy is: Among them, λ is the tip speed ratio; β is the blade pitch angle; R is the rotor radius; V is the wind speed; ρ is the air density; C p (λ,β) is the wind energy utilization coefficient: The parameter λ i The solution is:

4. The wind turbine fault diagnosis system according to claim 1, characterized in that: The transmission system model is: Jg=J ω / n 2 Among them, T ω is the starting torque; T g is the generator torque; n is the gear speed ratio; is the generator speed; ω r is the rotor speed; J g is the moment of inertia of the high-speed shaft side; J ω is the wind wheel moment of inertia.

5. The wind turbine fault diagnosis system according to claim 1, characterized in that: The variable pitch system model is: β min ≤β≤β max Among them, β and β ref They are the actual value of the pitch angle and the reference value of the pitch angle respectively; is the error between the reference value and the actual value; τ is the time constant; β min , β max are the minimum and maximum values ​​of the pitch angle respectively.

6. The wind turbine fault diagnosis system according to claim 1, characterized in that: The probabilistic neural network is trained using virtual data of different fault types generated by a data generation module. During the training process, a genetic algorithm is used to optimize the weights of the probabilistic neural network. The training process of the probabilistic neural network includes: Step 1: Initialize the genetic algorithm population size and use a random function to generate an N×D matrix between 0 and 1; Where N represents the population size, D represents the number of genes in each chromosome, and the number of genes is the same as the number of weights in the probabilistic neural network; Step 2: Use the probabilistic neural network as the fitness evaluation function, the genes in each chromosome as the weight of the probabilistic neural network, the training error output by the probabilistic neural network as the fitness, and select the chromosome with the smallest error to retain all its information directly to the next generation; Step 3: Use the roulette wheel method to select two chromosomes for crossover operation based on fitness, and select one chromosome for mutation operation based on mutation rate; Step 4: Determine whether the termination condition is met. If so, output the trained probabilistic neural network model, otherwise execute step 2.

7. The wind turbine fault diagnosis system according to claim 1, characterized in that: The system further comprises: The data acquisition module is used to collect the operating data of the wind turbine system.

8. A wind turbine fault diagnosis method, implemented based on the system according to any one of claims 1 to 7, the method comprising: Input the wind turbine operation data into the fault diagnosis module to obtain the fault diagnosis result; The training data of the fault diagnosis module is generated by a data generation module.

Images