Multi-source sensor nonlinear error correction method based on fan vibration fault detection

Abstract

The invention discloses a fan vibration fault detection-based multi-source sensor nonlinear error correction method, which comprises the following steps of: establishing a weight analysis model for analyzing relevance between fan vibration influence factors and fan vibration faults on the basis of data acquired by a fan multi-source sensor; determining the influence weight of each fan vibration influence factor on fan vibration fault detection according to the weight analysis model; according to the influence weight, main factors influencing the fan vibration fault detection are selected, and a multi-source sensor nonlinear error correction model is established based on the main factors influencing the fan vibration fault detection; and correcting the error of the multi-source sensor based on the nonlinear error correction model. According to the method, the influence weight of each influence factor on fan vibration fault detection is analyzed, the nonlinear error correction model of the multi-source sensor data is established based on the influence weight, and the model can effectively improve the accuracy of fan vibration fault monitoring.

Description

Technical Field

[0001] The invention relates to a sensor nonlinear correction method, in particular to a multi-source sensor nonlinear error correction method based on fan vibration fault detection. Background Art

[0002] Wind power generation technology has become an important renewable energy power generation method due to its advantages such as low cost and cleanliness. The safe and stable operation of wind turbines is crucial to the power generation capacity and operating efficiency of wind farms. However, wind farms are usually located in areas with complex natural environments. Wind turbines are subject to the influence of various external factors such as wind loads and vibrations. They are prone to problems such as fatigue and cracks, which affect the normal operation of wind turbines. Therefore, vibration fault detection of wind turbines is very important.

[0003] Magnetostrictive displacement sensors are widely used in vibration fault detection of wind turbines. The sensors can accurately measure the displacement and position of each component of the wind turbine, providing data support for vibration fault detection of the wind turbine. However, the input and output characteristics of magnetostrictive displacement sensors are nonlinear, and there are nonlinear errors, so they need to be corrected from the software.

[0004] The operation of a wind farm is a dynamic process. The relationship between the operating parameters of the wind turbine and the environmental data of the wind farm is complex, which will affect the accuracy of vibration fault detection of the wind turbine. Therefore, it is not enough to correct the data collected by the magnetostrictive displacement sensor alone. It is also necessary to consider the correlation between various influencing factors and perform nonlinear error correction on multi-source sensor data, which is also one of the focuses of this type of research. Summary of the invention

[0005] Purpose of the invention: The purpose of the present invention is to provide a method for correcting nonlinear errors of multi-source sensors based on fan vibration fault detection, so as to solve the problem of how to correct nonlinear errors of multi-source sensor data.

[0006] Technical solution: The multi-source sensor nonlinear error correction method based on fan vibration fault detection described in the present invention comprises the following steps:

[0007] Based on the data collected by the fan multi-source sensors, a weight analysis model is established to analyze the correlation between the factors affecting fan vibration and fan vibration failure.

[0008] Determine the influence weight of each fan vibration influencing factor on fan vibration fault detection according to the weight analysis model;

[0009] The main factors affecting the fan vibration fault detection are selected according to the influence weights, and a multi-source sensor nonlinear error correction model is established based on the main factors affecting the fan vibration fault detection;

[0010] Correct multi-source sensor errors based on nonlinear error correction model.

[0011] Preferably, the data collected by the wind turbine multi-source sensor includes wind turbine operating status and environmental monitoring data.

[0012] Preferably, the weight analysis model is:

[0013]

[0014] Among them, suppose that n wind turbine operation status monitoring and environment monitoring sensors collect data respectively i1 ,x i2 ,…,x in , each sensor collects m sets of data, i = 1, 2, 3, ..., m, is the fan vibration data predicted by the model; a1, a2, …, a n is the weight of each fan vibration influencing factor, and a0 is the disturbance value.

[0015] Preferably, determining the influence weight of each fan vibration influencing factor on the fan vibration fault detection includes:

[0016] Assume the loss function J of the weighted analysis model is

[0017]

[0018] In the formula, y i The actual vibration data of the fan collected by the fan vibration monitoring sensor;

[0019] Define the gradient as

[0020]

[0021] In the formula, is the partial derivative operator.

[0022] Then the update formula of the weight coefficient is

[0023] a new =a old -η▽J

[0024] In the formula, a new =[a0,a1,…,a n ] new is the weight coefficient after update; a old =[a0,a1,…,a n ] old is the weight coefficient before updating; η is the learning rate of model parameter updating;

[0025] When the loss function meets the conditions, stop training and output the weight coefficient a=[a0,a1,…,a n ], and obtain the influence weights of various fan vibration influencing factors on fan vibration fault detection.

[0026] Preferably, the method for selecting the main factors affecting fan vibration fault detection includes:

[0027] The three fan vibration influencing factors with the largest influence weights are selected as the main factors affecting fan vibration fault detection.

[0028] Preferably, the establishing of a multi-source sensor nonlinear error correction model comprises:

[0029] Each sensor corresponding to the main factors affecting fan vibration fault detection collects k sets of data, and the nonlinear error correction model of multi-source sensors is expressed as:

[0030]

[0031] in, is the corrected fan vibration monitoring situation; x i1 ,x i2 ,x i3 is the data obtained by the sensors corresponding to the three fan vibration influencing factors with the largest weights on fan vibration fault detection, i = 1, 2, 3, ..., k; a1, a2, a3 are the data corresponding to x i1 ,x i2 ,x i3 The corresponding weights of the three fan vibration influencing factors; b0, b1, b2, b3 are the coefficients of the nonlinear error correction model.

[0032] Preferably, the correcting the multi-source sensor errors based on the nonlinear error correction model includes:

[0033] Assume that the error function of the nonlinear error correction model is

[0034]

[0035] The goal of optimization is to minimize the error function:

[0036]

[0037] The simultaneous equations are solved to obtain the coefficients b0, b1, b2, and b3 of the nonlinear error correction model. The coefficients of the nonlinear error correction model are substituted into the nonlinear error correction model to realize the nonlinear error correction of multi-source sensors based on fan vibration fault detection.

[0038] Preferably, minimizing the error function includes making the partial derivative of the error function with respect to the four coefficients of the equation equal to 0.

[0039] Based on the above method, the present invention further discloses a computer-readable storage medium storing a computer program, wherein the computer program implements the steps of the above method when executed by a processor.

[0040] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0041] The present invention comprehensively considers the impact of wind turbine operating status data, wind farm environmental monitoring data and other multi-source sensor data on the accuracy of wind turbine vibration fault detection. First, the influence weights of various influencing factors on wind turbine vibration fault detection are analyzed, and a nonlinear error correction model of multi-source sensor data is established based on the influence weights. This model can effectively improve the accuracy of wind turbine vibration fault monitoring. DETAILED DESCRIPTION

[0042] The technical solution of the present invention is further described below.

[0043] The nonlinear error correction method of multi-source sensors based on fan vibration fault detection is as follows:

[0044] 1. Establish a weight analysis model for the impact of wind turbine multi-source sensor data on vibration fault detection

[0045] In order to ensure the stable operation of the wind farm, it is necessary to use sensors to collect a large amount of data, including wind turbine operating status data such as speed, active output, torque, and environmental monitoring data such as wind speed, wind direction, and temperature. These data, as factors affecting wind turbine vibration, will affect the accuracy of wind turbine vibration fault detection, but the degree of influence of different factors is different. Therefore, it is necessary to analyze the correlation between these factors and wind turbine vibration faults and determine the weight of each influencing factor. To this end, the present invention establishes a weight analysis model:

[0046] Assume that n wind turbine operating status monitoring and environmental monitoring sensors collect data x1, x2, …, x n , then the weight analysis model of each fan vibration influencing factor is:

[0047]

[0048] In the formula, is the fan vibration fault data predicted by the model, a1, a2, …, a n is the weight of each fan vibration influencing factor, and a0 is the disturbance value.

[0049] Each sensor collects m groups of data, and a matrix model is established between each fan vibration influencing factor and the fan vibration fault monitoring data:

[0050]

[0051] Among them, suppose that n wind turbine operation status monitoring and environment monitoring sensors collect data respectively i1 ,x i2 ,…,x in , each sensor collects m sets of data, i = 1, 2, 3, ..., m, is the fan vibration data predicted by the model; a1, a2, …, a n is the weight of each fan vibration influencing factor, and a0 is the disturbance value.

[0052] 2. Determine the weight of each fan vibration influencing factor based on the weight analysis model

[0053] The gradient method is used to analyze the weight analysis model established in step 1 to confirm the weight of each fan vibration influencing factor on fan vibration fault detection.

[0054] Assume the loss function J of the weighted analysis model is

[0055]

[0056] In the formula, y i The actual vibration data of the fan collected by the fan vibration monitoring sensor.

[0057] Define the gradient as

[0058]

[0059] In the formula, is the partial derivative operator.

[0060] Then we can get the gradient matrix

[0061]

[0062] Update the weight coefficients in the weight analysis model according to the gradient matrix

[0063]

[0064] In the formula, a new =[a0,a1,…,a n ] new is the weight coefficient after update; a old =[a0,a1,…,a n ] old is the weight coefficient before updating; η is the learning rate of model parameter updating.

[0065] When the loss function meets the conditions, stop training and output the weight coefficient a=[a0,a1,…,a n], and obtain the weights of various fan vibration influencing factors on fan vibration fault detection.

[0066] 3. Establish a multi-source sensor nonlinear error correction model based on fan vibration fault detection

[0067] Select the three fan vibration influencing factors with the largest weights in step 2 as the main factors affecting fan vibration fault detection. The nonlinear error correction model of multi-source sensors based on fan vibration fault detection can be expressed as

[0068]

[0069] in, is the corrected fan vibration monitoring situation, x1, x2, x3 are the data obtained by the sensors corresponding to the three fan vibration influencing factors with the largest influencing weights on fan vibration fault detection in step 2; a1, a2, a3 are the weights of the three fan vibration influencing factors corresponding to x1, x2, x3, and b0, b1, b2, b3 are the coefficients of the nonlinear error correction model.

[0070] Each sensor collects k groups of data and establishes a matrix model for nonlinear error correction of multi-source sensors:

[0071]

[0072] in, is the corrected fan vibration monitoring situation; x i1 ,x i2 ,x i3 is the data obtained by the sensors corresponding to the three fan vibration influencing factors with the largest weights on fan vibration fault detection, i = 1, 2, 3, ..., k; a1, a2, a3 are the data corresponding to x i1 ,x i2 ,x i3 The corresponding weights of the three fan vibration influencing factors.

[0073] 4. Correction of multi-source sensor errors based on nonlinear error correction model

[0074] Based on the nonlinear error correction model in step 3, the errors of multi-source sensors are corrected to improve the accuracy of vibration fault detection.

[0075] Assume that the error function of the nonlinear error correction model is

[0076]

[0077] The goal of optimization is to minimize the error function, that is, the partial derivative of the error function with respect to the four coefficients of the equation is 0:

[0078]

[0079] Arrange the equation system

[0080]

[0081] In the formula

[0082]

[0083]

[0084] Solve the system of simultaneous equations to obtain the coefficients of the nonlinear correction model

[0085] b0=D1 / D, b1=D2 / D, b2=D3 / D, b3=D4 / D

[0086] In the formula

[0087]

[0088] Substituting the nonlinear error correction model coefficients b0, b1, b2, and b3 into the nonlinear error correction model, the model can be used to realize the nonlinear error correction of multi-source sensors based on fan vibration fault detection.

Claims

1. A multi-source sensor nonlinear error correction method based on fan vibration fault detection, characterized in that: The steps include: Based on the data collected by the fan multi-source sensors, a weight analysis model is established to analyze the correlation between the factors affecting fan vibration and fan vibration failure. Determine the influence weight of each fan vibration influencing factor on fan vibration fault detection according to the weight analysis model; The main factors affecting the fan vibration fault detection are selected according to the influence weights, and a multi-source sensor nonlinear error correction model is established based on the main factors affecting the fan vibration fault detection; Correct multi-source sensor errors based on nonlinear error correction model.

2. The multi-source sensor nonlinear error correction method based on fan vibration fault detection according to claim 1 is characterized in that: The data collected by the wind turbine multi-source sensor includes wind turbine operating status and environmental monitoring data.

3. The multi-source sensor nonlinear error correction method based on fan vibration fault detection according to claim 1 is characterized in that: The weight analysis model is: Among them, suppose that n wind turbine operation status monitoring and environment monitoring sensors collect data respectively i1 ,x i2 ,…,x in , each sensor collects m sets of data, i = 1, 2, 3, ..., m, is the fan vibration data predicted by the model; a1, a2, …, a n is the weight of each fan vibration influencing factor, and a0 is the disturbance value.

4. The multi-source sensor nonlinear error correction method based on fan vibration fault detection according to claim 1 is characterized in that: Determining the influence weight of each fan vibration influencing factor on the fan vibration fault detection includes: Assume the loss function J of the weighted analysis model is In the formula, y i Actual vibration data of the fan collected by the fan vibration monitoring sensor; Define the gradient as In the formula, is the partial derivative operator; Then the update formula of the weight coefficient is In the formula, a new =[a0,a1,…,a n ] new is the weight coefficient after update; a old =[a0,a1,…,a n ] old is the weight coefficient before updating; η is the learning rate of model parameter updating; When the loss function meets the conditions, stop training and output the weight coefficient a=[a0,a1,…,a n ], and obtain the influence weights of various fan vibration influencing factors on fan vibration fault detection.

5. The multi-source sensor nonlinear error correction method based on fan vibration fault detection according to claim 1 is characterized in that: The method for selecting the main factors affecting fan vibration fault detection includes: The three fan vibration influencing factors with the largest influence weights are selected as the main factors affecting fan vibration fault detection.

6. The multi-source sensor nonlinear error correction method based on fan vibration fault detection according to claim 1 is characterized in that: The establishment of a multi-source sensor nonlinear error correction model comprises: Each sensor corresponding to the main factors affecting fan vibration fault detection collects k sets of data, and the nonlinear error correction model of multi-source sensors is expressed as: Among them, y i is the corrected fan vibration monitoring situation; x i1 ,x i2 ,x i3 is the data obtained by the sensors corresponding to the three fan vibration influencing factors with the largest weights on fan vibration fault detection, i = 1, 2, 3, ..., k; a1, a2, a3 are the data corresponding to x i1 ,x i2 ,x i3 The corresponding weights of the three fan vibration influencing factors; b0, b1, b2, b3 are the coefficients of the nonlinear error correction model.

7. The multi-source sensor nonlinear error correction method based on fan vibration fault detection according to claim 1 is characterized in that: The correction of multi-source sensor errors based on the nonlinear error correction model includes: Assume that the error function of the nonlinear error correction model is The goal of optimization is to minimize the error function: The simultaneous equations are solved to obtain the coefficients b0, b1, b2, and b3 of the nonlinear error correction model. The coefficients of the nonlinear error correction model are substituted into the nonlinear error correction model to realize the nonlinear error correction of multi-source sensors based on fan vibration fault detection.

8. The multi-source sensor nonlinear error correction method based on fan vibration fault detection according to claim 7 is characterized in that: The step of minimizing the error function includes making the partial derivatives of the error function with respect to the four coefficients of the equation equal to 0.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images