Intelligent state monitoring system with autonomous learning characteristic for tower drum of wind generating set

Abstract

The invention discloses a monitoring and diagnosing system for a tower drum of a wind generating set. According to the method, the information of the tower drum of the wind generating set is acquired from multiple dimensions, a system architecture with information acquisition, feature extraction, intelligent diagnosis and autonomous learning functions is constructed, a set of complete diagnosis standards for the running state of the tower drum of the wind generating set is generated, the timeliness and accuracy of fault diagnosis of the tower drum of the wind generating set are improved, and the fault diagnosis efficiency is improved. The influence of wind load on tower drum fault diagnosis is reduced, and a self-adaptive diagnosis strategy of the tower drum of the wind generating set is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of wind power generation, and relates to a method for monitoring the state of a wind turbine tower barrel. Background Art

[0002] With the expansion of the new energy market map and the continuous growth of the installed capacity of wind power generation, the construction of wind power projects is gradually developing towards the third and fourth types of inland areas. The tower barrels and blades of wind turbines are getting taller and longer. Major accidents such as the inclination and collapse of wind turbine tower barrels occur frequently, posing a serious threat to the safety production of power generation enterprises. The tower barrel is a very important part of the wind power industry chain. It not only supports the entire wind turbine, but also absorbs the vibration of the wind turbine. The problem of damage to the wind turbine tower barrel has attracted great attention at home and abroad. Therefore, the method for monitoring the damage of the wind turbine tower barrel has become a research hotspot in the engineering field in recent years. There are still some problems in the existing tower barrel monitoring methods. First, there is a lack of relatively mature judgment criteria for diagnosing the state of the tower barrel. Second, a single-factor fault may be affected by other factors. Monitoring only a single factor is not conducive to the timely discovery of tower barrel faults. Third, due to different wind loads, the degree of tower barrel faults is different, which is not conducive to the diagnosis of tower barrel faults. Fourth, at present, the maintenance of the tower barrel is in a fault-triggered mode. The tower barrel diagnosis method cannot be updated and iterated in a timely and effective manner, and the detection and diagnosis method lacks the ability of autonomous learning. Summary of the Invention

[0003] The present method proposes a monitoring and diagnosis system for the tower barrel of a wind power generation unit. The system monitors the bolt state information, tower barrel state information and wind load state information of the wind turbine, constructs an architecture with functions of information collection, feature extraction, intelligent diagnosis and autonomous learning, improves the timeliness and accuracy of the fault diagnosis of the tower barrel of the wind power generation unit, reduces the influence of wind load on the tower barrel fault diagnosis, and realizes the adaptive diagnosis strategy of the tower barrel of the wind power generation unit.

[0004] As Figure 1 shown, the system includes four levels, namely the wind farm information collection layer, the feature database, the intelligent diagnosis layer and the autonomous learning layer.

[0005] The wind farm information collection layer is responsible for collecting the bolt state information, wind load information and tower barrel state information of the wind power generation unit. The bolt state information includes the pre-tightening force of the fixed bolts of the first-layer tower barrel, the pre-tightening force of the fixed bolts of the second-layer tower barrel, the pre-tightening force of the fixed bolts of the third-layer tower barrel, and the pre-tightening force of the fixed bolts of the fourth-layer tower barrel. The wind load information includes the wind vibration coefficient at height z, the wind pressure height change coefficient, the wind load shape coefficient, and the basic wind pressure. The tower barrel state information includes the average value of the tower barrel verticality within a period of time, the effective value average of the nacelle swing within a period of time, and the effective value average of the tower base inclination within a period of time.

[0006] The feature database is responsible for extracting features from the collected bolt status information and tower status information to obtain feature factors. The bolt status feature fusion module fuses the collected bolt status information to obtain bolt feature factors, and takes the bolt feature factors as new data and inputs them into the intelligent diagnosis layer. The tower status feature fusion module fuses the collected tower status information to obtain tower feature factors, and takes the tower feature factors as new data and inputs them into the intelligent diagnosis layer.

[0007] The intelligent diagnosis layer is responsible for extracting the diagnostic factors of the operating status of the wind turbine tower and generating the diagnostic criteria for the operating status of the wind turbine tower. First, the collected wind load information is input into the tower dynamics feature estimation module to obtain the theoretical bolt feature function and tower feature function under the same wind load information; second, the theoretical bolt feature function is input into the bolt self-diagnosis module, and a correlation analysis is carried out with the bolt feature factors to obtain bolt statistical feature parameters, construct a bolt status diagnosis interval, perform coordinate transformation on the bolt status diagnosis interval to obtain bolt status diagnostic factors, and input the bolt status diagnostic factors into the status factor module; third, the theoretical tower feature function is input into the tower self-diagnosis module, and a correlation analysis is carried out with the tower feature factors to obtain tower statistical feature parameters, construct a tower status diagnosis interval, perform coordinate transformation on the tower status diagnosis interval to obtain tower status diagnostic factors, and input the tower status diagnostic factors into the status factor module; finally, the status factor module fuses the obtained bolt status diagnostic factors and tower status diagnostic factors to obtain the diagnostic factors for the operating status of the wind turbine tower, generates the diagnostic criteria for the tower status of the entire wind farm, and takes the diagnostic criteria as new data and inputs them into the autonomous learning layer.

[0008] The autonomous learning layer is responsible for autonomously iteratively learning the diagnostic criteria for the operating status of the wind turbine tower. The status of the tower is evaluated according to the criteria obtained by the intelligent diagnosis layer, and the on-site maintenance feedback results are compared with the diagnostic criteria. First, calculate the error of the status diagnosis interval, and achieve digital feedback by finding the error boundary value; second, calculate the statistical feature parameters of the status change factor, and achieve status feedback by updating the status change factor. Input the results of digital feedback and status feedback into the intelligent diagnosis layer to realize the iterative learning of the diagnostic system. Description of the Drawings

[0009] Figure 1 Intelligent Condition Monitoring System for Wind Turbine Tower with Autonomous Learning Feature Detailed Implementation Manner

[0010] The wind field information acquisition layer collects the bolt status information, wind load status information, and tower barrel status information of the wind turbine tower. The bolt status information includes the pre-tightening force F1 of the fixed bolts of the first-layer tower barrel, the pre-tightening force F2 of the fixed bolts of the second-layer tower barrel, the pre-tightening force F3 of the fixed bolts of the third-layer tower barrel, and the pre-tightening force F4 of the fixed bolts of the fourth-layer tower barrel. The tower barrel status information includes the average value b1 of the tower barrel verticality within time t, the average effective value b2 of the nacelle swing within time t, and the average effective value b3 of the tower base inclination within time t. The specific manifestation form of the wind load status information is as follows:

[0011] C = β z μ s μ z ω0

[0012] In the formula

[0013] C is the wind load factor;

[0014] β z is the wind vibration coefficient at height z;

[0015] μ s is the wind pressure height change coefficient;

[0016] μ z is the wind load shape coefficient;

[0017] ω0 is the basic wind pressure, approximately taken as 600 N / m 2 .

[0018] Input the bolt status information and tower barrel status information into the feature database, and perform feature fusion on the bolt status information. The specific manifestation form is as follows:

[0019]

[0020] In the formula

[0021] A is the bolt feature factor;

[0022] f is the data after fusing s kinds of bolt status information;

[0023] ω 1s is the weight of the s-th kind of bolt status information data;

[0024] σ1 2 is the variance after fusing s kinds of bolt status information data;

[0025] x1 is the true value after fusing s kinds of bolt status information data.

[0026] When determining the variance σ after fusing the s-th kind of bolt status information data 1s 2When using the mean value of the s-th bolt status information as the unbiased estimate of the fused data, the variance of the s-th bolt status information data is denoted as ω 1s Calculate according to the following formula:

[0027]

[0028] Fuse the tower barrel status information, and its specific manifestation is as follows:

[0029]

[0030] In the formula

[0031] B is the tower barrel characteristic factor;

[0032] b1 is the mean value of the tower barrel verticality within time t, b2 is the effective value mean of the nacelle swing within time t, and b3 is the effective value mean of the tower base inclination within time t;

[0033] d1 is the upper center radius of the tower barrel;

[0034] d2 is the lower center radius of the tower barrel;

[0035] h is the height of the tower barrel;

[0036] b is the data after fusing the s types of tower barrel status information;

[0037] ω 2s is the weight of the s-th tower barrel status information data;

[0038] σ2 2 is the variance after fusing the s types of tower barrel status information data;

[0039] x2 is the true value after fusing the s types of tower barrel status information data.

[0040] When determining the variance σ of the data after fusing the s-th tower barrel status information 2s 2 When using the mean value of the s-th tower barrel status information as the unbiased estimate of the fused data, the variance of the s-th tower barrel status information data is denoted as ω 2s Calculate according to the following formula:

[0041]

[0042] Input the wind load factor into the estimation of the tower barrel dynamic characteristics to obtain the bolt dynamic equation f(x) and the tower barrel dynamic equation g(x). Under the same wind load factor C, there is a theoretical value for the bolt factor A That is There is a theoretical value for the tower barrel factor B That is Input factors A and B obtained from the feature database into the intelligent diagnosis layer for experiments, and their manifestations are as follows:

[0043]

[0044] In the formula

[0045] A i is the i-th factor A;

[0046] B j is the j-th factor B;

[0047]

[0048] σ a 2 is the variance of factor A i ;

[0049] σ b 2 is the variance of factor B j ;

[0050] According to the obtained statistical characteristic parameters μ a , σ a 2 , the statistical characteristic parameters μ b , σ b 2 are used to obtain three state intervals:

[0051]

[0052] In the formula

[0053] Q1 represents interval one, Q2 represents interval two, and Q3 represents interval three;

[0054] l represents factors a and b;

[0055] p l1 represents the normal state probability parameter of factor l;

[0056] p l2 represents the warning state probability parameter of factor l;

[0057] p l3 represents the alarm state probability parameter of factor l;

[0058] When l = a, it is the bolt characteristic factor, μ l = μ a , σ l 2 = σ a 2 p l1 = p a1, p l2 = p a2 , p l3 = p a3 When A i ∈ Q1, the bolt characteristic factor is in a normal state. When A i ∈ Q2, the bolt characteristic factor is in a warning state. When A i ∈ Q3, the bolt characteristic factor is in an alarm state;

[0059] When l = b, it is the tower barrel characteristic factor, μ l = μ b , σ l 2 = σ b 2 , p l1 = p b1 , p l2 = p b2 , p l3 = p b3 When B j ∈ Q1, the tower barrel characteristic factor is in a normal state. When B j ∈ Q2, the tower barrel characteristic factor is in a warning state. When B j ∈ Q3, the tower barrel characteristic factor is in an alarm state.

[0060] According to the obtained state intervals, the state diagnosis factor P l is obtained simultaneously:

[0061]

[0062] wherein

[0063] When l = a, it is the bolt state diagnosis factor P a ;

[0064] When l = b, it is the tower barrel state diagnosis factor P b .

[0065] The obtained state diagnosis factors are simulated and integrated to obtain the operation state diagnosis factor P:

[0066]

[0067] wherein

[0068] p 11 is the probability that the tower barrel state remains normal;

[0069] p 12 is the probability that the tower barrel state changes from normal to warning;

[0070] p 13 is the probability that the tower barrel state changes from normal to alarm;

[0071] p 22 is the probability of maintaining the early warning for the tower barrel status;

[0072] p 23 is the probability of the tower barrel status changing from early warning to alarm.

[0073] Take the obtained state transition factor P as the operation status diagnosis factor of the tower barrel. If the change rate of the operation diagnosis factor exceeds M%, it is considered that the status of the tower barrel has changed.

[0074] The autonomous learning layer is responsible for conducting status evaluation and inspection on the obtained tower barrel status diagnosis criteria, collecting the feedback results of on-site maintenance. If the judgment is accurate, output the obtained status diagnosis factor and factor change rate as the tower barrel status diagnosis criteria; if the judgment is inaccurate, perform iterative update on the intelligent diagnosis layer through status feedback and digital feedback.

[0075] The process of status feedback is as follows:

[0076] G1: Collect the change rate M of the inaccurate diagnosis factor, g , with a total of m;

[0077] G2: Calculate the mean μ of the change rate of the diagnosis factor, g , that is,

[0078] G3: Calculate the variance σ of the change rate of the diagnosis factor, g 2 , that is,

[0079] G4: According to the on-site feedback situation, let M′ = M + σ g 2 or M′ = M - σ g 2 be the new change rate of the diagnosis factor, thereby realizing status feedback.

[0080] The process of digital feedback is as follows:

[0081] F1: Collect the parameter A of the inaccurate judgment, i , B j ,

[0082] F2: Randomly select 3 centers from A i , denoted as μ f (f = 1, 2, 3);

[0083] F3: Calculate the objective function of A i ,

[0084] F4: Repeat step F2 until the minimum value of the objective function is obtained, i.e.,

[0085] F5: Calculate the central values of each cluster through iterative calculation, and calculate the boundary value H through backtracking calculation av , where v = 1, 2, 3. That is, when v = 1, it represents the normal state; when v = 2, it represents the warning state; when v = 3, it represents the alarm state;

[0086] F6: According to the boundary value H av Iterate the state probability parameters, i.e.,

[0087] F7: Similarly, perform steps F2 - F6 on the parameter B j to obtain the probability parameters after boundary value iteration, and then realize digital feedback.

Claims

1. A smart condition monitoring system for the tower barrel of a wind turbine with autonomous learning characteristics fully monitors the bolt condition information, tower barrel condition information, and wind load condition information of the wind turbine, constructs an architecture with functions of information acquisition, feature extraction, intelligent diagnosis, and autonomous learning, effectively integrates theoretical feature information into the real environment, combines the tower barrel condition diagnosis results of the wind turbine with on-site feedback results, and proposes a method for monitoring the operating condition of the tower barrel of a wind turbine with autonomous learning ability.

2. The intelligent condition monitoring system for the wind turbine tower barrel with autonomous learning characteristics according to claim 1, characterized in that The bolt condition information is fused in the following manner: Where A is the bolt feature factor; f is the data after fusing s kinds of bolt condition information; ω 1s is the weight of the data of the s-th bolt status information; σ1 2 is the variance after data fusion of s kinds of bolt status information; x1 is the true value after fusing s kinds of bolt condition information data. When determining the variance σ after data fusion of the state information of the s-th bolt 1s 2 the mean value of the state information of the s-th bolt is used as the unbiased estimate of the fused data, and the variance of the state information data of the s-th bolt obtained by estimation is denoted as ω 1s Calculated according to the following formula:

3. The intelligent condition monitoring system for the wind turbine tower barrel with the characteristic of autonomous learning according to claim 1, wherein The tower barrel condition information is fused in the following manner: Where B is the tower barrel feature factor; b1 is the average value of the tower barrel verticality within t time, b2 is the effective value average of the nacelle swing within t time, and b3 is the effective value average of the tower base inclination within t time; d1 is the upper center radius of the tower barrel; d2 is the lower center radius of the tower barrel; h is the height of the tower barrel; b is the data after fusing s kinds of tower barrel condition information; ω 2s is the weight of the data of the s-th tower barrel status information; σ2 2 is the variance after data fusion of s kinds of tower barrel state information; x2 is the true value after fusing s kinds of tower barrel condition information data. When determining the variance σ after data fusion of the s-th tower barrel state information 2s 2 , the mean value of the s-th tower barrel state information is used as the unbiased estimate of the fused data, and the estimated variance of the s-th tower barrel state information data is denoted as ω 2s Calculate according to the following formula:

4. The intelligent condition monitoring system for the wind turbine tower barrel with autonomous learning characteristics according to claim 1, characterized in that The bolt condition diagnosis interval and the tower barrel condition diagnosis interval are constructed according to the following steps: Input the wind load factor into the dynamic characteristic estimation of the tower barrel to obtain the bolt dynamic equation f(x) and the tower barrel dynamic equation g(x). Under the same wind load factor C, there is a theoretical value for the bolt factor A That is There is a theoretical value for the tower barrel factor B That is Input the factors A and B obtained from the characteristic database into the intelligent diagnosis layer for experiments. Their manifestation forms are as follows: Where A i is the i-th factor A; B j is the j-th factor B; σ a 2 is the variance of factor A i ; σ b 2 is the variance of factor B j ; According to the obtained statistical characteristic parameters μ a , σ a 2 , the statistical characteristic parameters μ b , σ b 2 yield three state intervals: Where Q1 represents interval one, Q2 represents interval two, and Q3 represents interval three; l represents factors a and b; p l1 Represents the normal state probability parameter of factor l; p l2 Indicates the probability parameter of the factor l warning state; p l3 Indicates the alarm state probability parameter of factor l; When l = a, it is the bolt characteristic factor, μ l = μ a , σ l 2 = σ a 2 , p l1 = p a1 , p l2 = p a2 , p l3 = p a3 When A i ∈Q1, the bolt characteristic factor is in a normal state. When A i ∈Q2, the bolt characteristic factor is in a warning state. When A i ∈Q3, the bolt characteristic factor is in an alarm state; When l = b, it is the tower barrel characteristic factor, μ l = μ b , σ l 2 = σ b 2 , p l1 = p b1 , p l2 = p b2 , p l3 = p b3 When B j ∈Q1, the tower barrel characteristic factor is in a normal state. When B j ∈Q2, the tower barrel characteristic factor is in a warning state. When B j ∈Q3, the tower barrel characteristic factor is in an alarm state.

5. The intelligent condition monitoring system for the wind turbine tower barrel with autonomous learning characteristics according to claim 1, characterized in that The condition diagnosis factor is extracted and the tower barrel condition diagnosis standard is obtained in the following manner: Obtain the state diagnosis factor P according to the obtained state interval l : Where When l = a, it is the bolt state diagnosis factor P a ; When l = b, it is the tower barrel state diagnosis factor P b . The obtained condition diagnosis factors are simulated and integrated to obtain the operating condition diagnosis factor P: Where p 11 is the probability that the tower barrel state remains normal; p 12 is the probability that the tower barrel state changes from normal to warning; p 13 The probability that the tower barrel state changes from normal to alarm; p 22 The probability of maintaining the warning for the tower barrel state p 23 is the probability that the tower barrel state changes from early warning to alarm. The obtained state transition factor P is used as the operating condition diagnosis factor of the tower barrel. If the change rate of the operating diagnosis factor exceeds M%, it is considered that the tower barrel condition has changed.

6. The intelligent condition monitoring system for the wind turbine tower barrel with autonomous learning characteristics according to claim 1, characterized in that The condition feedback of the diagnosis method is carried out in the following manner: G1: Collect the change rate M of the diagnostic factors with inaccurate judgment, with a total of m; g , a total of m; G2: Calculate the mean μ of the change rate of the diagnostic factor g , that is G3: Calculate the variance σ of the change rate of the diagnostic factor g 2 , that is G4: According to the on-site feedback, let M′ = M + σ g 2 or M′ = M - σ g 2 as the change rate of the new diagnostic factor, thereby achieving state feedback.

7. The intelligent condition monitoring system for the wind turbine tower barrel with autonomous learning characteristics according to claim 1, characterized in that The digital feedback of the diagnosis method is carried out in the following manner: F1: Collect inaccurate parameter A for judgment i , B j , F2: Randomly select 3 centers from A i and denote them as μ f (f = 1, 2, 3); F3: Calculate A i objective function F4: Repeat step F2 until the minimum value of the objective function is obtained, i.e., F5: The central value of each cluster is obtained through iterative calculation, and the boundary value H is obtained through back-calculation av , where v = 1, 2, 3. That is, when v = 1, it represents the normal state; when v = 2, it represents the warning state; when v = 3, it represents the alarm state F6: According to the boundary value H av The iteration state probability parameter, namely F7: Similarly, parameter B j is processed through steps F2 - F6 to obtain the probability parameter after boundary value iteration, thereby achieving digital feedback.

Images