A wind turbine generator chamber cleaning anomaly identification method, system and computer equipment

Abstract

The present invention belongs to the technical field of wind turbine generator bore sweep anomaly identification, and specifically relates to a wind turbine generator bore sweep anomaly identification method, which uses three conditions to determine whether there is a bore sweep anomaly: 1) using the average wind speed before starting to determine whether the generator has an anomaly such as a bore sweep; 2) determining whether the generator has an anomaly such as a bore sweep by comparing the temperature rise of the generator under each power segment during the power generation process; 3) determining whether the generator has an anomaly such as a bore sweep by comparing the dirtiness of the inlet and outlet of the generator cooling system. By using big data analysis methods to conduct a longitudinal comparison of the entire field of wind turbines over a period of operation, it is effectively identified that the unit has a generator bore sweep risk, which can greatly improve accuracy. The data used are the unit's conventional SCADA operation data and the conventional cameras already in the cabin, and no additional sensor information needs to be installed, which has good effectiveness.

Description

Technical Field

[0001] The present invention belongs to the technical field of wind turbine bore cleaning anomaly recognition, and in particular relates to a wind turbine bore cleaning anomaly recognition method, system and computer equipment. Background Art

[0002] The generator is a core component of a wind turbine. When wind blows the impeller, it drives the gearbox and the generator rotor connected to the gearbox through the bearings. The rotor then establishes a torque transmission relationship with the stator through the magnetic field, thereby converting kinetic energy into electrical energy. As generators become larger and larger, the diameter of the generator is getting larger and larger. To reduce magnetic resistance and save costs, the design gap between the stator and rotor has also tended to gradually decrease. If the bearings are misaligned or the load impact causes eccentricity or structural deformation, the air gap between the stator and rotor will decrease. In extreme cases, the stator and rotor surfaces will rub against each other, which is the so-called bore scraping phenomenon.

[0003] Boring has become a major cause of damage to major generator components. Because the gap between the stator and rotor exists within the motor, even if it does occur, it can be difficult for maintenance personnel to detect. Over time, this can cause rotor structure loss, insulation wear, and eventually insulation breakdown. In severe cases, this can lead to fires and other serious safety incidents.

[0004] After searching, for example, patent documents CN207382074U - A single-bearing wind turbine stator and rotor anti-sweeping structure, CN201536255U - Generator, motor sweeping protector, CN201515212U - Motor or generator barrel protection monitoring device, their solutions are mainly to add additional equipment for anti-sweeping protection.

[0005] Currently, there are few methods for motor equipment bore scanning and identification, and there is no way to identify and protect existing batch-operated units. Summary of the Invention

[0006] The purpose of the present invention is to provide a method, system and computer equipment for identifying wind turbine chamber cleaning anomalies, which solve the problem that batch-operated units have no way to be identified and protected.

[0007] The present invention is achieved through the following technical solutions:

[0008] A method for identifying abnormalities in a wind turbine sweeping chamber comprises the following steps:

[0009] S1. Acquire real-time monitoring data and unit operation data of the target wind farm and perform data cleaning. The cleaned data includes startup data of each unit, temperature rise data of the generator under power generation status, and images of the air inlet and outlet of the generator cooling system;

[0010] S2. Input the startup data of each unit into the constructed average wind speed model at the startup time to calculate the startup average wind speed Viavg of each unit and the startup average wind speed Vavg of all units in the site;

[0011] The generator temperature rise data of each unit in the power generation state is input into the constructed generator temperature rise calculation model to calculate the average generator temperature rise Tki in each power compartment of each unit and the average generator temperature rise Tkavg in each power section of all units in the site;

[0012] Input the air inlet and outlet images of the generator cooling system of each unit into the image recognition model to identify the number of dirty photos Nbi and the total number of photos Ni;

[0013] Where i = 1, 2...N, i represents the unit number; k = 1, 2...K, k represents the number of compartments;

[0014] S3. The determination of the chamber cleaning abnormality is determined by the following three conditions:

[0015] Condition 1: Compare the average startup wind speed Viavg of each unit with the average startup wind speed Vavg of all units in the site. If the wind speed exceeds the threshold, condition 1 is considered to be met.

[0016] Condition 2: Compare the average temperature rise Tki of each generator in each power section of each unit with the average temperature rise Tavg of all generators in the entire site. If the number of compartments exceeding the temperature rise threshold exceeds the preset number of compartments, then condition 2 is considered to be met.

[0017] Condition 3: Based on the number of dirty photos Nbi and the total number of photos Ni, calculate the ratio of negative samples to total samples for each unit. If the ratio exceeds the set threshold, it is considered that condition 3 is met;

[0018] When condition 1 or condition 2 is met, and condition 3 is met at the same time, it is determined that the wind turbine generator chamber cleaning is abnormal.

[0019] Furthermore, in S1, the cleaning of the unit operation data specifically includes: deleting duplicate values, supplementing missing values, and / or eliminating interrupted data;

[0020] The specific cleaning of real-time monitoring data is to remove blurry images or images with abnormal photo locations.

[0021] Furthermore, in S2, the startup data of each unit is input into the constructed average wind speed model at the startup time to calculate the startup average wind speed Viavg of each unit and the startup average wind speed Vavg of all units in the site; specifically:

[0022] 2.11) Identify the status of each unit's operating data and identify the data marked as startup;

[0023] 2.12) Extract the wind speed sequence of all units in the site within tmin before each start-up time, and calculate the average wind speed V i of the units during this tmin. j , i=1,2...N, i represents the unit number; j=1,2...M, j represents the number of startups;

[0024] 2.13) Average startup wind speed V i for all units on site j Sort, remove x% maximum and X% minimum values, and retain qualified Vi j ;

[0025] 2.14) Calculate the average starting wind speed Viavg of each unit based on the data after elimination, and then calculate the qualified Vi j Take the average value to get the average starting wind speed Vavg of all the units in the site.

[0026] Furthermore, in S2, the generator temperature rise data of each unit in the power generation state is input into the constructed generator temperature rise calculation model to calculate the average generator temperature rise Tki in each power compartment of each unit and the average generator temperature rise Tkavg in each power section of all units in the site; specifically:

[0027] 2.21) Identify the power generation state and the start-up state, extract the temperature rise data of the generators of all units in the site under the power generation state, and record the power generation power P corresponding to the temperature rise of the generators;

[0028] 2.22) Divide the temperature rise into bins based on power generation: Divide the power range 0-Pr into k intervals, where Pr is the rated power;

[0029] Calculate the average generator temperature rise Tki in each power compartment of each unit;

[0030] 2.23) Based on Tki obtained in 2.22), calculate the average temperature rise of the generators at each power range of all units in the field, and express it as T1avg, T2avg, ..., Tkavg;

[0031] Tkavg=(Tk1+Tk2+…+TkN) / N, where N is the unit number.

[0032] Furthermore, in S2, the images of the air inlet and outlet of the generator cooling system of each unit are input into the image recognition model to identify the number of dirty photos Nbi and the total number of photos Ni; specifically:

[0033] 2.31) Graphic classification: Classify the images of the generator cooling system inlet and outlet, and mark the time and unit;

[0034] 2.32) Graphics cleaning: Remove invalid images based on graphic recognition methods;

[0035] 2.33) Pattern Recognition: Using image recognition and machine learning methods, the system compares and identifies the degree of dirtiness of the generator cooling system's air inlet and outlet images;

[0036] 2.34) Identify the number of dirty photos Nbi and the total number of photos Ni.

[0037] Furthermore, in S3, the expression of condition 1 is: Viavg-Vavg>V1; V1 is the wind speed threshold.

[0038] Furthermore, in S3, the expression of condition 2 is: Tki-Tkavg>T0, where T0 is the temperature rise threshold;

[0039] And the number of compartments exceeding the temperature rise threshold T0 is greater than or equal to y, where y is the preset number of compartments.

[0040] Further, prompts are given based on the results of S3, specifically:

[0041] When condition 1 or condition 2 is met, and condition 3 is also met, an operation and maintenance prompt will be issued;

[0042] The threshold value in condition 3 is set to K0, K0 is set to the third gear, the ratios are K1, K2 and K3, and K3>K2>K1:

[0043] When K0 selects K1, it will report the abnormality of the chamber cleaning.

[0044] When K0 selects K2, an abnormal cleaning warning is issued;

[0045] When K0 selects K3, it reports that the chamber cleaning is abnormal and urgent.

[0046] A wind turbine generator chamber cleaning anomaly recognition system, comprising:

[0047] Data acquisition module, used to obtain real-time monitoring data and unit operation data of the target wind farm;

[0048] The data preprocessing module is used to clean real-time monitoring data and unit operation data. The cleaned data includes each unit's startup data, generator temperature rise data under power generation status, and images of the generator cooling system's inlet and outlet air vents.

[0049] The average wind speed model is used to calculate the startup data of each unit to obtain the average startup wind speed Viavg of each unit and the average startup wind speed Vavg of all units in the site;

[0050] The generator temperature rise calculation model is used to calculate the generator temperature rise data of each unit in the power generation state, and obtain the average generator temperature rise Tki of each power segment of each unit, and the average generator temperature rise Tavg of all units in the site;

[0051] The image recognition model is used to identify the images of the inlet and outlet of the generator cooling system of each unit, and identify the number of dirty photos Nbi and the total number of photos Ni;

[0052] Where i = 1, 2...N, i represents the unit number; k = 1, 2...K, k represents the number of compartments;

[0053] The first judgment module is used to compare the average startup wind speed Viavg of each unit with the average startup wind speed Vavg of all units in the site. If the wind speed exceeds the threshold, it is determined that condition 1 is met;

[0054] The second judgment module is used to compare the average temperature rise Tki of the generator in each power section of each unit with the average temperature rise Tkavg of the generators of all units in the site. If the number of compartments exceeding the temperature rise threshold exceeds the preset number of compartments, it is determined that condition 2 is met;

[0055] The third judgment module is used to calculate the ratio of negative samples to total samples of each unit based on the number of dirty photos Nbi and the total number of photos Ni. If the ratio exceeds the set threshold, it is determined that condition 3 is met;

[0056] When condition 1 or condition 2 is met, and condition 3 is met at the same time, it is determined that the wind turbine generator chamber cleaning is abnormal.

[0057] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the wind turbine generator bore cleaning anomaly identification method are implemented.

[0058] Compared with the prior art, the present invention has the following beneficial technical effects:

[0059] The present invention discloses a method for identifying abnormalities in the bore sweeping of wind turbines. Based on physical mechanisms, the method analyzes the phenomena after abnormal generator bore sweeping, and simultaneously judges the generator bore sweeping status from the specific performance during the startup process, the power generation process, and the entire operation. This method can greatly improve the accuracy. The data used are the conventional SCADA operation data of the unit and the conventional camera in the cabin. No additional sensor information is required, and the method has good effectiveness. The method uses a big data analysis method to conduct a longitudinal comparison of the entire wind turbine over a period of operation, effectively identifying the risk of generator bore sweeping in the unit. Whether the bore sweeping is abnormal is judged based on the following three conditions:

[0060] 1) Judgment condition 1: Use the average wind speed before starting to determine whether the generator has abnormalities such as sweeping;

[0061] 2) Judgment condition 2: by comparing the temperature rise of the generator at each power range during the power generation process, it is determined whether the generator has a chamber sweeping abnormality;

[0062] 3) Judgment condition 3: By comparing the contamination of the generator cooling system inlet and outlet, when the generator has a scuffed bore, the conformal coating on the inner surface of its stator and rotor will fall off and adhere to the generator cooling system outlet, directly manifesting as contamination.

[0063] The algorithm consists of three parts: 1. Checking whether the wind speed at startup is higher than that of similar units on site; 2. Checking whether the temperature rise of the generator during startup and power generation is higher than that of similar units on site; and 3. Using image recognition to identify abnormal contamination at the air outlet of the generator cooling system in the engine room. Based on these three conditions, if either condition 1 or 2 is met, and condition 3 is also met, it can be determined that the chamber has been scavenged.

[0064] Furthermore, the algorithm carried by this method can be transplanted into computer equipment to conduct regular inspections of all units in the field and issue early warning information to prompt operation and maintenance personnel to conduct further inspections and take action in advance to avoid the expansion of the bore-sweeping problem. For example, for units with minor bore-sweeping, alignment adjustments or fixed rotor structural parts can be made. For units with severe bore-sweeping, the tower needs to be removed and replaced to prevent further insulation breakdown or fire. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 The present invention is a flow chart of a method for identifying abnormalities in a wind turbine chamber. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solutions and advantages of the present invention more clear, the following is a further detailed description with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. That is, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments.

[0067] The components described and illustrated in the drawings and embodiments of the present invention may be arranged and designed in a variety of different configurations. Therefore, the detailed description of the embodiments of the present invention provided in the following drawings is not intended to limit the scope of the claimed invention, but merely represents a selected embodiment of the present invention. All other embodiments derived by those skilled in the art based on the drawings and embodiments of the present invention without inventive effort shall fall within the scope of protection of the present invention.

[0068] The present invention discloses a method for identifying abnormalities in a wind turbine generator's bore sweeping, comprising the following steps:

[0069] S1. Acquire real-time monitoring data and unit operation data of the target wind farm and perform data cleaning. The cleaned data includes startup data of each unit, temperature rise data of the generator under power generation status, and images of the air inlet and outlet of the generator cooling system;

[0070] S2. Input the startup data of each unit into the constructed average wind speed model at the startup time to calculate the startup average wind speed Viavg of each unit and the startup average wind speed Vavg of all units in the site;

[0071] The generator temperature rise data of each unit in the power generation state is input into the constructed generator temperature rise calculation model to calculate the average generator temperature rise Tki in each power compartment of each unit and the average generator temperature rise Tkavg in each power section of all units in the site;

[0072] Input the air inlet and outlet images of the generator cooling system of each unit into the image recognition model to identify the number of dirty photos Nbi and the total number of photos Ni;

[0073] Where i = 1, 2...N, i represents the unit number; k = 1, 2...M, k represents the number of compartments;

[0074] S3. The determination of the chamber cleaning abnormality is determined by the following three conditions:

[0075] Condition 1: Compare the average startup wind speed Viavg of each unit with the average startup wind speed Vavg of all units in the site. If the wind speed exceeds the threshold, condition 1 is considered to be met.

[0076] Condition 2: Compare the average temperature rise Tki of each generator in each power section of each unit with the average temperature rise Tavg of all generators in the entire site. If the number of compartments exceeding the temperature rise threshold exceeds the preset number of compartments, then condition 2 is considered to be met.

[0077] Condition 3: Based on the number of dirty photos Nbi and the total number of photos Ni, calculate the ratio of negative samples to total samples for each unit. If the ratio exceeds the set threshold, it is considered that condition 3 is met;

[0078] When condition 1 or 2 is met, if condition 3 is also met, it is determined that the wind turbine generator has an abnormality in its bore cleaning.

[0079] The features and performance of the present invention are further described in detail below with reference to the embodiments.

[0080] like Figure 1As shown, the present invention is mainly divided into four major steps: data preparation, model calculation, abnormality identification, and operation and maintenance prompts. It can be implemented in the field controller (i.e., the centralized control and maintenance center), and when problems are detected, it prompts on-site personnel to further check and perform operation and maintenance work.

[0081] 1. Data preparation

[0082] 1) Structured data preparation: Collect and store operational data for the target wind farm, requiring no less than one month of data to fully capture various operational states and external wind conditions, making the data more effective.

[0083] 2) Unstructured data preparation: Graphically capture the real-time monitoring data from the target wind farm's nacelle cameras. The capture frequency can be automatically defined, such as 10 images per day, and saved as images.

[0084] 3) The returned data and images are placed on the field server according to the specified location and number for storage;

[0085] 4) Data cleaning work, including deleting duplicate values, supplementing missing values, and removing interrupted data.

[0086] 2. Model calculation

[0087] The model calculation includes three parts: the average wind speed at startup, the temperature rise of the generator in the startup and power generation states, and the identification of contamination at the inlet and outlet of the generator cooling system in the cabin.

[0088] 2.1 Identification of average wind speed at startup

[0089] For units with bore scraping, due to friction between the stator and rotor, the starting torque is relatively large, and the corresponding starting wind speed will be lower than that of other units.

[0090] 1) Identify the startup status of each unit data. The operation data channel contains records of the standby status. Normal units are divided into several states: standby, startup, power generation, shutdown, and maintenance. Data marked as startup needs to be identified;

[0091] 2) Calculate the average wind speed V i of all units in the site within the first 10 minutes after each startup. j , i=1,2...N, i represents the unit number; j=1,2...M, j represents the number of starts. Note that the total number of starts for each unit is different.

[0092] For example, for Unit 1, a total of 100 startups are identified. The wind speed information within 10 minutes before each startup is extracted and averaged, that is, the average wind speed values of Unit 1 before the 100 startups are obtained, V11, V12, V13...V1100 , if unit 2# has 103 starts, then we get V21, V22, V23...V2 103 Similarly, we can get the average wind speed before each start-up of all units on the site. Assuming that the last unit has been started 89 times, the total number of starts of 50 units is 5000 times, then the Vi j The total number of times is 100+103+…+89=5000 times.

[0093] 3) Sort the starting wind speeds Vij of all units on site, remove the x% maximum value and the x% minimum value, and convert them into Vij'.

[0094] Assume V11 = 2.1 m / s, V12 = 3.5 m / s, ... V1 100 =3.2m / s; V21=4.8m / s, V22=0.2m / s…Sort these 5000 Vij values, find the maximum and minimum values that account for 1% (here x% is assumed to be 1%) and filter them out. After filtering out, 4900 data remain.

[0095] Remove extreme abnormal data caused by data or debugging reasons to avoid affecting the average wind speed at startup.

[0096] 4) Calculate the average startup wind speed of each unit based on these 4900 data. Since abnormal data has been filtered out in step 3), the startup times of each unit have changed. For example, the startup wind speeds of unit 1# are V11, V13…V1 100 , because V12 is too large and has been screened out, the starting wind speed value of unit 1# is less than 100, but only 99. Then take the average starting wind speed of unit 1#, and get V1avg=(V11+V13+V14……+V1 100 ) / 99, and similarly, we can get the average starting wind speeds of other units V2avg, V3avg, ... V50avg (get 50 values, each representing the average starting wind speed of a unit). Then we can calculate the average starting wind speed of the 4900 Vi j Take the average value to get the average wind speed information Vavg for the whole field.

[0097] 2.2 Calculation of generator temperature rise in power generation state

[0098] During power generation, the friction between the stator and rotor generates heat. Therefore, under the same power operation conditions, the temperature rise of its bearings is higher than that of other units. Similarly, during the startup process, even if it is not generating electricity, the temperature of its generator is also higher.

[0099] 1) Identify the power generation state and the startup state, extract and calculate the generator temperature rise (generator temperature - ambient temperature) of all units in the site under the power generation state, and record the power generation power P corresponding to the temperature rise;

[0100] 2) Divide the temperature rise into bins according to the generated power, and calculate the average generator temperature rise Ti k avg for each power segment of each unit, where i = 1, 2 ... N, i represents the unit number; k = 1, 2 ... K, k represents the number of bins;

[0101] Since the power range is 0-Pr (Pr is the rated power, which is also the maximum power), for example, taking 10% of Pr as a bin, the average bearing temperature rise in the bin is calculated. For example, the average generator temperature rise corresponding to the intervals of 0-10% Pr, 10%-20% Pr...90%-100% Pr is T1i, T2i,...T10i. That is, 10 quantities are calculated for each unit. Taking 50 units as an example, 500 temperature variables can be calculated, corresponding to the average generator temperature rise of each unit in each power range;

[0102] 3) Calculate the average temperature rise of the generators of all units in each power range:

[0103] T1avg=(T11+T12+…+T1N) / N, where N is the unit number;

[0104] T2avg=(T21+T22+…+T2N) / N, where N is the unit number;

[0105] Similarly, we can obtain T10avg=(T101+T102+…+T10N) / N.

[0106] 2.3 Identification of dirt at the inlet and outlet of the generator cooling system

[0107] Perform image recognition to identify whether there is abnormal dirt at the air outlet of the engine room generator cooling system;

[0108] 1) Graphic classification: classify the returned data and mark the time and unit;

[0109] 2) Image cleaning: Using image recognition technology, remove invalid images such as blurry images and images with abnormal photo positions;

[0110] 3) Pattern recognition: Using image recognition technology and machine learning methods, the degree of dirtiness of a specific photographed area (here, the air inlet and outlet of the generator cooling system) is compared and identified.

[0111] 4) Record the number of dirty photos Nbi (negative samples) and the total number of photos Ni, (i = 1, 2...N, i represents the group number).

[0112] When the machine learning method is learning, it scores each unit based on its level of dirtiness and selects images of the most contaminated units to be pushed to the operation and maintenance personnel for identification and judgment. The results are manually fed back and the algorithm in the machine learning method is readjusted until the algorithm is mature.

[0113] 2. Abnormal identification

[0114] The determination of bearing abnormality requires a combined judgment based on three conditions:

[0115] 1) Condition 1: Compare the average wind speed for each unit at startup obtained from the above calculation results with the average wind speed for the entire site. If it exceeds a certain threshold V1, then condition 1 is considered to be met:

[0116] Viavg-Vavg>V1

[0117] 2) Condition 2: Compare the average temperature rise of each unit's generator obtained from the above calculation results with the average temperature rise of all generators in the field. Each bin is required to be compared separately, that is, whether the difference between Tki and Tkavg exceeds T0 (k = 1, 2, ... 10 bins). If 7 bins meet the requirement that the difference exceeds T0 (here assuming the preset number of bins = 5, that is, 7 out of 10 bin temperatures exceed T0), then condition 2 is considered to be met.

[0118] 3) Condition 3: Based on the identification rate of dirt at the inlet and outlet of the generator cooling system, the ratio of negative samples to total samples of each unit is calculated, Ki = Nbi / Ni. If Ki exceeds a certain threshold K0, condition 3 is considered to be met.

[0119] 3. Operation and maintenance tips

[0120] 1) When one of conditions 1 and 2 is met, if condition 3 is also met, an operation and maintenance prompt will be given.

[0121] 2) The threshold K0 in condition 3 can be set to three levels: K1, K2, and K3, and K3>K2>K1:

[0122] When K0 selects K1, it will report the abnormality of the chamber cleaning.

[0123] When K0 selects K2, an abnormal cleaning warning is issued;

[0124] When K0 selects K3, it will report the emergency of chamber cleaning abnormality;

[0125] On-site operation and maintenance are carried out based on these three levels of information prompts.

[0126] The present invention also discloses a wind turbine generator chamber cleaning anomaly recognition system, comprising:

[0127] Data acquisition module, used to obtain real-time monitoring data and unit operation data of the target wind farm;

[0128] The data preprocessing module is used to clean real-time monitoring data and unit operation data. The cleaned data includes each unit's startup data, generator temperature rise data under power generation status, and images of the generator cooling system's inlet and outlet air vents.

[0129] The average wind speed model is used to calculate the startup data of each unit to obtain the average startup wind speed Viavg of each unit and the average startup wind speed Vavg of all units in the site;

[0130] The generator temperature rise calculation model is used to calculate the generator temperature rise data of each unit in the power generation state, and obtain the average generator temperature rise Tki of each power segment of each unit, and the average generator temperature rise Tavg of all units in the site;

[0131] The image recognition model is used to identify the images of the inlet and outlet of the generator cooling system of each unit, and identify the number of dirty photos Nbi and the total number of photos Ni;

[0132] Where i = 1, 2...N, i represents the unit number; K = 1, 2...M, K represents the number of compartments;

[0133] The first judgment module is used to compare the average startup wind speed Viavg of each unit with the average startup wind speed Vavg of all units in the site. If the wind speed exceeds the threshold, it is determined that condition 1 is met;

[0134] The second judgment module is used to compare the average temperature rise Tki of the generator in each power section of each unit with the average temperature rise Tkavg of the generators of all units in the site. If the number of compartments exceeding the temperature rise threshold exceeds the preset number of compartments, it is determined that condition 2 is met;

[0135] The second judgment module is used to calculate the ratio of negative samples to total samples of each unit based on the number of dirty photos Nbi and the total number of photos Ni. If the ratio exceeds the set threshold, it is determined that condition 3 is met;

[0136] When any one of the conditions is met, it is determined that the wind turbine generator chamber cleaning is abnormal.

[0137] In an exemplary embodiment, a computer device is further provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the wind turbine generator bore sweeping anomaly identification method when executing the computer program. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0138] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A method for identifying abnormalities in a wind turbine sweeping chamber, characterized in that: The following steps are involved: S1. Acquire real-time monitoring data and unit operation data of the target wind farm and perform data cleaning. The cleaned data includes startup data of each unit, temperature rise data of the generator under power generation status, and images of the air inlet and outlet of the generator cooling system; S2. Input the startup data of each unit into the constructed average wind speed model at the startup time to calculate the startup average wind speed Viavg of each unit and the startup average wind speed Vavg of all units in the site; The generator temperature rise data of each unit in the power generation state is input into the constructed generator temperature rise calculation model to calculate the average generator temperature rise Tki in each power compartment of each unit and the average generator temperature rise Tkavg in each power section of all units in the site; Input the air inlet and outlet images of the generator cooling system of each unit into the image recognition model to identify the number of dirty photos Nbi and the total number of photos Ni; Where i = 1, 2...N, i represents the unit number; k = 1, 2...K, k represents the number of compartments; S3. The determination of the chamber cleaning abnormality is determined by the following three conditions: Condition 1: Compare the average startup wind speed Viavg of each unit with the average startup wind speed Vavg of all units in the site. If the wind speed exceeds the threshold, condition 1 is considered to be met. Condition 2: Compare the average temperature rise Tki of each generator in each power section of each unit with the average temperature rise Tavg of all generators in the entire site. If the number of compartments exceeding the temperature rise threshold exceeds the preset number of compartments, then condition 2 is considered to be met. Condition 3: Based on the number of dirty photos Nbi and the total number of photos Ni, calculate the ratio of negative samples to total samples for each unit. If the ratio exceeds the set threshold, it is considered that condition 3 is met; When condition 1 or condition 2 is met, and condition 3 is met at the same time, it is determined that the wind turbine generator chamber cleaning is abnormal.

2. A method for identifying abnormalities in a wind turbine sweeping chamber according to claim 1, characterized in that: In S1, the cleaning of unit operation data specifically includes: deleting duplicate values, supplementing missing values and / or eliminating interrupted data; The specific cleaning of real-time monitoring data is to remove blurry images or images with abnormal photo locations.

3. The method for identifying abnormalities in a wind turbine sweeping chamber according to claim 1, characterized in that: In S2, the startup data of each unit is input into the average wind speed model constructed at the startup time, and the startup average wind speed Viavg of each unit and the startup average wind speed Vavg of all units in the site are calculated; specifically: 2.11) Identify the status of each unit's operating data and identify the data marked as startup; 2.12) Extract the wind speed sequence of all units in the site within tmin before each start-up time, and calculate the average wind speed V i of the units during this tmin. j , i=1,2...N, i represents the unit number; j=1,2...M, j represents the number of startups; 2.13) Average starting wind speed V i for all units on site j Sort, remove x% maximum and X% minimum values, and retain qualified Vi j ; 2.14) Calculate the average starting wind speed Viavg of each unit based on the data after elimination, and then calculate the qualified Vi j Take the average value to get the average starting wind speed Vavg of all the units in the site.

4. The method for identifying abnormalities in a wind turbine sweeping chamber according to claim 1, characterized in that: In S2, the generator temperature rise data of each unit in the power generation state is input into the constructed generator temperature rise calculation model to calculate the average generator temperature rise Tki in each power compartment of each unit and the average generator temperature rise Tkavg in each power section of all units in the site; specifically: 2.21) Identify the power generation state and the start-up state, extract the temperature rise data of the generators of all units in the site under the power generation state, and record the power generation power P corresponding to the temperature rise of the generators; 2.22) Divide the temperature rise into bins based on power generation: Divide the power range 0-Pr into k intervals, where Pr is the rated power; Calculate the average generator temperature rise Tki in each power compartment of each unit; 2.23) Based on Tki obtained in 2.22), calculate the average temperature rise of the generator at each power range of the entire unit, and express it as T1avg, T2avg, ..., Tkavg; Tkavg=(Tk1+Tk2+…+TkN) / N, where N is the unit number.

5. The method for identifying abnormalities in a wind turbine sweeping chamber according to claim 1, characterized in that: In S2, the images of the air inlet and outlet of the generator cooling system of each unit are input into the image recognition model to identify the number of dirty photos Nbi and the total number of photos Ni; specifically: 2.31) Graphic classification: Classify the images of the generator cooling system inlet and outlet, and mark the time and unit; 2.32) Graphics cleaning: Remove invalid images based on graphic recognition methods; 2.33) Pattern Recognition: Using image recognition and machine learning methods, the system compares and identifies the degree of dirtiness of the generator cooling system's air inlet and outlet images; 4) Identify the number of dirty photos Nbi and the total number of photos Ni.

6. The method for identifying abnormalities in a wind turbine sweeping chamber according to claim 1, characterized in that: In S3, the expression of condition 1 is: Viavg-Vavg>V1; V1 is the wind speed threshold.

7. The method for identifying abnormalities in a wind turbine sweeping chamber according to claim 1, characterized in that: In S3, the expression of condition 2 is: Tki-Tkavg>T0, where T0 is the temperature rise threshold; And the number of compartments exceeding the temperature rise threshold T0 is greater than or equal to y, where y is the preset number of compartments.

8. The method for identifying abnormalities in a wind turbine sweeping chamber according to claim 1, characterized in that: Prompt based on the results of S3, specifically: When condition 1 or condition 2 is met, and condition 3 is also met, an operation and maintenance prompt will be issued; The threshold value in condition 3 is set to K0, K0 is set to the third gear, the ratios are K1, K2 and K3, and K3>K2>K1: When K0 selects K1, it will report the abnormality of the chamber cleaning; When K0 selects K2, an abnormal cleaning warning will be issued; When K0 selects K3, it reports that the chamber cleaning is abnormal and urgent.

9. A wind turbine generator sweeping abnormality identification system, characterized in that: include: Data acquisition module, used to obtain real-time monitoring data and unit operation data of the target wind farm; The data preprocessing module is used to clean real-time monitoring data and unit operation data. The cleaned data includes each unit's startup data, generator temperature rise data under power generation status, and images of the generator cooling system's inlet and outlet air vents. The average wind speed model is used to calculate the startup data of each unit to obtain the average startup wind speed Viavg of each unit and the average startup wind speed Vavg of all units in the site; The generator temperature rise calculation model is used to calculate the generator temperature rise data of each unit in the power generation state, and obtain the average generator temperature rise Tki of each power segment of each unit, and the average generator temperature rise Tavg of all units in the site; The image recognition model is used to identify the images of the inlet and outlet of the generator cooling system of each unit, and identify the number of dirty photos Nbi and the total number of photos Ni; Where i = 1, 2...N, i represents the unit number; k = 1, 2...K, K represents the number of compartments; The first judgment module is used to compare the average startup wind speed Viavg of each unit with the average startup wind speed Vavg of all units in the site. If the wind speed exceeds the threshold, it is determined that condition 1 is met; The second judgment module is used to compare the average temperature rise Tki of the generator in each power section of each unit with the average temperature rise Tkavg of the generators of all units in the site. If the number of compartments exceeding the temperature rise threshold exceeds the preset number of compartments, it is determined that condition 2 is met; The third judgment module is used to calculate the ratio of negative samples to total samples of each unit based on the number of dirty photos Nbi and the total number of photos Ni. If the ratio exceeds the set threshold, it is determined that condition 3 is met; When condition 1 or condition 2 is met, and condition 3 is met at the same time, it is determined that the wind turbine generator chamber cleaning is abnormal.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the wind turbine generator bore cleaning abnormality identification method according to any one of claims 1 to 8 are implemented.

Images