A wind turbine gearbox filter blockage early warning diagnosis method and system

Abstract

The application provides a wind turbine gearbox filter blockage early warning diagnosis method and system, including the following steps: step 1, obtaining historical data of a to-be-tested wind turbine; step 2, calculating a filter pressure difference threshold; step 3, obtaining real-time operation data of the to-be-tested wind turbine; step 4, calculating a real-time residual mean of the filter pressure difference; and step 5, judging whether the to-be-tested wind turbine gearbox filter is abnormal according to the obtained real-time residual mean and the obtained filter pressure difference threshold; the application can find abnormalities in advance, reduce filter element damage probability and reduce spare parts consumption.

Description

Technical Field

[0001] This invention belongs to the field of wind turbine gearbox filter blockage early warning technology, specifically relating to a wind turbine gearbox filter blockage early warning diagnosis method and system. Background Technology

[0002] Intelligent operation and maintenance of wind farms has become a development trend in the industry, and the ability to detect potential fault risks in advance through unit operation data is currently a hot topic in the industry.

[0003] As a crucial component of the transmission system of wind turbines, the gearbox's auxiliary system is also very important. Gearbox filter failure often leads to a continuous increase in gearbox lubricating oil temperature, ultimately resulting in shutdown and power generation loss. Severely clogged gearbox filters can also damage the gearbox oil pump.

[0004] Currently, there is no reliable early warning system for gearbox filters in the industry. Usually, it is done through regular manual inspections, and by the time abnormalities are discovered, the filter element has already been damaged to varying degrees. Summary of the Invention

[0005] The purpose of this invention is to provide a method for early warning and diagnosis of gearbox filter blockage in wind turbines, which solves the problem of the lag in existing gearbox fault early warning systems.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] The present invention provides a method for early warning and diagnosis of gearbox filter blockage in wind turbines, comprising the following steps:

[0008] Step 1: Obtain historical data of the wind turbine under test. The historical data includes gearbox outlet oil temperature, gearbox filter inlet pressure, gearbox filter outlet pressure, and filter differential pressure.

[0009] Step 2: Divide the obtained historical data into four datasets;

[0010] Step 3: Model the four datasets respectively, and calculate the filter differential pressure threshold for each dataset based on the model.

[0011] Step 4: Obtain the real-time operating data of the wind turbine under test. The real-time operating data includes the real-time outlet oil temperature of the gearbox, the real-time inlet pressure of the gearbox filter, and the real-time outlet pressure of the gearbox filter.

[0012] Step 5: Divide the obtained real-time running data according to Step 2 to obtain four datasets;

[0013] Step 6: Using the model corresponding to each dataset obtained in Step 3, calculate the mean real-time residual of the filter pressure difference for each of the four datasets of real-time running data.

[0014] Step 7: Based on the real-time residual mean obtained in Step 6 and the filter differential pressure threshold obtained in Step 3, determine whether the gearbox filter of the wind turbine under test is abnormal.

[0015] In a preferred embodiment, in step 2, the obtained historical data is divided into four datasets. The specific method is as follows:

[0016] Based on the threshold values ​​of the fine filter bypass pressure valve and the thermal valve of the gearbox filter in the wind turbine under test, and combined with the gearbox outlet oil temperature and the gearbox filter inlet pressure, the historical data is divided into four datasets.

[0017] A preferred embodiment involves dividing the historical data into four datasets, specifically as follows:

[0018] The first historical dataset is formed by extracting data from historical data where the gearbox filter inlet pressure is less than the threshold of the fine filter bypass pressure valve and the gearbox outlet oil temperature is less than the threshold of the thermal valve.

[0019] Data from historical data showing that the gearbox filter inlet pressure is greater than the threshold of the fine filter bypass pressure valve and the gearbox outlet oil temperature is less than the threshold of the thermal valve are used to form a second historical dataset.

[0020] Data from historical data showing that the gearbox filter inlet pressure is greater than the threshold of the fine filter bypass pressure valve and the gearbox outlet oil temperature is greater than the threshold of the thermal valve are used to form a third historical dataset.

[0021] The fourth historical dataset is formed by extracting data from historical data where the gearbox filter inlet pressure is greater than the threshold of the fine filter bypass pressure valve and the gearbox outlet oil temperature is greater than the threshold of the thermal valve.

[0022] In a preferred embodiment, in step 3, the four obtained datasets are modeled separately, specifically by:

[0023] Univariate linear regression was performed on all gearbox filter inlet pressures and all filter differential pressures in the first and second historical datasets, respectively, to obtain the first and second models.

[0024] Normalize all gearbox outlet oil temperatures, all gearbox filter inlet pressures, and all filter differential pressures in the third and fourth historical datasets respectively to obtain the normalized gearbox outlet oil temperatures, gearbox filter inlet pressures, and filter differential pressures.

[0025] Multiple linear regression fitting was performed on the normalized gearbox outlet oil temperature, gearbox filter inlet pressure, and filter differential pressure in the third and fourth historical datasets, respectively, to obtain the third and fourth models.

[0026] In a preferred embodiment, in step 3, the filter differential pressure threshold corresponding to each dataset is calculated based on the obtained model. The specific method is as follows:

[0027] The method for calculating the filter differential pressure threshold corresponding to the first historical dataset is as follows:

[0028] By combining the inlet pressure of each gearbox filter in the first historical dataset with the first model, an estimated value of the filter differential pressure is calculated.

[0029] The residual corresponding to the pressure difference of each filter is calculated based on the estimated value of the filter pressure difference.

[0030] The first filter pressure difference threshold corresponding to the first historical dataset is calculated based on the residual corresponding to each filter pressure difference.

[0031] The method for calculating the filter differential pressure threshold corresponding to the second historical dataset is as follows:

[0032] By combining the inlet pressure of each gearbox filter in the second historical dataset with the second model, an estimate of the filter differential pressure is calculated.

[0033] The residual corresponding to the pressure difference of each filter is calculated based on the estimated value of the filter pressure difference.

[0034] The second filter pressure difference threshold corresponding to the second historical dataset is calculated based on the residual corresponding to each filter pressure difference.

[0035] The method for calculating the filter differential pressure threshold corresponding to the third historical dataset is as follows:

[0036] Normalize all gearbox filter inlet pressure, gearbox outlet oil temperature and gearbox filter outlet pressure in the third historical dataset to obtain normalized gearbox filter inlet pressure, gearbox outlet oil temperature and gearbox filter outlet pressure.

[0037] The estimated value of the filter differential pressure is calculated by combining the normalized gearbox filter inlet pressure and the normalized gearbox outlet oil temperature with the third model.

[0038] The residual corresponding to each filter pressure difference is calculated based on the estimated value of the filter pressure difference and the normalized gearbox filter outlet pressure.

[0039] The third filter pressure difference threshold corresponding to the third historical dataset is calculated based on the residual corresponding to each filter pressure difference.

[0040] The method for calculating the filter differential pressure threshold corresponding to the fourth historical dataset is as follows:

[0041] Normalize all gearbox filter inlet pressure, gearbox outlet oil temperature and gearbox filter outlet pressure in the fourth historical dataset to obtain normalized gearbox filter inlet pressure, gearbox outlet oil temperature and gearbox filter outlet pressure.

[0042] The estimated value of the filter differential pressure is calculated by combining the normalized gearbox filter inlet pressure and the normalized gearbox outlet oil temperature with the fourth model.

[0043] The residual corresponding to each filter pressure difference is calculated based on the estimated value of the filter pressure difference and the normalized gearbox filter outlet pressure.

[0044] The fourth filter pressure threshold corresponding to the fourth historical dataset is calculated based on the residual corresponding to each filter pressure difference.

[0045] In a preferred embodiment, in step 6, using the model corresponding to each dataset obtained in step 3, the average real-time residual of the filter pressure difference corresponding to each of the four datasets of real-time running data is calculated respectively. Specifically:

[0046] The four datasets are the first real-time dataset, the second real-time dataset, the third real-time dataset, and the fourth real-time dataset;

[0047] The method for calculating the mean real-time residual of the filter pressure difference corresponding to the first real-time dataset is as follows:

[0048] By combining the real-time inlet pressure of each gearbox filter in the first real-time dataset with the first model, the real-time estimate of the filter pressure difference is calculated.

[0049] The real-time residual corresponding to the pressure difference of each filter is calculated based on the real-time estimated value of the filter pressure difference;

[0050] The mean real-time residual of the first filter pressure difference corresponding to the first real-time dataset is calculated based on the real-time residual corresponding to each filter pressure difference.

[0051] The method for calculating the mean real-time residual of the filter pressure difference corresponding to the second real-time dataset is as follows:

[0052] By combining the real-time inlet pressure of each gearbox filter in the second real-time dataset with the second model, the real-time estimate of the filter differential pressure is calculated.

[0053] The real-time residual corresponding to the pressure difference of each filter is calculated based on the real-time estimated value of the filter pressure difference;

[0054] The mean real-time residual of the second filter pressure difference corresponding to the second real-time dataset is calculated based on the real-time residual corresponding to each filter pressure difference.

[0055] The method for calculating the mean real-time residual of the filter pressure difference corresponding to the third real-time dataset is as follows:

[0056] The real-time inlet pressure, real-time outlet oil temperature and real-time outlet pressure of each gearbox filter in the third real-time dataset are normalized to obtain the normalized real-time inlet pressure, real-time outlet oil temperature and real-time outlet pressure of the gearbox filter.

[0057] By combining the real-time inlet pressure of each normalized gearbox filter and the real-time outlet oil temperature of each normalized gearbox with the third model, the real-time estimate of the filter differential pressure is calculated.

[0058] The real-time residual corresponding to the pressure difference of each filter is calculated based on the real-time estimated value of the filter pressure difference and the normalized real-time outlet pressure of the gearbox filter.

[0059] The mean real-time residual of the third filter pressure corresponding to the third real-time dataset is calculated based on the real-time residual corresponding to each filter pressure difference.

[0060] The method for calculating the mean real-time residual of the filter pressure difference corresponding to the fourth real-time dataset is as follows:

[0061] The real-time inlet pressure, real-time outlet oil temperature and real-time outlet pressure of each gearbox filter in the fourth real-time dataset are normalized to obtain the normalized real-time inlet pressure, real-time outlet oil temperature and real-time outlet pressure of the gearbox filter.

[0062] By combining the real-time inlet pressure of each normalized gearbox filter and the real-time outlet oil temperature of each normalized gearbox with the fourth model, the real-time estimate of the filter differential pressure is calculated.

[0063] The real-time residual corresponding to the pressure difference of each filter is calculated based on the real-time estimated value of the filter pressure difference and the normalized real-time outlet pressure of the gearbox filter.

[0064] The mean real-time residual of the fourth filter pressure corresponding to the fourth real-time dataset is calculated based on the real-time residual corresponding to each filter pressure difference.

[0065] In a preferred embodiment, in step 7, based on the average real-time residual value obtained in step 6 and the filter differential pressure threshold obtained in step 3, it is determined whether the gearbox filter of the wind turbine under test is abnormal. The specific method is as follows:

[0066] like This indicates that the coarse filter element is clogged;

[0067] like and If the coarse filter cartridge is blocked, the fine filter cartridge may also be blocked.

[0068] like and The fine filter cartridge is clogged, while the coarse filter cartridge is not clogged;

[0069] like and The coarse filter element is not clogged, and the fine filter element is not clogged.

[0070] This preferred embodiment provides a wind turbine gearbox filter blockage early warning and diagnostic system, including:

[0071] The historical data acquisition unit is used to acquire historical data of the wind turbine under test, including gearbox outlet oil temperature, gearbox filter inlet pressure, gearbox filter outlet pressure, and filter differential pressure; and to acquire real-time operating data of the wind turbine under test, including real-time gearbox outlet oil temperature, real-time gearbox filter inlet pressure, and real-time gearbox filter outlet pressure.

[0072] The data partitioning unit is used to divide the obtained historical data and real-time running data into four corresponding datasets.

[0073] The threshold calculation unit is used to model the four datasets respectively and calculate the filter differential pressure threshold for each dataset based on the model.

[0074] The residual mean calculation unit uses the model corresponding to each dataset to calculate the real-time residual mean of the filter pressure difference corresponding to the four datasets of real-time running data.

[0075] The anomaly detection unit is used to determine whether the gearbox filter of the wind turbine under test is abnormal based on the real-time residual average value of the obtained filter differential pressure and the obtained filter differential pressure threshold.

[0076] Compared with the prior art, the beneficial effects of the present invention are:

[0077] A method for warning and diagnosing the blockage of a gearbox filter in a wind turbine provided by the present invention establishes a mathematical model under different operating conditions according to the working principle of the oil-water cooling system of the gearbox and in combination with historical data. Then, the data to be measured for the unit in the near future is obtained, classified according to the operating conditions, and respectively compared with the mathematical model established from the historical data under this condition through calculation, accurately positioning the blocked part of the gearbox filter and giving a prediction result. Since the present invention performs real-time calculation and monitoring, it can thus detect abnormalities in advance, reduce the probability of filter element damage, and reduce the consumption of spare parts. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 It is a schematic diagram of the oil-water cooling of the gearbox;

[0079] Figure 2 It is a flow chart of the warning method of the present invention. <​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​Form the second historical dataset D2, and use the quartile method to clean the data of the second historical dataset D2. This dataset is for lubricating oil passing through the coarse filter element → oil-water exchanger.

[0086] (3) Select T, P from the data of P I b I and P D , and use T, P I and P D to form the third historical dataset D3, and use the quartile method to clean the data of the third historical dataset D3. This dataset is for lubricating oil passing through the fine filter element → coarse filter element → oil-water exchanger and oil-water exchange bypass.

[0087] (4) Select T, P from the data of P I >a and T > b I and P D , and use T, P I to form the fourth historical dataset D4, and use the quartile method to clean the data of the fourth historical dataset D4. This dataset is for lubricating oil passing through the coarse filter element → oil-water exchanger and oil-water exchange bypass.

[0088] Step 3, establish the following model for the first historical dataset D1:

[0089] S31, perform unary linear regression fitting on all the independent variable gearbox filter inlet pressures P in the first historical dataset D1 I and all the filter differential pressure dependent variables P D using the least squares method to obtain the first model, that is, the linear equation f1(x).

[0090] S32, combine each P in the first historical dataset D1 I with the linear equation f1(x), calculate the corresponding estimated value P` of the filter differential pressure dependent variable D , and calculate the residual E1 corresponding to each filter differential pressure dependent variable as E1 = P D - P` D [[ID=4�5]].

[0091] ] S33, calculate the mean of the residuals and the variance of the residuals

[0092] S34, calculate the threshold of the first filter differential pressure corresponding to the first historical dataset

[0093] Step 4, establish the following model for the second historical dataset D2:

[0094] S41, for all P in the second historical dataset D2 I and all PD Using the least squares method to perform univariate linear regression fitting, we obtain the second model, namely the linear equation f2(x).

[0095] S42, each independent variable P in the second historical dataset D2 I By combining the linear equation f2(x), the corresponding filter pressure differential dependent variable estimate P` is calculated. D And calculate the residual E2 = P corresponding to the differential pressure dependent variable of each filter. D -P` D .

[0096] S43, Calculate the mean residual. and residual variance

[0097] S44, Calculate the threshold of the second filter pressure difference corresponding to the second historical dataset.

[0098] Step 5, build the model for the third historical dataset D3 as follows:

[0099] S51, for all T and P values ​​in the third historical dataset D3 I and P D The min-max normalization method was used to process the data, resulting in normalized T″ and P. I "and P D ", and record (T 3max |T 3min ),

[0100] Normalization formula:

[0101] S52, for all normalized T″ and P″ I "and P D "Using the least squares method to perform multiple linear regression fitting, we obtain the third model, namely the linear equation f3(x)."

[0102] S53, using each normalized T″ and P I ", combined with f3(x), the corresponding filter pressure differential dependent variable estimate P` is obtained. D And calculate the residual E3 = P corresponding to the differential pressure dependent variable of each filter. D "-P" D .

[0103] S54, Calculate the mean residual. and residual variance

[0104] S55, Calculate the threshold of the third filter pressure difference corresponding to the third historical dataset.

[0105] Step 6, establish the following model for the fourth historical data set D4:

[0106] S61, for all T and P in the fourth historical data set D4 I and P D Use the min-max normalization method for processing to obtain the normalized T″, P I ″ and P D ″, and record (T 4max |T 4min ),

[0107] Normalization formula:

[0108] S62, for all the normalized T″, P I ″ and P D ″ Use the least squares method for multiple linear regression fitting to obtain the fourth model, that is, the linear equation f4(x).

[0109] S63, use each normalized T″ and P I ″, combined with f4(x) to calculate the corresponding estimated value of the filter differential pressure dependent variable P` D , and calculate the residual E4 corresponding to each filter differential pressure dependent variable as E4 = P D ″ - P` D .

[0110] S64, calculate the residual mean and the residual variance

[0111] S65, calculate the threshold of the fourth filter differential pressure corresponding to the fourth data set

[0112] Step 7, obtain the real-time operation data of the wind turbine to be measured from the SCADA system. The real-time operation data includes the real-time oil temperature t at the outlet of the gearbox, the real-time inlet pressure p of the gearbox filter i , the real-time outlet pressure p of the gearbox filter o , and calculate the real-time differential pressure p of the filter d .

[0113] Step 8, data screening. According to the factory values of the filter: the threshold a of the fine filter bypass pressure valve and the threshold b of the thermal valve, divide the data into four data sets:

[0114] (1) Select p i < a and t < b from the data, and select p i and p d , and pi and p d to form the first real-time data set d1.

[0115] (2) Select p from the data where p i > a and t < b, and select p i and p d , and use p i and p d to form the second real-time data set d2.

[0116] (3) Select t, p from the data where p i b, and select t, p i and p d , and use t, p i and p d to form the third real-time data set d3, and perform data cleaning and normalization according to (T 3max |T 3min ) in step 5 to obtain the normalized t″, p i ″ and p d ″.

[0117] (4) Select t, p from the data where p i > a and t > b, and select t, p i d and p, and use t, p i d and p 4max to form the fourth data set d4, and perform data cleaning and normalization according to (T 4min |T i ) in step 6 to obtain the normalized t″, p d ″ and p i ″.

[0118] Step 9, for the first real-time data set d1, combine each p d in the first real-time data set d1 with f1(x) in step 3 to calculate the corresponding real-time estimated value p` d of the filter differential pressure dependent variable, and calculate the corresponding real-time residual e1 = p d - p` i and the mean real-time residual of the first filter differential pressure corresponding to the first real-time data set

[0119] Step 10, for the second real-time data set d2, combine each p d in the second real-time data set d2 with f2(x) in step 4 to calculate the corresponding real-time estimated value p` d of the filter differential pressure dependent variable, and calculate the corresponding real-time residual e2 = pd -p` d and the mean real-time residual of the second filter pressure difference corresponding to the second real-time dataset.

[0120] Step 11, for the third real-time dataset d3, use the normalized values ​​of each t″ and p i Combining f3(x) from step 5, the corresponding real-time estimate of the filter pressure differential dependent variable p` is calculated. d And calculate the real-time residual e3 = p corresponding to the differential pressure dependent variable of each filter. d "-p` d and the mean real-time residual of the third filter pressure difference corresponding to the third real-time dataset.

[0121] Step 12, for the fourth real-time dataset d4, use the normalized values ​​of each t″ and p i Combining f4(x) from step 6, the corresponding real-time estimate of the filter pressure differential dependent variable p` is calculated. d And calculate the real-time residual e4 = p for each filter pressure differential dependent variable. d "-p` d and the mean real-time residual of the fourth filter pressure difference corresponding to the fourth real-time dataset.

[0122] Step 13, Early Warning Judgment:

[0123] (1) Since the gearbox oil passes through the coarse filter element in datasets d2 and d4, therefore if This indicates that the coarse filter element is clogged;

[0124] (2) Since in datasets d2 and d4, the gearbox oil passes through the coarse filter element, if... Then it is determined that the coarse filter element is blocked; in the data sets d1 and d3, the gearbox oil passes through the coarse filter element and the fine filter element, if simultaneously The fine filter cartridge may also become clogged;

[0125] (3) Since the gearbox oil passes through the coarse filter element in datasets d2 and d4, if... The coarse filter element is not clogged; data d1 and d3 are concentrated, indicating that the gearbox oil passes through both the coarse and fine filter elements. If simultaneously... The fine filter cartridge will become clogged;

[0126] (4) Since the gearbox oil passes through the coarse filter element in datasets d2 and d4, if... The coarse filter element is not clogged; data d1 and d3 are concentrated, indicating that the gearbox oil passes through both the coarse and fine filter elements. If simultaneously... The fine filter element is not clogged.

[0127] In this invention, reliable historical data is first used to perform linear regression fitting on locally linear datasets D1, D2, D3, and D4, obtaining linear equations f1(x), f2(x), f3(x), and f4(x) and residuals. Since the residuals of the linear regression equations follow a normal distribution, the boundary values ​​with a 99.7% probability interval of the residual distribution are [mean - 3 * variance, mean + 3 * variance]. Because filter blockage will increase the filter pressure difference, the upper boundary value [mean + 3 * variance] is used to calculate the residuals Th1, Th2, Th3, and Th4. For the real-time running datasets d1, d2, d3, and d4, the estimated values ​​of all points are calculated based on the filter inlet pressure and the regression equations f1(x), f2(x), f3(x), and f4(x), and the mean residual is calculated by subtracting the estimated value from the actual value. The residuals are analyzed as follows:

[0128] Result 1:

[0129] In datasets d2 and d4, the gearbox oil only passes through the coarse filter element. If... If the residual exceeds the upper boundary value of the residual distribution interval with a probability of 99.7%, that is, under the same filter inlet pressure, the actual filter pressure difference is too high and exceeds the reasonable range of filter pressure difference in historical sample data, then the coarse filter element is considered to be blocked.

[0130] Result 2:

[0131] In datasets d1 and d3, the gearbox oil passes through a coarse filter element and a fine filter element. If... This means that the residual has exceeded the upper boundary value of the residual distribution interval with a probability of 99.7%. In other words, under the same filter inlet pressure, the actual filter pressure difference is too high and exceeds the reasonable range of filter pressure difference in historical sample data. Since the gearbox oil passes through both the coarse filter element and the fine filter element at the same time, both the coarse filter and the fine filter element may be blocked. At this time, result 1 should be used in conjunction with the judgment. If result 1 is true, it is determined that the coarse filter element is blocked and the fine filter element may be blocked. If result 1 is not true, it is determined that the fine filter element is blocked.

[0132] Result 3: and

[0133] If the residuals in the test data d1, d2, d3, and d4 are all within the boundary range with a probability of 99.7% in the distribution interval, then the filter pressure difference is considered to be within a reasonable range, and it is determined that neither the coarse filter nor the fine filter element is blocked.

[0134] This preferred embodiment provides a wind turbine gearbox filter blockage early warning and diagnostic system, including:

[0135] The historical data acquisition unit is used to acquire historical data of the wind turbine under test, including gearbox outlet oil temperature, gearbox filter inlet pressure, gearbox filter outlet pressure, and filter differential pressure; and to acquire real-time operating data of the wind turbine under test, including real-time gearbox outlet oil temperature, real-time gearbox filter inlet pressure, and real-time gearbox filter outlet pressure.

[0136] The data partitioning unit is used to divide the obtained historical data and real-time running data into four corresponding datasets.

[0137] The threshold calculation unit is used to model the four datasets respectively and calculate the filter differential pressure threshold for each dataset based on the model.

[0138] The residual mean calculation unit uses the model corresponding to each dataset to calculate the real-time residual mean of the filter pressure difference corresponding to the four datasets of real-time running data.

[0139] The anomaly detection unit is used to determine whether the gearbox filter of the wind turbine under test is abnormal based on the real-time residual average value of the obtained filter differential pressure and the obtained filter differential pressure threshold.

[0140] Because the gearbox lubricating oil cooling system controls the lubricating oil through different circuits based on oil pressure and temperature, there is an uncertain nonlinear relationship between the pressure difference across the gearbox filter and the oil pressure and temperature before the filter during overall analysis. This makes it difficult to effectively analyze whether there are any abnormalities in the pressure difference across the filter. In this invention, the principle of local linearization of nonlinear relationships is used to split the data into four cases according to conditions: the first historical dataset D1 (lubricating oil passes through the fine filter element → coarse filter element → oil-water exchanger), the second historical dataset D2 (lubricating oil passes through the coarse filter element → oil-water exchanger), the third historical dataset D3 (lubricating oil passes through the fine filter element → coarse filter element → oil-water exchanger and oil-water exchange bypass), and the fourth historical dataset D4 (lubricating oil passes through the coarse filter element → oil-water exchanger and oil-water exchange bypass). In each dataset, the pressure difference across the gearbox filter conforms to a linear relationship with the oil pressure and temperature before the filter. This makes it easy to analyze whether there are any abnormalities in the pressure difference across the filter, thereby determining whether the filter element is blocked.

Claims

1. A wind turbine gearbox filter blockage early warning diagnostic method, characterized by, The method comprises the following steps: Step 1, obtaining historical data of the wind turbine to be tested, the historical data comprising gear box outlet oil temperature, gear box filter inlet pressure, gear box filter outlet pressure and filter differential pressure; Step 2, dividing the obtained historical data to obtain four data sets; Step 3, modeling the obtained four data sets respectively, and calculating the filter differential pressure threshold corresponding to each data set according to the obtained model; Step 4, obtaining real-time running data of the wind turbine to be tested, the real-time running data comprising real-time gear box outlet oil temperature, real-time gear box filter inlet pressure and real-time gear box filter outlet pressure; Step 5, dividing the obtained real-time running data according to step 2 to obtain four data sets; Step 6, using the model corresponding to each data set obtained in step 3 to calculate the real-time residual mean of the filter differential pressure corresponding to the four data sets of the real-time running data respectively; Step 7, judging whether the gear box filter of the wind turbine to be tested is abnormal according to the real-time residual mean obtained in step 6 and the filter differential pressure threshold obtained in step 3; in step 2, the obtained historical data is divided to obtain four data sets, and the specific method is as follows: According to the fine filter bypass pressure valve threshold and the thermal valve threshold of the gear box filter in the wind turbine to be tested, the historical data is divided to obtain four data sets in combination with the gear box outlet oil temperature and the gear box filter inlet pressure; in step 2, the obtained historical data is divided to obtain four data sets, and the specific method is as follows: Data in which the gear box filter inlet pressure is less than the fine filter bypass pressure valve threshold and the gear box outlet oil temperature is less than the thermal valve threshold is extracted from the historical data to form a first historical data set; Data in which the gear box filter inlet pressure is greater than the fine filter bypass pressure valve threshold and the gear box outlet oil temperature is less than the thermal valve threshold is extracted from the historical data to form a second historical data set; Data in which the gear box filter inlet pressure is less than the fine filter bypass pressure valve threshold and the gear box outlet oil temperature is greater than the thermal valve threshold is extracted from the historical data to form a third historical data set; Data in which the gear box filter inlet pressure is greater than the fine filter bypass pressure valve threshold and the gear box outlet oil temperature is greater than the thermal valve threshold is extracted from the historical data to form a fourth historical data set; in step 3, the obtained four data sets are modeled respectively, and the specific method is as follows: The gear box filter inlet pressure and all filter differential pressures in the first historical data set and the second historical data set are subjected to one-dimensional linear regression fitting respectively to obtain a first model and a second model; The gear box outlet oil temperature, the gear box filter inlet pressure and the filter differential pressure in the third historical data set and the fourth historical data set are subjected to normalization processing respectively to obtain normalized gear box outlet oil temperature, gear box filter inlet pressure and filter differential pressure; The normalized gearbox filter inlet pressure, gearbox outlet oil temperature and gearbox filter outlet pressure in the third historical data set and the fourth historical data set are subjected to multiple linear regression fitting, respectively, to obtain a third model and a fourth model; in step 3, the filter differential pressure threshold corresponding to each data set is calculated according to the obtained model, wherein the calculation method of the filter differential pressure threshold corresponding to the first historical data set is as follows: The filter differential pressure estimate value is calculated by combining each gearbox filter inlet pressure in the first historical data set with the first model; The residual corresponding to each filter differential pressure is calculated according to the obtained filter differential pressure estimate value; The first filter differential pressure threshold corresponding to the first historical data set is calculated according to the residual corresponding to each filter differential pressure; The calculation method of the filter differential pressure threshold corresponding to the second historical data set is as follows: The filter differential pressure estimate value is calculated by combining each gearbox filter inlet pressure in the second historical data set with the second model; The residual corresponding to each filter differential pressure is calculated according to the obtained filter differential pressure estimate value; The second filter differential pressure threshold corresponding to the second historical data set is calculated according to the residual corresponding to each filter differential pressure; The calculation method of the filter differential pressure threshold corresponding to the third historical data set is as follows: The normalized gearbox filter inlet pressure, gearbox outlet oil temperature and gearbox filter outlet pressure in the third historical data set and the fourth historical data set are subjected to multiple linear regression fitting, respectively, to obtain a third model and a fourth model; in step 3, the filter differential pressure threshold corresponding to each data set is calculated according to the obtained model, wherein the calculation method of the filter differential pressure threshold corresponding to the first historical data set is as follows: The normalized gearbox filter inlet pressure, gearbox outlet oil temperature and gearbox filter outlet pressure in the third historical data set and the fourth historical data set are subjected to multiple linear regression fitting, respectively, to obtain a third model and a fourth model; in step 3, the filter differential pressure threshold corresponding to each data set is calculated according to the obtained model, wherein the calculation method of the filter differential pressure threshold corresponding to the first historical data set is as follows: The filter differential pressure estimate value is calculated by combining each normalized gearbox filter inlet pressure, each normalized gearbox outlet oil temperature with the third model; The residual corresponding to each filter differential pressure is calculated according to the obtained filter differential pressure estimate value and the normalized gearbox filter outlet pressure; The third filter differential pressure threshold corresponding to the third historical data set is calculated according to the residual corresponding to each filter differential pressure; The calculation method of the filter differential pressure threshold corresponding to the fourth historical data set is as follows: The normalized gearbox filter inlet pressure, gearbox outlet oil temperature and gearbox filter outlet pressure in the third historical data set and the fourth historical data set are subjected to multiple linear regression fitting, respectively, to obtain a third model and a fourth model; in step 3, the filter differential pressure threshold corresponding to each data set is calculated according to the obtained model, wherein the calculation method of the filter differential pressure threshold corresponding to the first historical data set is as follows: The normalized gearbox filter inlet pressure, gearbox outlet oil temperature and gearbox filter outlet pressure in the third historical data set and the fourth historical data set are subjected to multiple linear regression fitting, respectively, to obtain a third model and a fourth model; in step 3, the filter differential pressure threshold corresponding to each data set is calculated according to the obtained model, wherein the calculation method of the filter differential pressure threshold corresponding to the first historical data set is as follows: The filter differential pressure estimate value is calculated by combining each normalized gearbox filter inlet pressure, each normalized gearbox outlet oil temperature with the fourth model; 2. A wind turbine gearbox filter blockage early warning diagnostic method according to claim 1, characterised in that, The residual corresponding to each filter differential pressure is calculated according to the obtained filter differential pressure estimate value and the normalized gearbox filter outlet pressure; The fourth filter differential pressure threshold corresponding to the fourth historical data set is calculated according to the residual corresponding to each filter differential pressure. In step 6, the real-time residual mean of the filter differential pressure corresponding to the four data sets of the real-time running data is calculated respectively by using the model corresponding to each data set obtained in step 3, specifically: The four data sets are the first real-time data set, the second real-time data set, the third real-time data set and the fourth real-time data set. The calculation method of the real-time residual mean of the filter pressure difference corresponding to the first real-time data set is: The real-time estimated value of the filter pressure difference is calculated by combining each gearbox filter real-time inlet pressure in the first real-time data set with the first model; The real-time residual corresponding to each filter pressure difference is calculated according to the real-time estimated value of the filter pressure difference; The real-time residual mean of the first filter pressure difference corresponding to the first real-time data set is calculated according to the real-time residual corresponding to each filter pressure difference obtained; The calculation method of the real-time residual mean of the filter pressure difference corresponding to the second real-time data set is: The real-time estimated value of the filter pressure difference is calculated by combining each gearbox filter real-time inlet pressure in the second real-time data set with the second model; The real-time residual corresponding to each filter pressure difference is calculated according to the real-time estimated value of the filter pressure difference; The real-time residual mean of the second filter pressure difference corresponding to the second real-time data set is calculated according to the real-time residual corresponding to each filter pressure difference obtained; The calculation method of the real-time residual mean of the filter pressure difference corresponding to the third real-time data set is: The normalized gearbox filter real-time inlet pressure, gearbox outlet real-time oil temperature and gearbox filter real-time outlet pressure in the third real-time data set are normalized to obtain the normalized gearbox filter real-time inlet pressure, gearbox outlet real-time oil temperature and gearbox filter real-time outlet pressure; The real-time estimated value of the filter pressure difference is calculated by combining each normalized gearbox filter real-time inlet pressure, each normalized gearbox real-time outlet oil temperature with the third model; The real-time residual corresponding to each filter pressure difference is calculated according to the real-time estimated value of the filter pressure difference and the normalized gearbox filter real-time outlet pressure; The real-time residual mean of the third filter pressure difference corresponding to the third real-time data set is calculated according to the real-time residual corresponding to each filter pressure difference obtained; The calculation method of the real-time residual mean of the filter pressure difference corresponding to the fourth real-time data set is: The normalized gearbox filter real-time inlet pressure, gearbox real-time outlet oil temperature and gearbox filter real-time outlet pressure in the fourth real-time data set are normalized to obtain the normalized gearbox filter real-time inlet pressure, gearbox real-time outlet oil temperature and gearbox filter real-time outlet pressure; The real-time estimated value of the filter pressure difference is calculated by combining each normalized gearbox filter real-time inlet pressure, each normalized gearbox real-time outlet oil temperature with the fourth model; The real-time residual corresponding to each filter pressure difference is calculated according to the real-time estimated value of the filter pressure difference and the normalized gearbox filter real-time outlet pressure; The real-time residual mean of the fourth filter pressure difference corresponding to the fourth real-time data set is calculated according to the real-time residual corresponding to each filter pressure difference obtained.

3. A wind turbine gearbox filter blockage early warning diagnostic method according to claim 2, characterised in that, In step 7, according to the real-time residual mean obtained in step 6 and the filter pressure difference threshold value obtained in step 3, it is judged whether the gearbox filter of the wind turbine to be tested is abnormal, and the specific method is: If then the coarse filter cartridge is identified as clogged; If and then it is determined that the coarse filter cartridge is clogged and the fine filter cartridge can also be clogged; If and then the fine filter cartridge is clogged and the coarse filter cartridge is not clogged; If and then the coarse filter cartridge is not clogged and the fine filter cartridge is not clogged; wherein, is a real-time residual mean of a first filter differential pressure corresponding to a first real-time data set; is a real-time residual mean of a second filter differential pressure corresponding to a second real-time data set; is a real-time residual mean of a third filter differential pressure corresponding to a third real-time data set; is a real-time residual mean of a fourth filter differential pressure corresponding to a fourth real-time data set; is a threshold value of a first filter differential pressure corresponding to a first historical data set; is a threshold value of a second filter differential pressure corresponding to a second historical data set; is a threshold value of a third filter differential pressure corresponding to a third historical data set; is a threshold value of a fourth filter differential pressure corresponding to a fourth historical data set.

4. A wind turbine generator set gearbox filter blockage early warning diagnostic system characterized by, The system of the wind turbine gearbox filter blockage early warning diagnosis method according to claim 1 comprises: The historical data acquisition unit is configured to acquire historical data of the wind turbine to be tested, wherein the historical data comprises gear box outlet oil temperature, gear box filter inlet pressure, gear box filter outlet pressure and filter differential pressure; and acquire real-time operation data of the wind turbine to be tested, wherein the real-time operation data comprises gear box real-time outlet oil temperature, gear box filter real-time inlet pressure and gear box filter real-time outlet pressure. The data division unit is configured to divide the obtained historical data and real-time operation data respectively, and obtain four corresponding data sets respectively. The threshold calculation unit is configured to model the four data sets respectively, and calculate the filter differential pressure threshold corresponding to each data set according to the obtained model. The residual mean calculation unit is configured to calculate the real-time residual mean of the filter differential pressure corresponding to the four data sets of the real-time operation data respectively by using the model corresponding to each data set. The abnormality judgment unit is configured to judge whether the gear box filter of the wind turbine to be tested is abnormal according to the obtained real-time residual mean of the filter differential pressure and the obtained filter differential pressure threshold.

Images