A wind turbine main drive system state evaluation method and system

Abstract

The application provides a wind turbine main drive system state evaluation method and system, relates to the wind turbine state evaluation technical field, and particularly relates to the following: a SCADA data set is collected and the data set is divided; the collected SCADA data is preprocessed; the operating condition of the wind turbine main drive system is divided; the wind turbine main drive system state parameters under different operating conditions are secondarily processed; the state parameter prediction model of the wind turbine main drive system under different operating conditions is established; and the state evaluation model is established, and the main drive system state evaluation grade result is output, so that the technical problem of inaccurate evaluation results caused by human subjective factors in the prior art wind turbine state evaluation method is solved.

Description

Technical Field

[0001] This invention relates to the field of wind turbine generator condition assessment technology, specifically to a method and system for assessing the condition of the main drive system of a wind turbine generator. Background Technology

[0002] Wind turbines are typically installed in harsh geographical environments and are constantly exposed to various severe weather conditions such as wind, rain, frost, and snow, which increases the failure rate of wind turbines. In particular, the main drive system of wind turbines accounts for 40%-60% of the downtime caused by failures in the main drive system. Once the turbine is shut down for repair, it not only consumes a lot of financial resources but also takes a long time, which also increases operating costs.

[0003] Currently, most condition assessments of wind turbine generators focus on the entire turbine or individual components such as gearboxes and main bearings, neglecting the crucial role and importance of the main drive system in the overall turbine operation. Assessments of the main drive system are rare. Furthermore, existing condition assessment methods typically rely on wind speed or clustering to simply categorize the turbine's operating conditions, sometimes without any categorization at all. This approach leads to inaccurate condition assessment results. Traditional assessment methods often employ binary classification, resulting in excessive subjective factors that make it difficult for the final assessment to accurately reflect the turbine's actual operating condition.

[0004] Therefore, this paper proposes a condition assessment method for the main drive system of wind turbines, which can provide a more detailed understanding of the operating status of wind turbines, remind engineers to carry out timely maintenance, reduce turbine downtime, lower operation and maintenance costs, and increase economic benefits. This method is of great significance to the maintenance of wind turbines and the development of the wind power industry. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides a method and system for assessing the condition of a wind turbine's main drive system, which solves the technical problem mentioned in the background that current wind turbine condition assessment methods suffer from subjective human factors and inaccurate assessment results.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for assessing the condition of a wind turbine main drive system, the method comprising:

[0009] Collect SCADA datasets and divide the datasets;

[0010] Preprocess the collected SCADA data;

[0011] The operating conditions of the main drive system of the wind turbine are divided into different working conditions.

[0012] Secondary processing is performed on the status parameters of the main drive system of the wind turbine under different operating conditions;

[0013] Establish a state parameter prediction model for the main drive system of a wind turbine under different operating conditions;

[0014] Establish a condition assessment model and output the condition assessment level results of the main drive system.

[0015] Preferably, the step of collecting and dividing the SCADA dataset includes: firstly, collecting the historical data of the SCADA system under different health states of the wind turbine main drive system, and dividing the dataset into two parts;

[0016] One dataset is used to train the neural network state prediction model of the main drive system, i.e., the training dataset; the other dataset is used for experimental verification, i.e., the experimental dataset.

[0017] Finally, real-time data is collected and input into the neural network prediction model of the main drive system to perform a state assessment of the wind turbine's main drive system.

[0018] Preferably, the preprocessing of the collected SCADA data includes: data cleaning and state parameter correlation analysis.

[0019] Preferably, the data cleaning and state parameter correlation analysis include: the data cleaning includes: cleaning out invalid, abnormal, or missing data based on the wind power curve of the wind turbine; the state parameter correlation analysis includes: selecting parameters that are strongly correlated with the operating state of the wind turbine's main drive system from the collected dataset, and selecting strongly correlated (|r|>0.6) and highly correlated (|r|>0.8) state parameters based on the Pearson correlation coefficient method.

[0020] The formula for calculating the Pearson correlation coefficient is as follows:

[0021]

[0022] In the formula: E(XY) is the common expectation of X and Y; E(X) is the expectation of X; E(Y) is the expectation of Y;

[0023] When the total number of samples X and Y is N, the formula for calculating the correlation coefficient between the total samples X and Y is as follows:

[0024]

[0025] In the formula: N is the total number of samples, It is for X i The standard score of the sample, σ X The standard deviation of the sample is 1. This is the sample average.

[0026] Preferably, the operation state of the wind turbine main drive system is divided into operating conditions, including: based on the collected historical data, the wind turbine's operating conditions are initially divided into operating condition 1 (shutdown stage), operating condition 2 (low power stage), operating condition 3 (rapid power increase stage), operating condition 4 (constant power stage), and operating condition 5 (shutdown stage due to wind speed exceedance) based on the actual wind power curve.

[0027] Secondly, for the rapid power increase stage of operating condition 3 and the constant power stage of operating condition 4, operating condition parameters were selected. Data clustering algorithms were used to further subdivide operating condition 3 and operating condition 4 into operating conditions such as 3-1, 3-2, ..., 3-m and 4-1, 4-2, ..., 4-n. At the same time, an operating condition identification model was established on this basis to identify the operating condition from real-time data during online evaluation.

[0028] Preferably, the secondary processing of the state parameters of the wind turbine main drive system under different operating conditions includes: after the operating conditions are divided, the Pearson correlation coefficient method is used to perform a second correlation analysis on the state parameters in each operating condition. Based on the calculation results, the state parameters with strong correlation (|r|>0.6) and extreme correlation (|r|>0.8) are selected, and redundant state parameters are removed.

[0029] Preferably, the secondary processing of the state parameters of the wind turbine main drive system under different operating conditions further includes: standardizing the selected state parameters using a linear function normalization method.

[0030] The calculation formula is as follows:

[0031]

[0032] In the formula: X norm Refers to standardized data; X max The maximum data in the dataset; X represents the smallest data point in the dataset; X represents the data to be standardized.

[0033] Preferably, the establishment of the state parameter prediction model of the wind turbine main drive system under different operating conditions includes: establishing a state parameter prediction model, wherein the state parameter prediction model is specifically: establishing a BiGRU state parameter prediction model based on attention mechanism optimized by chaotic particle swarm optimization (CPSO), and sequentially inputting the datasets under different operating conditions into the prediction model using deep learning methods to train the prediction model.

[0034] The experimental dataset was preprocessed, the operating conditions were divided into working conditions, and the state parameters were processed in a secondary manner. Finally, the data was input into the state parameter prediction model of the wind turbine main drive system under the corresponding working conditions. The accuracy of the prediction model was verified by comparing the output predicted values ​​with the actual values.

[0035] Preferably, the step of establishing a state assessment model and outputting the state assessment level result of the main drive system includes: establishing an evaluation state assessment structure system for the main drive system, with the target layer C being the main drive system, and the project layer C... i That is, important components in the main drive system, secondary indicator level C ij That is, for each component's key parameters, a three-scale method is used to measure the importance of the indicators in the decision-making process at the same level, and the initial weight W is calculated. I The influence weight W was calculated using the Decision and Experiment Evaluation Test (DEMATEL) method. D The subjective weight is obtained by combining the initial weight and the influence weight; the objective weight W is calculated using the entropy weight method. K A state evaluation model was constructed using the fuzzy comprehensive evaluation method combined with comprehensive weights. The established evaluation model calculates sequentially from the index layer and finally obtains the state level of the wind turbine main drive system.

[0036] The present invention also provides a condition assessment system for a wind turbine main drive system, the system comprising:

[0037] Data collection and partitioning module: used to collect SCADA datasets and partition them;

[0038] Data preprocessing module: used to preprocess the collected SCADA data;

[0039] Operating condition identification module: used to classify the operating conditions of the main drive system of wind turbine generator;

[0040] Parameter secondary processing module: used to perform secondary processing on the status parameters of the main drive system of the wind turbine under different operating conditions;

[0041] Prediction Model Building Module: Used to build state parameter prediction models for the main drive system of wind turbines under different operating conditions;

[0042] Condition assessment module: Used to establish a condition assessment model and output the condition assessment level results of the main drive system.

[0043] Beneficial effects

[0044] This invention provides a method and system for assessing the condition of a wind turbine's main drive system. It offers the following advantages:

[0045] 1. The wind turbine active system state assessment method provided by this invention can better understand the current operating state of the main drive system than traditional state assessment.

[0046] 2. The present invention provides a method for evaluating the status of an active system of a wind turbine, which divides the operating conditions of the main drive system into two categories and evaluates the operating status of the main drive system under the current conditions, which helps researchers to better understand the current situation of the main drive system.

[0047] 3. This invention uses an improved scaling method combined with DEMATEL to calculate subjective weights, and then combines these with objective weights calculated using the entropy weight method to obtain our final weights. This minimizes the influence of subjective human judgment and makes the evaluation results more accurate. It also solves the problem that there is currently no mature technology for evaluating the status of the main drive system. Attached Figure Description

[0048] Figure 1 A flowchart of a method for assessing the condition of a wind turbine main drive system provided by the present invention;

[0049] Figure 2 This invention provides a structural diagram of a wind turbine main drive system condition assessment system. Detailed Implementation

[0050] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0051] like Figure 1 As shown, a method for assessing the condition of a wind turbine main drive system includes:

[0052] Collect SCADA datasets and divide the datasets;

[0053] Preprocess the collected SCADA data;

[0054] The operating conditions of the main drive system of the wind turbine are divided into different working conditions.

[0055] Secondary processing is performed on the status parameters of the main drive system of the wind turbine under different operating conditions;

[0056] Establish a state parameter prediction model for the main drive system of a wind turbine under different operating conditions;

[0057] Establish a condition assessment model and output the condition assessment level results of the main drive system.

[0058] Preferably, the step of collecting and dividing the SCADA dataset includes: firstly, collecting the historical data of the SCADA system under different health states of the wind turbine main drive system, and dividing the dataset into two parts;

[0059] One dataset is used to train the neural network state prediction model of the main drive system, i.e., the training dataset; the other dataset is used for experimental verification, i.e., the experimental dataset.

[0060] Finally, real-time data is collected and input into the neural network prediction model of the main drive system to perform a state assessment of the wind turbine's main drive system.

[0061] Preferably, the preprocessing of the collected SCADA data includes: data cleaning and state parameter correlation analysis.

[0062] Preferably, the data cleaning and state parameter correlation analysis include: the data cleaning includes: cleaning out invalid, abnormal, or missing data based on the wind power curve of the wind turbine; the state parameter correlation analysis includes: selecting parameters that are strongly correlated with the operating state of the wind turbine's main drive system from the collected dataset, and selecting strongly correlated (|r|>0.6) and highly correlated (|r|>0.8) state parameters based on the Pearson correlation coefficient method.

[0063] The formula for calculating the Pearson correlation coefficient is as follows:

[0064]

[0065] In the formula: E(XY) is the common expectation of X and Y; E(X) is the expectation of X; E(Y) is the expectation of Y;

[0066] When the total number of samples X and Y is N, the formula for calculating the correlation coefficient between the total samples X and Y is as follows:

[0067]

[0068] In the formula: N is the total number of samples, It is for X i The standard score of the sample, σ X The standard deviation of the sample is 1. This is the sample average.

[0069] Preferably, the operation state of the wind turbine main drive system is divided into operating conditions, including: based on the collected historical data, the wind turbine's operating conditions are initially divided into operating condition 1 (shutdown stage), operating condition 2 (low power stage), operating condition 3 (rapid power increase stage), operating condition 4 (constant power stage), and operating condition 5 (shutdown stage due to wind speed exceedance) based on the actual wind power curve.

[0070] Secondly, for the rapid power increase stage of operating condition 3 and the constant power stage of operating condition 4, operating condition parameters were selected. Data clustering algorithms were used to further subdivide operating condition 3 and operating condition 4 into operating conditions such as 3-1, 3-2, ..., 3-m and 4-1, 4-2, ..., 4-n. At the same time, an operating condition identification model was established on this basis to identify the operating condition from real-time data during online evaluation.

[0071] Preferably, the secondary processing of the state parameters of the wind turbine main drive system under different operating conditions includes: after the operating conditions are divided, the Pearson correlation coefficient method is used to perform a second correlation analysis on the state parameters in each operating condition. Based on the calculation results, the state parameters with strong correlation (|r|>0.6) and extreme correlation (|r|>0.8) are selected, and redundant state parameters are removed.

[0072] Preferably, the secondary processing of the state parameters of the wind turbine main drive system under different operating conditions further includes: standardizing the selected state parameters using a linear function normalization method.

[0073] The calculation formula is as follows:

[0074]

[0075] In the formula: X norm Refers to standardized data; X max The maximum data in the dataset; X represents the smallest data point in the dataset; X represents the data to be standardized.

[0076] Preferably, the establishment of the state parameter prediction model of the wind turbine main drive system under different operating conditions includes: establishing a state parameter prediction model, wherein the state parameter prediction model is specifically: establishing a BiGRU state parameter prediction model based on attention mechanism optimized by chaotic particle swarm optimization (CPSO), and sequentially inputting the datasets under different operating conditions into the prediction model using deep learning methods to train the prediction model.

[0077] The experimental dataset was preprocessed, the operating conditions were divided into working conditions, and the state parameters were processed in a secondary manner. Finally, the data was input into the state parameter prediction model of the wind turbine main drive system under the corresponding working conditions. The accuracy of the prediction model was verified by comparing the output predicted values ​​with the actual values.

[0078] Preferably, the accuracy of the verification prediction model includes, but is not limited to, mean squared error, root mean squared error, and mean absolute error, which are hereinafter referred to as MSE, RMSE, and MAE, respectively.

[0079] The formula for calculating MSE is as follows:

[0080]

[0081] M t =observed t -predicted t

[0082] In the formula: observed t The predicted data represents the actual value. t M represents the predicted value output by the prediction model. t The value is the residual between the predicted value and the actual value; t is the number of data points in the dataset.

[0083] The RMSE calculation formula is as follows:

[0084]

[0085] M t =observed t -predicted t

[0086] In the formula: observed t The predicted data represents the actual value. t M represents the predicted value output by the prediction model. t The value is the residual between the predicted value and the actual value; t is the number of data points in the dataset.

[0087] The MAE calculation formula is as follows:

[0088]

[0089] M t =observed t -predicted t

[0090] In the formula: observed t The predicted data represents the actual value. t M represents the predicted value output by the prediction model. tThe value is the residual between the predicted value and the actual value; t is the number of data points in the dataset.

[0091] The residual M calculated under the corresponding working conditions t Select the largest residual M t(max) As the status alarm threshold under the current operating conditions, a status alarm is issued when the residual between the actual value and the predicted value of the input data exceeds this threshold.

[0092] Preferably, the step of establishing a state assessment model and outputting the state assessment level result of the main drive system includes: establishing an evaluation state assessment structure system for the main drive system, with the target layer C being the main drive system, and the project layer C... i That is, important components in the main drive system, secondary indicator level C ij That is, for each component's key parameters, a three-scale method is used to measure the importance of the indicators in the decision-making process at the same level, and the initial weight W is calculated. I The influence weight W was calculated using the Decision and Experiment Evaluation Test (DEMATEL) method. D The subjective weight is obtained by combining the initial weight and the influence weight; the objective weight W is calculated using the entropy weight method. K A state evaluation model was constructed using the fuzzy comprehensive evaluation method combined with comprehensive weights. The established evaluation model calculates sequentially from the index layer and finally obtains the state level of the wind turbine main drive system.

[0093] Preferably, the scale values ​​and their corresponding meanings are as follows: [-1 0 1] represent the relative importance of the indicators compared to each other [not important, equally important, important], to construct the judgment matrix B = (b ij ) n×n Calculate the initial weight W I ;

[0094] Secondly, we use the Decision Experiment and Evaluation Experiment (DEMATEL) method to calculate the influence weight W. D The specific process is as follows;

[0095] (1) The direct influence matrix D = (d) between indicators at different levels of the component ij ) n×n Here we define [0 1 2 3] to represent [no impact, similar impact, some impact, and significant impact] respectively, to evaluate the importance of each indicator and thus establish a direct impact matrix;

[0096] (2) Standardize the influence matrix D to obtain matrix X;

[0097]

[0098] (3) Calculate the comprehensive influence matrix E and influence weight W among the indicators at each level. D,

[0099] E = X + X 2 +...+X n =(e ij ) n×n

[0100]

[0101] (4) Calculate the corrected subjective weights

[0102] Next, the objective weight W is calculated using the entropy weight method. K The specific steps are as follows:

[0103] (1) Standardize the parameters based on their properties:

[0104] Positive indicators:

[0105]

[0106] Reverse indicators:

[0107]

[0108] (2) Calculate the information entropy of each indicator:

[0109]

[0110] Where: A = -ln(n) -1 B = ln p ij ,

[0111] (3) Calculate the weight of each indicator:

[0112]

[0113] Furthermore, the establishment of the condition assessment model also includes the calculation of the degree of degradation. Since parameters such as temperature and power are easily affected by wind speed, using a fixed threshold to calculate the degree of degradation would reduce the accuracy of the assessment. Therefore, a dynamic method for calculating the degree of degradation is adopted, and the specific calculation formula is as follows:

[0114]

[0115] For other parameter indicators, the degree of degradation is calculated using traditional degradation calculation methods based on fixed thresholds.

[0116] Finally, we constructed our evaluation model using the fuzzy comprehensive evaluation method combined with the comprehensive weights we calculated. The operating status level of the wind turbine's main drive system was divided into 5 levels: {Healthy, Sub-healthy, Qualified, Attention, Serious}.

[0117] A = W·G

[0118] In the formula, the comprehensive weight G is the membership matrix. g ij The membership degree of the indicator at the current state level is represented by the selected membership function, where A is the evaluation result and · is the operator.

[0119] Preferably, the calculation also includes starting from the index layer based on the established evaluation model, and finally obtaining the state level membership matrix of our target layer, namely the wind turbine main drive system.

[0120] A = {x1 x2 x3 x4 x5}, where x1 to x5 reflect the degree of membership of the current main drive system in the {healthy, sub-healthy, qualified, attention, serious} state level. When real-time data is collected, after the first data processing, working condition identification, the second data processing, and state parameter prediction, the real-time state assessment result is finally obtained through the evaluation model.

[0121] The present invention also provides a condition assessment system for a wind turbine main drive system, the system comprising:

[0122] Data collection and partitioning module: used to collect SCADA datasets and partition them;

[0123] Data preprocessing module: used to preprocess the collected SCADA data;

[0124] Operating condition identification module: used to classify the operating conditions of the main drive system of wind turbine generator;

[0125] Parameter secondary processing module: used to perform secondary processing on the status parameters of the main drive system of the wind turbine under different operating conditions;

[0126] Prediction Model Building Module: Used to build state parameter prediction models for the main drive system of wind turbines under different operating conditions;

[0127] Condition assessment module: Used to establish a condition assessment model and output the condition assessment level results of the main drive system.

[0128] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method of condition assessment of a main drive system of a wind power generator, characterized in that, The method includes: Collect SCADA datasets and divide the datasets; The collected SCADA data is preprocessed, including data cleaning and state parameter correlation analysis; the data cleaning includes cleaning invalid, abnormal or missing data according to the wind power curve of the wind turbine; the state parameter correlation analysis includes selecting parameters with strong correlation with the operation state of the main drive system of the wind turbine from the collected data set, and calculating and selecting strong correlation and extremely correlated state parameters according to the Pearson correlation coefficient method; The formula for calculating the Pearson correlation coefficient is as follows: where: is the expectation of X, Y together; is the expectation of X; is the expectation of Y; When the total number of samples X and Y is N, the formula for calculating the correlation coefficient between the total samples X and Y is as follows: where: N is the total number of samples, is the standard score for the sample, is the standard deviation of the X sample, is the mean of the X sample; is the standard score for the sample, is the standard deviation of the Y sample, is the mean of the Y sample; The operating conditions of the main drive system of the wind turbine are divided into the following categories: Based on the collected historical data and the actual wind power curve, the operating conditions of the wind turbine are initially divided into the following categories: Condition 1: shutdown stage; Condition 2: low power stage; Condition 3: rapid power increase stage; Condition 4: constant power stage; and Condition 5: shutdown stage when wind speed exceeds the limit. Secondly, for the rapid power increase stage of operating condition 3 and the constant power stage of operating condition 4, operating condition parameters were selected. Data clustering algorithms were used to further subdivide operating condition 3 and operating condition 4 into operating condition 3-1, 3-2, ..., 3-m and operating condition 4-1, 4-2, ..., 4-n. At the same time, an operating condition identification model was established on this basis to identify the operating condition from real-time data during online evaluation. The state parameters of the fan main transmission system under different working conditions are twice processed, including: after the working condition division is completed, Pearson correlation coefficient method is used to perform correlation analysis on the state parameters in each working condition again, and according to the calculation results, the state parameters with strong correlation and extreme correlation are selected, and the redundant state parameters are removed; Establish a state parameter prediction model for the main drive system of a wind turbine under different operating conditions; Establish a condition assessment model and output the condition assessment level results of the main drive system.

2. The wind generator main drive system condition assessment method according to claim 1, characterized by, The process of collecting and dividing the SCADA dataset includes: firstly, collecting the historical multivariate time series data of the SCADA system under different health states of the wind turbine main drive system, and dividing the dataset into two parts. One dataset is used to train the state parameter prediction model of the main drive system, i.e., the training dataset; the other dataset is used for experimental verification, i.e., the experimental dataset. Finally, real-time data is collected and input into the state parameter prediction model of the main drive system to perform state assessment of the wind turbine main drive system.

3. The wind power generator main drive system condition assessment method according to claim 1, characterized by, The secondary processing of the state parameters of the main drive system of the wind turbine under different operating conditions also includes: standardizing the selected state parameters using a linear function normalization method. The calculation formula is as follows: wherein: denotes the normalized data; is the maximum data of the data set; is the minimum data of the data set; is the data to be normalized.

4. The wind generator main drive system condition assessment method according to claim 1, characterized by, The establishment of a state parameter prediction model for the main drive system of a wind turbine under different operating conditions includes: establishing a state parameter prediction model, specifically: establishing a BiGRU state parameter prediction model based on an attention mechanism optimized by the chaotic particle swarm optimization (CPSO) algorithm, and sequentially inputting the training datasets under different operating conditions into the prediction model using deep learning methods to train the prediction model. The experimental dataset was preprocessed, the operating conditions were divided into working conditions, and the state parameters were processed in a secondary manner. Finally, the data was input into the state parameter prediction model of the wind turbine main drive system under the corresponding working conditions. The accuracy of the prediction model was verified by comparing the output predicted values ​​with the actual values.

5. The wind power generator main drive system condition assessment method according to claim 1, characterized by, The establishment of the state assessment model and the output of the state assessment level results of the main drive system include: establishing an evaluation state assessment structure system for the main drive system, with a target layer. C That is, the main drive system, at the project level. That is, important components in the main drive system, secondary indicator layer That is, for each component's key parameters, a three-scale method is used to measure the importance of the indicators in the decision-making process at the same level, and the initial weights are calculated. ; Calculate influence weights using the Decision and Experiment Evaluation Test (DEMATEL) method. Subjective weights are obtained by combining initial weights and influence weights; objective weights are calculated using the entropy weight method. A state assessment model was constructed using the fuzzy comprehensive evaluation method combined with comprehensive weights. The established assessment model calculates sequentially from the index layer and finally obtains the state level of the wind turbine main drive system.

6. A wind generator main drive system condition assessment system, characterized by, The system includes: a data collection and segmentation module: used to collect SCADA datasets and segment the datasets; The data preprocessing module is used to preprocess the collected SCADA data, including data cleaning and state parameter correlation analysis. Data cleaning includes removing invalid, abnormal, or missing data based on the wind power curve of the wind turbine. State parameter correlation analysis includes selecting parameters with strong correlation to the operating state of the wind turbine's main drive system from the collected dataset, and calculating and selecting strongly correlated parameters using the Pearson correlation coefficient method. and highly relevant State parameters ； The formula for calculating the Pearson correlation coefficient is as follows: In the formula: For the common expectation of X and Y; Let X be the expectation; Let Y be the expectation; When the total number of samples X and Y is N, the formula for calculating the correlation coefficient between the total samples X and Y is as follows: In the formula: N The total number of samples, Yes Standard scores of the sample Let X be the sample standard deviation. X is the sample mean. Yes Standard scores of the sample Let Y be the standard deviation of the sample. Y represents the sample mean; Operating condition identification module: used to classify the operating conditions of the main drive system of wind turbine generator, including: based on the collected historical data, firstly, based on the actual wind power curve, the operating conditions of the wind turbine are initially divided into operating condition 1 shutdown stage, operating condition 2 low power stage, operating condition 3 rapid power increase stage, operating condition 4 constant power stage and operating condition 5 shutdown stage when wind speed exceeds the limit. Secondly, for the rapid power increase stage of operating condition 3 and the constant power stage of operating condition 4, operating condition parameters were selected. Data clustering algorithms were used to further subdivide operating condition 3 and operating condition 4 into operating condition 3-1, 3-2...3-m and operating condition 4-1, 4-2...4-n. At the same time, an operating condition identification model was established on this basis to identify the operating condition from real-time data during online evaluation. The parameter secondary processing module is used to perform secondary processing on the state parameters of the wind turbine main drive system under different operating conditions. This includes: after the operating conditions are divided, using the Pearson correlation coefficient method, performing a second correlation analysis on the state parameters in each operating condition, and selecting strongly correlated parameters based on the calculation results. and highly relevant The state parameters are determined, and redundant state parameters are removed. Prediction Model Building Module: Used to build state parameter prediction models for the main drive system of wind turbines under different operating conditions; Condition assessment module: Used to establish a condition assessment model and output the condition assessment level results of the main drive system.

Images