A wind turbine condition monitoring and maintenance decision support system

Abstract

The present invention belongs to the technical field of wind turbine status monitoring, and discloses a wind turbine status monitoring and maintenance decision support system, comprising: collecting temperature fluctuation indicators, comprehensive vibration indicators, comprehensive wear health indicators and comprehensive environmental adaptability indicators; correlating and analyzing the temperature fluctuation indicators, comprehensive vibration indicators, comprehensive wear health indicators and comprehensive environmental adaptability indicators to obtain a comprehensive health index of the wind turbine; evaluating the current status of the wind turbine according to the comprehensive health index, judging whether the equipment has an abnormality, and generating an equipment status signal; triggering an alarm mechanism according to the equipment status signal, and generating an abnormality assessment report at the same time; while improving the accuracy and sensitivity of wind turbine status monitoring, optimizing the maintenance strategy, significantly enhancing the environmental adaptability of the system, improving operation and maintenance efficiency, equipment reliability and economic benefits, and providing technical support for realizing intelligent operation and maintenance.

Description

Technical Field

[0001] The present invention relates to the technical field of wind turbine condition monitoring, and more particularly to a wind turbine condition monitoring and maintenance decision support system. Background Art

[0002] With the continuous increase in installed wind power capacity, ensuring the safe and stable operation of wind turbines has become a critical issue. Wind turbines operate in complex and variable environments, including extreme temperature fluctuations, strong winds, and humidity variations. These factors can lead to performance degradation or even failure. Therefore, effective condition monitoring and maintenance decision support for wind turbines are crucial.

[0003] Currently, wind turbine condition monitoring primarily relies on regular inspections, sensor monitoring, and predictive maintenance software. However, regular inspections are inefficient and difficult to capture real-time changes; sensor monitoring focuses on a single indicator, lacks comprehensive assessment, and lacks real-time performance and sensitivity; predictive maintenance suffers from insufficient data integration and poor environmental adaptability, leading to inaccurate comprehensive assessments, potentially increasing unnecessary maintenance times, and limited reliability in complex environments. Therefore, existing technologies urgently need to improve comprehensive assessments, real-time response, and environmental adaptability. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned objectives, the present invention provides the following technical solution: a wind turbine condition monitoring and maintenance decision support system, comprising:

[0005] Data construction module: used to collect temperature fluctuation indicators, comprehensive vibration indicators, comprehensive wear and health indicators and comprehensive environmental adaptability indicators;

[0006] Data analysis module: used to calculate the comprehensive health index of key components through temperature fluctuation index, comprehensive vibration index, comprehensive wear health index and comprehensive environmental adaptability index, and perform correlation analysis on the comprehensive health index of each key component to obtain the comprehensive health index of the wind turbine;

[0007] Status assessment module: used to assess the current status of the wind turbine based on the comprehensive health index, determine whether the equipment is abnormal, and generate equipment status signals;

[0008] Abnormal alarm module: used to trigger the alarm mechanism according to the equipment status signal and generate abnormal assessment report.

[0009] Furthermore, the temperature fluctuation index is collected in the following manner:

[0010] In a monitoring cycle, the monitoring cycle is divided into n moments, and the temperature data of each key component in the monitoring cycle at each moment is collected using thermocouples and arranged in time sequence to obtain a time series array;

[0011] Among them, key components include the wind turbine's main shaft, bearings, generator, hydraulic brake assembly and gearbox;

[0012] Set a time window with a step size of wt. For any time t in the time series array, calculate the average value in the time window before time t, record it as the moving average temperature, then obtain the absolute difference between the moving average temperature and the temperature data at time t, and record the absolute difference as the temperature fluctuation index at time t.

[0013] Furthermore, the comprehensive vibration index is collected in the following ways:

[0014] Step S31: Using the start time of the monitoring period as the initial time, using the accelerometer to collect the vibration signal data of each key component in real time, and preprocessing the vibration signal data between the initial time and time t, including using median filtering to remove high-frequency noise and normalization;

[0015] Step S32: Using a high-pass filter and a low-pass filter to perform convolution processing on the preprocessed vibration signal data, respectively obtaining a first-level low-pass filtering result Yd1[n] and a first-level high-pass filtering result Yg1[n], which are filtered vibration signal data and still maintain the same length sequence as the preprocessed vibration signal data;

[0016] Where n is a discrete time index, representing the nth sampling point in the vibration signal data;

[0017] Step S33: sampling Yd1[n] and Yg1[n] respectively, including taking a sample from every other sampling point in Yd1[n] to obtain an approximate coefficient sequence Ad1, and taking a sample from every other sampling point in Yg1[n] to obtain a detail coefficient sequence Dg1;

[0018] Among them, a sample represents the data at a sampling point;

[0019] Step S34: using a high-pass filter and a low-pass filter to perform convolution processing on the approximate coefficient sequence Ad1 to obtain a secondary high-pass filtering result Yg2[n] and a secondary low-pass filtering result Yd2[n];

[0020] Process Yd2[n] and Yg2[n] using the same sampling method as step S33 to obtain the approximate coefficient sequence Ad2 and the detail coefficient sequence Dg2;

[0021] Step S35: Repeat the iterative step S34, using a high-pass filter and a low-pass filter to convolve and sample the approximate coefficient sequence Adh-1 of the previous level until the set number of times H is reached, to obtain the approximate coefficient sequence AdH and detail coefficient sequence DgH of each level, where h represents the level;

[0022] Step S36: For each level h, calculate the energy distribution of the detail coefficients at each level ,in, is the detail coefficient at the hth level, Indicates the number of detail coefficients at level h, represents the total energy at level h;

[0023] Take the mean of the total energy at all levels as the time-frequency domain feature of the key component at time t;

[0024] Step S37: Arranging the pre-processed vibration signal data in time series to form a vibration time series array;

[0025] Take the mean of the absolute values ​​of all data in the vibration time series array as the absolute average value, calculate the ratio of the maximum absolute value in the vibration time series array to the absolute average value, and obtain the vibration pulse index of the key component at time t;

[0026] Step S38: For each key component, create a comprehensive vibration index of the key component using the time-frequency domain characteristics and the vibration pulse index , where WFD represents the time-frequency domain feature and TFF represents the vibration pulse index.

[0027] Furthermore, the comprehensive wear health index is collected in the following ways:

[0028] A magnetic probe is used to absorb metal chips from the oil of each key component. The magnetic field change on the magnetic probe at time t is measured by a magnetic induction sensor. The magnetic induction sensor outputs a corresponding current signal, which is compared with the calibration curve to obtain the corresponding metal element concentration.

[0029] The standard curve is the relationship curve between the current signal and the magnetic field change established in the laboratory;

[0030] The calibration curve is drawn as follows:

[0031] Prepare a series of standard oil samples with known metal element concentrations, place a magnetic probe into each standard oil sample, and adsorb the magnetic particles;

[0032] A magnetic induction sensor is used to record the magnetic field changes on the magnetic probe after each experiment and output the corresponding current signal;

[0033] Record the metal element concentration of each standard oil sample and its corresponding current signal, and draw a standard curve with the metal element concentration as the horizontal axis and the current signal as the vertical axis;

[0034] The difference in metal element concentration between any moment t and the previous moment t-1 within the monitoring period is taken as the concentration difference, and arranged in time sequence to obtain a concentration difference array;

[0035] Using a time window with a step size of wt, for any moment t on the concentration difference array, calculate the average value in the time window before moment t as the wear metal content trend of the key component at moment t;

[0036] Constructing a comprehensive wear health index by trending the wear metal content of key components , where WMCT represents the wear metal content trend, represents the oil health index of key components at time t, is a preset fixed scale factor, ;

[0037] The collection methods of oil health indicators include:

[0038] Use a viscometer to measure the viscosity of the oil in each key component at time t, and take the ratio of the oil viscosity at time t to the oil viscosity at the initial time as the change in the oil viscosity in each key component;

[0039] The oil contamination degree at time t is obtained by measuring the number of pollutants in the oil of each key component using a particle counter. The ratio of the oil contamination degree at time t to the oil contamination degree at the initial time is taken as the oil contamination degree change.

[0040] Use acid-base titration to measure the acid value of the oil in each key component at time t, and take the ratio of the oil acid value at time t to the acid value of the new oil as the oil oxidation amount;

[0041] The oil health index of each key component at time t is obtained by multiplying the change in oil viscosity, oil contamination, and oil oxidation.

[0042] Furthermore, the comprehensive environmental adaptability index is collected in the following ways:

[0043] Step S51: collecting environmental data of the wind turbine generator system at each moment during the monitoring period through a weather station, including temperature, humidity, wind speed, particulate matter, salt spray, rainfall, and air pressure, to form an original environmental data set;

[0044] Step S52: Standardize each environmental factor in the original environmental data set. , calculate its probability distribution , where i represents the i-th environmental factor, j is the index of the moment in the monitoring period, and n represents the number of moments in the monitoring period;

[0045] Step S53: Calculate the information entropy of each environmental factor based on the probability distribution of each environmental factor ,in, is the normalization coefficient;

[0046] Step S54: Calculate the weight of each environmental factor based on the information entropy of each environmental factor , where m is the number of environmental factors;

[0047] Step S55: Based on each environmental factor and the corresponding weight, a weighted summation formula is used to calculate the comprehensive environmental adaptability index of the environment in which the wind turbine generator system is located at time t.

[0048] Furthermore, the method of calculating the health index of key components by using the temperature fluctuation index, the comprehensive vibration index, the comprehensive wear health index and the comprehensive environmental adaptability index includes:

[0049] Step S61: standardizing the temperature fluctuation index, comprehensive vibration index, comprehensive wear health index, and comprehensive environmental adaptability index at each time t to form an index data set;

[0050] Step S62: For each indicator type, set an ideal optimal indicator value as a reference indicator;

[0051] For each indicator u at time t, calculate the absolute difference between it and the reference indicator ,in, is the value of the u-th indicator at time t, is the value of the reference indicator at time t;

[0052] According to the absolute difference of each indicator at time t, find the minimum absolute difference and maximum value ;

[0053] Step S63: Calculate the correlation coefficient of each indicator at time t based on the minimum and maximum absolute differences ,in, represents the correlation coefficient of the u-th indicator at time t;

[0054] The mean correlation coefficient of all indicators at time t is taken as the health index of each key component at time t.

[0055] Furthermore, the method of performing correlation analysis on the health index of each key component to obtain the comprehensive health index of the wind turbine generator system includes:

[0056] Step S71: taking every two key components as a key component pair, and obtaining the mechanical connection relationship between each key component pair;

[0057] For key component pairs that do not have mechanical connection relationships, obtain their functional dependencies;

[0058] Step S72: constructing a graph structure, taking each key component as a node in the graph, and taking the mechanical connections or functional dependencies between key components as edges between nodes;

[0059] Step S73: taking the comprehensive health index of each key component at time t as the initial feature of the corresponding node;

[0060] For each node, calculate the average of the initial features of the node and all adjacent nodes as the initial new feature representation;

[0061] Step S74: For each node v, perform a linear transformation based on the new feature representation ,in, For new feature representation, represents the characteristic coefficient, q represents the current number of iterations, represents the bias term;

[0062] According to the current linear transformation, the ReLU activation function is applied to obtain the new feature representation of the node ,in Represents the ReLU activation function;

[0063] Step S75: iterate step S4 until the difference between the new feature representation obtained in the current iteration and the new feature representation obtained in the previous iteration is less than a preset feature difference threshold, and then stop iterating;

[0064] Step S76: Based on the new feature representations of all nodes, the comprehensive health index of the wind turbine is calculated by weighted average method;

[0065] Characteristic coefficient and bias The way to obtain is:

[0066] Obtain the historical health index and corresponding real comprehensive health index of each key component of the wind turbine;

[0067] The Xavier method is used to and Assign an initial value;

[0068] Based on the training data set, the initial feature coefficients and bias terms are applied to calculate a temporary predicted comprehensive health index;

[0069] The mean square error is used as the loss function to evaluate the difference between the temporary predicted comprehensive health index and the true comprehensive health index. The gradient of the loss function relative to the feature coefficient and bias term is calculated through the back propagation algorithm. According to the calculated gradient, the feature coefficient and bias term are updated using the gradient descent method to minimize the loss function. This process is repeated until the value of the loss function no longer changes, and the optimal feature coefficient and bias term are obtained.

[0070] Furthermore, the method of evaluating the current state of the wind turbine generator system based on the comprehensive health index, determining whether the equipment is abnormal, and generating the equipment status signal includes:

[0071] The equipment status signal includes the equipment normal signal and the equipment abnormal signal;

[0072] Set a normal operating threshold of the comprehensive health index. If the comprehensive health index is less than the normal operating threshold, the wind turbine is determined to be in a normal state and a normal equipment signal is generated.

[0073] If the comprehensive health index is greater than or equal to the normal operating threshold, the wind turbine is determined to be in an abnormal state and an equipment abnormality signal is generated.

[0074] Furthermore, the method of setting the normal operation threshold of the comprehensive health index includes:

[0075] According to the comprehensive health index at time t-1 , calculate the benchmark value ;

[0076] Where a represents the smoothing factor of the reference value, It represents the baseline value of the comprehensive health index at time t-2, and is defined as is the average comprehensive health index of the wind turbine in the first h moments during the monitoring period, express The rate of change at time t-2 is defined as is the estimated value of the comprehensive health index change rate of the wind turbine at the first h moments in the monitoring period, It represents the comprehensive health index of the wind turbine at the first moment in the monitoring period. express The baseline value at time t-1;

[0077] according to The baseline value at time t-1 is calculated The rate of change at time t-1 , where b represents the smoothing factor of the rate of change;

[0078] Pick and The sum of the two is used as the normal operating threshold of the comprehensive health index of the wind turbine at time t.

[0079] Furthermore, the method of triggering an alarm mechanism based on the device status signal and generating an abnormality assessment report includes:

[0080] If an abnormal signal from the equipment is detected, an alarm mechanism is triggered. The alarm time, CHI value, health index of each key component, and the specific values ​​of each indicator are recorded. Through the fault tree analysis method, with the CHI value exceeding the normal threshold as the top event, the sub-events that cause the CHI increase are decomposed, and the basic events that affect the CHI increase are identified.

[0081] With time as the horizontal axis and the CHI value at each moment as the vertical axis, draw a CHI change trend graph;

[0082] The CHI change trend chart, the alarm time, the health index of each key component, the specific value of each indicator and the basic events are integrated into an abnormality assessment report, which is sent to maintenance personnel in the form of an alarm notification.

[0083] The technical effects and advantages of the wind turbine condition monitoring and maintenance decision support system of the present invention are as follows:

[0084] This invention collects high-frequency, high-quality data from various parts of a wind turbine (such as the gearbox and generator) and the external environment (such as wind speed and temperature). It constructs key features that characterize equipment status changes from multiple dimensions, including temperature fluctuation indicators, comprehensive vibration indicators, comprehensive wear and tear health indicators, and comprehensive environmental adaptability indicators. These features are then correlated and analyzed to produce a Comprehensive Health Index (CHI) that comprehensively reflects the overall health of the equipment. Unlike traditional single-indicator monitoring methods, this method integrates multi-source, heterogeneous data to improve the accuracy and reliability of condition assessment and enable early warning. Based on historical data analysis of the CHI, the normal operating threshold of the CHI is dynamically updated to ensure that it adapts to the actual operating conditions of the equipment. This not only improves monitoring accuracy but also enables the timely identification of potential problems, guiding specific maintenance and management actions, and ensuring safe and stable equipment operation. This invention places particular emphasis on monitoring and assessment of environmental conditions, ensuring reliable condition monitoring services for users in all circumstances, overcoming the reliability issues of existing technologies in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 A schematic diagram of a wind turbine condition monitoring and maintenance decision support system according to the present invention;

[0086] Figure 2Schematic diagram of a wind turbine condition monitoring and maintenance decision support method according to the present invention. DETAILED DESCRIPTION

[0087] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0088] Example 1

[0089] See also Figure 1 As shown, the wind turbine condition monitoring and maintenance decision support system described in this embodiment includes:

[0090] Data construction module: used to collect temperature fluctuation indicators, comprehensive vibration indicators, comprehensive wear and health indicators and comprehensive environmental adaptability indicators;

[0091] Data analysis module: used to calculate the comprehensive health index of key components through temperature fluctuation index, comprehensive vibration index, comprehensive wear health index and comprehensive environmental adaptability index, and perform correlation analysis on the comprehensive health index of each key component to obtain the comprehensive health index of the wind turbine;

[0092] Status assessment module: used to assess the current status of the wind turbine based on the comprehensive health index, determine whether the equipment is abnormal, and generate equipment status signals;

[0093] Abnormal alarm module: used to trigger the alarm mechanism according to the equipment status signal and generate abnormal assessment report;

[0094] The temperature fluctuation index is collected in the following ways:

[0095] In a monitoring cycle, the monitoring cycle is divided into n moments, and the temperature data of each key component in the monitoring cycle at each moment is collected using thermocouples and arranged in time sequence to obtain a time series array;

[0096] Among them, key components include the wind turbine's main shaft, bearings, generator, hydraulic brake assembly and gearbox;

[0097] Set a time window with a step size of wt. For any time t in the time series array, calculate the average value in the time window before time t, record it as the moving average temperature, and then obtain the absolute difference between the moving average temperature and the temperature data at time t. Record the absolute difference as the temperature fluctuation index at time t.

[0098] By setting a time window and calculating the moving average temperature and temperature fluctuation index of each key component at each moment, this method can dynamically and real-time monitor the temperature changes of key wind turbine components. Utilizing time-series temperature data collected by thermocouples, combined with moving average and temperature fluctuation indicators, it can more accurately capture abnormal temperature fluctuations, provide early warning of potential failures, and optimize maintenance strategies. This method not only improves the ability to perceive the health status of equipment, but also supports intelligent maintenance decisions based on status, thereby effectively improving system reliability and safety, reducing unplanned downtime, and extending the service life of key components.

[0099] The comprehensive vibration index collection method includes:

[0100] Step S31: Using the start time of the monitoring period as the initial time, using the accelerometer to collect the vibration signal data of each key component in real time, and preprocessing the vibration signal data between the initial time and time t, including using median filtering to remove high-frequency noise and normalization;

[0101] Step S32: Using a high-pass filter and a low-pass filter to perform convolution processing on the preprocessed vibration signal data, respectively obtaining a first-level low-pass filtering result Yd1[n] and a first-level high-pass filtering result Yg1[n], which are filtered vibration signal data and still maintain the same length sequence as the preprocessed vibration signal data;

[0102] Where n is a discrete time index, representing the nth sampling point in the vibration signal data;

[0103] Step S33: Sampling Yd1[n] and Yg1[n] respectively, including taking a sample from every other sampling point in Yd1[n] to obtain an approximate coefficient sequence Ad1, and taking a sample from every other sampling point in Yg1[n] to obtain a detail coefficient sequence Dg1 (for example, assuming that the length of the original vibration signal data is N, the lengths of Ad1 and Dg1 are both N / 2);

[0104] Among them, a sample represents the data at a sampling point;

[0105] Step S34: using a high-pass filter and a low-pass filter to perform convolution processing on the approximate coefficient sequence Ad1 to obtain a secondary high-pass filtering result Yg2[n] and a secondary low-pass filtering result Yd2[n];

[0106] Process Yd2[n] and Yg2[n] using the same sampling method as step S33 to obtain the approximate coefficient sequence Ad2 and the detail coefficient sequence Dg2;

[0107] Step S35: Repeat the iterative step S34, use the high-pass filter and low-pass filter to convolve and sample the approximate coefficient sequence Adh-1 of the previous level until the set number of times H is reached, and obtain the approximate coefficient sequence Ad of each level H and detail coefficient sequence Dg H , where h represents the level;

[0108] Step S36: For each level h, calculate the energy distribution of the detail coefficients at each level ,in, is the detail coefficient at the hth level, Indicates the number of detail coefficients at level h, represents the total energy at level h;

[0109] Take the mean of the total energy at all levels as the time-frequency domain feature of the key component at time t;

[0110] Step S37: Arranging the pre-processed vibration signal data in time series to form a vibration time series array;

[0111] Take the mean of the absolute values ​​of all data in the vibration time series array as the absolute average value, calculate the ratio of the maximum absolute value in the vibration time series array to the absolute average value, and obtain the vibration pulse index of the key component at time t;

[0112] Step S38: For each key component, create a comprehensive vibration index of the key component using the time-frequency domain characteristics and the vibration pulse index , where WFD represents the time-frequency domain feature and TFF represents the vibration pulse index;

[0113] It should be noted that the comprehensive vibration index obtained by combining the vibration pulse index with the time-frequency domain characteristics can provide a more comprehensive and sensitive condition monitoring perspective. This approach combines the advantages of time-frequency domain and frequency domain analysis, and can provide a more robust and reliable condition monitoring solution in complex and changing working environments.

[0114] The comprehensive wear health indicator is collected in the following ways:

[0115] A magnetic probe is used to absorb metal chips from the oil of each key component. The magnetic field change on the magnetic probe at time t is measured by a magnetic induction sensor. The magnetic induction sensor outputs a corresponding current signal, which is compared with the calibration curve to obtain the corresponding metal element concentration.

[0116] The standard curve is the relationship curve between the current signal and the magnetic field change established in the laboratory;

[0117] The calibration curve is drawn as follows:

[0118] Prepare a series of standard oil samples with known metal element concentrations, place a magnetic probe into each standard oil sample, and adsorb the magnetic particles;

[0119] A magnetic induction sensor is used to record the magnetic field changes on the magnetic probe after each experiment and output the corresponding current signal;

[0120] Record the metal element concentration of each standard oil sample and its corresponding current signal, and draw a standard curve with the metal element concentration as the horizontal axis and the current signal as the vertical axis;

[0121] The difference in metal element concentration between any moment t and the previous moment t-1 within the monitoring period is taken as the concentration difference, and arranged in time sequence to obtain a concentration difference array;

[0122] Using a time window with a step size of wt, for any moment t on the concentration difference array, calculate the average value in the time window before moment t as the wear metal content trend of the key component at moment t;

[0123] Constructing a comprehensive wear health index by trending the wear metal content of key components , where WMCT represents the wear metal content trend, represents the oil health index of key components at time t, is a preset fixed scale factor, ;

[0124] Methods for obtaining the oil health index include:

[0125] Use a viscometer to measure the viscosity of the oil in each key component at time t, and take the ratio of the oil viscosity at time t to the oil viscosity at the initial time as the change in the oil viscosity in each key component;

[0126] The oil contamination degree at time t is obtained by measuring the number of pollutants in the oil of each key component using a particle counter. The ratio of the oil contamination degree at time t to the oil contamination degree at the initial time is taken as the oil contamination degree change.

[0127] Use acid-base titration to measure the acid value of the oil in each key component at time t, and take the ratio of the oil acid value at time t to the acid value of the new oil as the oil oxidation amount;

[0128] The oil health index of each key component at time t is obtained by multiplying the change in oil viscosity, oil contamination, and oil oxidation.

[0129] Combining wear metal content trends with the oil health index to calculate a comprehensive wear index allows for a more accurate assessment of equipment wear, enabling early warning of failures, optimizing maintenance strategies, extending the life of critical components, and improving overall system reliability and safety. This approach not only provides a comprehensive, sensitive, and multi-dimensional assessment but also reduces maintenance costs through data-driven decision-making, ensuring wind turbines are always in optimal working condition, reducing unplanned downtime, and preventing potential safety incidents.

[0130] The collection method of the comprehensive environmental adaptability index includes:

[0131] Step S51: collecting environmental data of the wind turbine generator system at each moment during the monitoring period through a weather station, including temperature, humidity, wind speed, particulate matter, salt spray, rainfall, and air pressure, to form an original environmental data set;

[0132] Step S52: Standardize each environmental factor in the original environmental data set. , calculate its probability distribution , where i represents the i-th environmental factor, j is the index of the moment in the monitoring period, and n represents the number of moments in the monitoring period;

[0133] Step S53: Calculate the information entropy of each environmental factor based on the probability distribution of each environmental factor ,in, Is the normalization coefficient, ensuring that the value of information entropy is in the range of [0,1];

[0134] Step S54: Calculate the weight of each environmental factor based on the information entropy of each environmental factor , where m is the number of environmental factors, represents the coefficient of variation of the i-th environmental factor;

[0135] Step S55: Based on each environmental factor and its corresponding weight, use the weighted summation formula to calculate the comprehensive environmental adaptability index of the environment where the wind turbine is located at time t. , where j=t;

[0136] The comprehensive environmental adaptability index of wind turbines calculated through the above steps can comprehensively and quantitatively evaluate the operational adaptability of wind turbines in complex and changing environments. This method not only considers the influence of multiple environmental factors such as temperature, humidity, and wind speed, but also assigns reasonable weights to each factor through standardization, probability distribution, and information entropy calculation to ensure the scientificity and accuracy of the evaluation results. The resulting comprehensive environmental adaptability index can more accurately reflect the actual operating conditions of wind turbines, optimize operation and maintenance strategies, provide early warning of potential risks, improve system reliability and safety, and support intelligent decision-making, thereby effectively reducing maintenance costs and extending equipment life.

[0137] The method of calculating the health index of key components by using the temperature fluctuation index, the comprehensive vibration index, the comprehensive wear health index and the comprehensive environmental adaptability index includes:

[0138] Step S61: standardizing the temperature fluctuation index, comprehensive vibration index, comprehensive wear health index, and comprehensive environmental adaptability index at each time t to form an index data set;

[0139] Step S62: For each indicator type, an ideal optimal indicator value is set (the setting of the ideal optimal indicator value is discussed by relevant industry personnel and formulated according to industry standards or specifications) as a reference indicator;

[0140] For each indicator u at time t, calculate the absolute difference between it and the reference indicator ,in, is the value of the u-th indicator at time t, is the value of the reference indicator at time t, where the reference indicator is represented by the number 0;

[0141] According to the absolute difference of each indicator at time t, find the minimum absolute difference and maximum value ; That is, among the 4 absolute differences of the 4 indicators, find the largest and smallest absolute differences;

[0142] Step S63: Calculate the correlation coefficient of each indicator at time t based on the minimum and maximum absolute differences ,in, represents the correlation coefficient of the u-th indicator at time t;

[0143] Take the mean of the correlation coefficients of all indicators at time t as the health index of the key components of the wind turbine at time t;

[0144] The method of performing correlation analysis on the health index of each key component to obtain the comprehensive health index of the wind turbine includes:

[0145] Step S71: Taking each two key components as a key component pair, obtain the mechanical connection relationship between each key component pair (this refers to the situation where two key components are directly connected through a physical connection or mechanical interface. This connection is usually an explicit, physical connection, meaning that the action of one component directly affects the state of the other component. For example, a gearbox and a generator are usually directly connected through a shaft. Therefore, there is a mechanical connection relationship between them. If a problem with the gearbox occurs, the generator may not operate normally).

[0146] For pairs of critical components that have no mechanical connection, obtain their functional dependencies (this refers to two critical components that have no direct physical connection but are functionally dependent on each other, that is, the normal operation of one component depends on the working status of the other component, even if there is no direct mechanical connection between them. For example, for non-direct-drive wind turbines, the main shaft transmits the mechanical energy generated by the rotation of the blades to the gearbox. If there is a problem with the main shaft, the gearbox may not be able to normally receive and process the input torque, thereby affecting efficiency and life);

[0147] Step S72: Construct a graph structure, with each key component as a node in the graph, and the mechanical connections or functional dependencies between the key components as edges between the nodes (for example, there is a direct mechanical connection between the gearbox and the generator, which can be represented as an edge; there is an electrical connection between the generator and the inverter, which can also be represented as an edge. Assume that we have the following key components: node JD1: gearbox, node JD2: generator, node JD3: blade, node JD4: inverter, the connection relationship between these components can be represented as: edge (JD1, JD2): mechanical connection between gearbox and generator, edge (JD2, JD4): electrical connection between generator and inverter, edge (JD2, JD3): functional dependency between generator and blade);

[0148] Step S73: taking the comprehensive health index of each key component at time t as the initial feature of the corresponding node;

[0149] For each node, calculate the average of the initial features of the node and all adjacent nodes as the initial new feature representation;

[0150] Step S74: For each node v, perform a linear transformation based on the new feature representation ,in, For new feature representation, represents the characteristic coefficient, q represents the current number of iterations, represents the bias term;

[0151] According to the current linear transformation, the ReLU activation function is applied to obtain the new feature representation of the node ,in Represents the ReLU activation function;

[0152] Step S75: iterate step S4 until the difference between the new feature representation obtained in the current iteration and the new feature representation obtained in the previous iteration is less than a preset feature difference threshold, and then stop iterating;

[0153] Step S76: Based on the new feature representations of all nodes, the comprehensive health index of the wind turbine is calculated by weighted average method;

[0154] Characteristic coefficient and bias The way to obtain is:

[0155] Collect training data sets, including historical health indices of key components of wind turbines and corresponding real comprehensive health indices;

[0156] The Xavier method is used to and Assign an initial value;

[0157] Based on the training data set, the initial feature coefficients and bias terms are applied to calculate a temporary predicted comprehensive health index;

[0158] Using mean square error as the loss function, the difference between the temporary predicted comprehensive health index and the true comprehensive health index is evaluated. The gradient of the loss function with respect to the characteristic coefficients and bias terms is calculated through the back-propagation algorithm. Based on the calculated gradient, the characteristic coefficients and bias terms are updated using the gradient descent method to minimize the loss function. This process is repeated until the value of the loss function no longer changes, and the optimal characteristic coefficients and bias terms are obtained.

[0159] It should be noted that by combining multiple key indicators, the comprehensive health index provides a comprehensive perspective for equipment status assessment, taking into account not only internal mechanical wear but also the influence of external environmental conditions. It can more accurately reflect the overall health of the equipment and avoid the one-sidedness that may be caused by a single indicator. This index supports real-time monitoring and early warning, can dynamically monitor the operating status of wind turbines, promptly identify and address potential problems, achieve early warning, and reduce the probability of sudden failures. Based on historical data and trend analysis of the comprehensive health index, maintenance strategies can be optimized and scientific and reasonable predictive maintenance plans can be formulated to reduce maintenance costs and downtime, while improving equipment reliability and service life. In addition, the comprehensive health index helps improve operational efficiency. Through performance optimization and decision support, it ensures that wind turbines operate within a safe range and reduces failure rates and downtime losses.

[0160] The method of evaluating the current status of the wind turbine according to the comprehensive health index, determining whether the equipment is abnormal, and generating the equipment status signal includes:

[0161] The equipment status signal includes the equipment normal signal and the equipment abnormal signal;

[0162] setting a normal operating threshold of the comprehensive health index, and comparing the comprehensive health index with the normal operating threshold;

[0163] If the comprehensive health index is less than the normal operating threshold, the wind turbine is determined to be in a normal state and a normal equipment signal is generated;

[0164] If the comprehensive health index is greater than or equal to the normal operating threshold, the wind turbine is determined to be in an abnormal state and an equipment abnormality signal is generated;

[0165] The method for setting the normal operation threshold of the comprehensive health index includes:

[0166] According to the comprehensive health index at time t-1 , calculate the benchmark value ;

[0167] Where a represents the smoothing factor of the reference value (between 0 and 1, usually 0.5), It represents the baseline value of the comprehensive health index at time t-2, and is defined as is the average comprehensive health index of the wind turbine in the first h moments during the monitoring period, express The rate of change at time t-2 is defined as is the estimated rate of change of the comprehensive health index of the wind turbine at the first h moments in the monitoring period, It represents the comprehensive health index of the wind turbine at the first moment in the monitoring period. express The baseline value at time t-1;

[0168] according to The baseline value at time t-1 is calculated The rate of change at time t-1 , where b represents the smoothing factor of the rate of change (between 0 and 1, usually 0.5);

[0169] Pick and The sum of is used as the normal operating threshold of the comprehensive health index of the wind turbine at time t;

[0170] The method of triggering the alarm mechanism according to the device status signal and generating the abnormality assessment report includes:

[0171] When an abnormal signal from equipment is detected, the alarm mechanism is automatically activated, recording the alarm time, CHI value, health index of each key component, and the specific values ​​of each indicator. Through fault tree analysis, with the CHI value exceeding the normal threshold as the top event, the sub-events that caused the CHI increase are decomposed, and the basic events that affected the CHI increase are identified;

[0172] With time as the horizontal axis and the CHI value at each moment as the vertical axis, draw a CHI change trend graph;

[0173] The CHI change trend chart, the alarm time, the health index of each key component, the specific value of each indicator, and basic events are integrated into an abnormality assessment report, which is then sent to maintenance personnel in the form of an alarm notification.

[0174] Provides an intuitive and easy-to-use visual operation panel to display relevant status information of wind turbines, making it easier for maintenance personnel to understand and operate.

[0175] This embodiment collects high-frequency, high-quality data from different parts of the wind turbine (such as the gearbox and generator) and the external environment (such as wind speed and temperature). From this data, important features that can characterize equipment status changes from multiple dimensions are constructed, including temperature fluctuation indicators, comprehensive vibration indicators, comprehensive wear and health indicators, and comprehensive environmental adaptability indicators. Correlation analysis is then performed on these indicators to obtain a comprehensive health index (CHI) that comprehensively reflects the overall health of the equipment. Unlike traditional single-indicator monitoring methods, this method integrates multi-source heterogeneous data to improve the accuracy and reliability of status assessment and enable early warning. Based on historical data analysis of the CHI, the normal operating threshold of the CHI is dynamically updated to ensure that it adapts to the actual operating conditions of the equipment. This not only improves monitoring accuracy but also enables the timely identification of potential problems, guiding specific maintenance and management actions, and ensuring safe and stable operation of the equipment. This embodiment places particular emphasis on monitoring and assessment of environmental conditions, ensuring reliable status monitoring services for users in all circumstances, overcoming the reliability issues of existing technologies in complex environments.

[0176] Example 2

[0177] See also Figure 2 As shown, for the parts not described in detail in this embodiment, please refer to the description of embodiment 1. A method for monitoring the state of a wind turbine and supporting maintenance decisions is provided, including:

[0178] S1. Collect temperature fluctuation index, comprehensive vibration index, comprehensive wear health index and comprehensive environmental adaptability index;

[0179] S2. Correlation analysis is performed on the temperature fluctuation index, the comprehensive vibration index, the comprehensive wear health index, and the comprehensive environmental adaptability index to obtain a comprehensive health index of the wind turbine;

[0180] S3. Evaluate the current status of the wind turbine according to the comprehensive health index, determine whether the equipment is abnormal, and generate an equipment status signal;

[0181] S4. According to the equipment status signal, the alarm mechanism is triggered and an abnormality assessment report is generated.

[0182] Example 3

[0183] This embodiment discloses an electronic device, including 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 operation mode of the above-mentioned method for wind turbine condition monitoring and maintenance decision support is implemented.

[0184] Since the electronic device introduced in this embodiment is an electronic device used to implement a method for wind turbine status monitoring and maintenance decision support in the embodiment of this application, based on the method for wind turbine status monitoring and maintenance decision support introduced in the embodiment of this application, those skilled in the art can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application will not be described in detail here. As long as those skilled in the art implement the electronic device used in the method for wind turbine status monitoring and maintenance decision support in the embodiment of this application, they are within the scope of protection to be provided by this application.

[0185] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0186] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that for users of ordinary skill in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A wind turbine condition monitoring and maintenance decision support system, characterized in that: include: Data construction module: used to collect temperature fluctuation indicators, comprehensive vibration indicators, comprehensive wear and health indicators and comprehensive environmental adaptability indicators; The collection methods of comprehensive vibration indicators include: Step S31: Using the start time of the monitoring period as the initial time, using the accelerometer to collect the vibration signal data of each key component in real time, and preprocessing the vibration signal data between the initial time and time t, including using median filtering to remove high-frequency noise and normalization; Step S32: Using a high-pass filter and a low-pass filter to perform convolution processing on the preprocessed vibration signal data, respectively obtaining a first-level low-pass filtering result Yd1[n] and a first-level high-pass filtering result Yg1[n], which are filtered vibration signal data and still maintain the same length sequence as the preprocessed vibration signal data; Where n is a discrete time index, representing the nth sampling point in the vibration signal data; Step S33: sampling Yd1[n] and Yg1[n] respectively, including taking a sample from every other sampling point in Yd1[n] to obtain an approximate coefficient sequence Ad1, and taking a sample from every other sampling point in Yg1[n] to obtain a detail coefficient sequence Dg1; Among them, a sample represents the data at a sampling point; Step S34: using a high-pass filter and a low-pass filter to perform convolution processing on the approximate coefficient sequence Ad1 to obtain a secondary high-pass filtering result Yg2[n] and a secondary low-pass filtering result Yd2[n]; Process Yd2[n] and Yg2[n] using the same sampling method as step S33 to obtain the approximate coefficient sequence Ad2 and the detail coefficient sequence Dg2; Step S35: Repeat the iterative step S34, using a high-pass filter and a low-pass filter to convolve and sample the approximate coefficient sequence Adh-1 of the previous level until the set number of times H is reached, to obtain the approximate coefficient sequence AdH and detail coefficient sequence DgH of each level, where h represents the level; Step S36: For each level h, calculate the energy distribution of the detail coefficients at each level ,in, is the detail coefficient at the hth level, Indicates the number of detail coefficients at level h, represents the total energy at level h; Take the mean of the total energy at all levels as the time-frequency domain feature of the key component at time t; Step S37: Arranging the pre-processed vibration signal data in time series to form a vibration time series array; Take the mean of the absolute values ​​of all data in the vibration time series array as the absolute average value, calculate the ratio of the maximum absolute value in the vibration time series array to the absolute average value, and obtain the vibration pulse index of the key component at time t; Step S38: For each key component, create a comprehensive vibration index of the key component using the time-frequency domain characteristics and the vibration pulse index , where WFD represents the time-frequency domain feature and TFF represents the vibration pulse index; Data analysis module: used to calculate the comprehensive health index of key components through temperature fluctuation index, comprehensive vibration index, comprehensive wear health index and comprehensive environmental adaptability index, and perform correlation analysis on the comprehensive health index of each key component to obtain the comprehensive health index of the wind turbine; Status assessment module: used to assess the current status of the wind turbine based on the comprehensive health index, determine whether the equipment is abnormal, and generate equipment status signals; Abnormal alarm module: used to trigger the alarm mechanism according to the equipment status signal and generate abnormal assessment report.

2. A wind turbine condition monitoring and maintenance decision support system according to claim 1, characterized in that: The temperature fluctuation index is collected in the following ways: In a monitoring cycle, the monitoring cycle is divided into n moments. The temperature data of each key component in the monitoring cycle at each moment is collected using thermocouples and arranged in time sequence to obtain a time series array. Among them, key components include the wind turbine's main shaft, bearings, generator, hydraulic brake assembly and gearbox; Set a time window with a step size of wt. For any time t in the time series array, calculate the average value in the time window before time t, record it as the moving average temperature, then obtain the absolute difference between the moving average temperature and the temperature data at time t, and record the absolute difference as the temperature fluctuation index at time t.

3. A wind turbine condition monitoring and maintenance decision support system according to claim 2, characterized in that: The comprehensive wear health indicator is collected in the following ways: A magnetic probe is used to absorb metal chips from the oil of each key component. The magnetic field change on the magnetic probe at time t is measured by a magnetic induction sensor. The magnetic induction sensor outputs a corresponding current signal, which is compared with the calibration curve to obtain the corresponding metal element concentration. The calibration curve is the relationship curve between the current signal and the magnetic field change established in the laboratory; The calibration curve is drawn as follows: Prepare a series of standard oil samples with known metal element concentrations, place a magnetic probe into each standard oil sample, and adsorb the magnetic particles; A magnetic induction sensor is used to record the magnetic field changes on the magnetic probe after each experiment and output the corresponding current signal; Record the metal element concentration of each standard oil sample and its corresponding current signal, and draw a calibration curve with the metal element concentration as the horizontal axis and the current signal as the vertical axis; The difference in metal element concentration between any moment t and the previous moment t-1 within the monitoring period is taken as the concentration difference, and arranged in time sequence to obtain a concentration difference array; Using a time window with a step size of wt, for any moment t on the concentration difference array, calculate the average value in the time window before moment t as the wear metal content trend of the key component at moment t; Constructing a comprehensive wear health index by trending the wear metal content of key components , where WMCT represents the wear metal content trend, represents the oil health index of key components at time t, is a preset fixed scale factor, ; The collection methods of oil health indicators include: Use a viscometer to measure the viscosity of the oil in each key component at time t, and take the ratio of the oil viscosity at time t to the oil viscosity at the initial time as the change in the oil viscosity in each key component; The oil contamination degree at time t is obtained by measuring the number of pollutants in the oil of each key component using a particle counter. The ratio of the oil contamination degree at time t to the oil contamination degree at the initial time is taken as the oil contamination degree change. Use acid-base titration to measure the acid value of the oil in each key component at time t, and take the ratio of the oil acid value at time t to the acid value of the new oil as the oil oxidation amount; The oil health index of each key component at time t is obtained by multiplying the change in oil viscosity, oil contamination, and oil oxidation.

4. A wind turbine condition monitoring and maintenance decision support system according to claim 3, characterized in that: The collection method of the comprehensive environmental adaptability index includes: Step S51: collecting environmental data of the wind turbine generator system at each moment during the monitoring period through a weather station, including temperature, humidity, wind speed, particulate matter, salt spray, rainfall, and air pressure, to form an original environmental data set; Step S52: Standardize each environmental factor in the original environmental data set. , calculate its probability distribution , where i represents the i-th environmental factor, j is the index of the moment in the monitoring period, and n represents the number of moments in the monitoring period; Step S53: Calculate the information entropy of each environmental factor based on the probability distribution of each environmental factor ,in, is the normalization coefficient; Step S54: Calculate the weight of each environmental factor based on the information entropy of each environmental factor , where m is the number of environmental factors; Step S55: Based on each environmental factor and the corresponding weight, a weighted summation formula is used to calculate the comprehensive environmental adaptability index of the environment in which the wind turbine generator system is located at time t.

5. A wind turbine condition monitoring and maintenance decision support system according to claim 4, characterized in that: The method of calculating the health index of key components by using the temperature fluctuation index, the comprehensive vibration index, the comprehensive wear health index and the comprehensive environmental adaptability index includes: Step S61: standardizing the temperature fluctuation index, comprehensive vibration index, comprehensive wear health index, and comprehensive environmental adaptability index at each time t to form an index data set; Step S62: For each indicator type, set an ideal optimal indicator value as a reference indicator; For each indicator u at time t, calculate the absolute difference between it and the reference indicator ,in, is the value of the u-th indicator at time t, is the value of the reference indicator at time t; According to the absolute difference of each indicator at time t, find the minimum absolute difference and maximum value ; Step S63: Calculate the correlation coefficient of each indicator at time t based on the minimum and maximum absolute differences ,in, represents the correlation coefficient of the u-th indicator at time t; The mean correlation coefficient of all indicators at time t is taken as the health index of each key component at time t.

6. A wind turbine condition monitoring and maintenance decision support system according to claim 5, characterized in that: The method of performing correlation analysis on the health index of each key component to obtain the comprehensive health index of the wind turbine includes: Step S71: taking every two key components as a key component pair, and obtaining the mechanical connection relationship between each key component pair; For key component pairs that do not have mechanical connection relationships, obtain their functional dependencies; Step S72: constructing a graph structure, taking each key component as a node in the graph, and taking the mechanical connections or functional dependencies between key components as edges between nodes; Step S73: taking the comprehensive health index of each key component at time t as the initial feature of the corresponding node; For each node, calculate the average of the initial features of the node and all adjacent nodes as the initial new feature representation; Step S74: For each node v, perform a linear transformation based on the new feature representation ,in, For new feature representation, represents the characteristic coefficient, q represents the current number of iterations, represents the bias term; According to the current linear transformation, the ReLU activation function is applied to obtain the new feature representation of the node ,in Represents the ReLU activation function; Step S75: iterate step S74 until the difference between the new feature representation obtained in the current iteration and the new feature representation obtained in the previous iteration is less than a preset feature difference threshold, and then stop iterating; Step S76: Based on the new feature representations of all nodes, the comprehensive health index of the wind turbine is calculated by weighted average method; Characteristic coefficient and bias The way to obtain is: Collect training data sets, including historical health indices of key components of wind turbines and corresponding real comprehensive health indices; The Xavier method is used to and Assign an initial value; Based on the training data set, the initial feature coefficients and bias terms are applied to calculate a temporary predicted comprehensive health index; The mean square error is used as the loss function to evaluate the difference between the temporary predicted comprehensive health index and the true comprehensive health index. The gradient of the loss function relative to the feature coefficient and bias term is calculated through the back propagation algorithm. According to the calculated gradient, the feature coefficient and bias term are updated using the gradient descent method to minimize the loss function. This process is repeated until the value of the loss function no longer changes, and the optimal feature coefficient and bias term are obtained.

7. A wind turbine condition monitoring and maintenance decision support system according to claim 6, characterized in that: The method of evaluating the current status of the wind turbine according to the comprehensive health index, determining whether the equipment is abnormal, and generating the equipment status signal includes: The equipment status signal includes the equipment normal signal and the equipment abnormal signal; Set a normal operating threshold of the comprehensive health index. If the comprehensive health index is less than the normal operating threshold, the wind turbine is determined to be in a normal state and a normal equipment signal is generated. If the comprehensive health index is greater than or equal to the normal operating threshold, the wind turbine is determined to be in an abnormal state and an equipment abnormality signal is generated.

8. A wind turbine condition monitoring and maintenance decision support system according to claim 7, characterized in that: The method for setting the normal operation threshold of the comprehensive health index includes: According to the comprehensive health index at time t-1 , calculate the benchmark value ; Where a represents the smoothing factor of the reference value, It represents the baseline value of the comprehensive health index at time t-2, and is defined as is the average comprehensive health index of the wind turbine in the first h moments during the monitoring period, express The rate of change at time t-2 is defined as is the estimated value of the comprehensive health index change rate of the wind turbine at the first h moments in the monitoring period, It represents the comprehensive health index of the wind turbine at the first moment in the monitoring period. express The baseline value at time t-1; according to The baseline value at time t-1 is calculated The rate of change at time t-1 , where b represents the smoothing factor of the rate of change; Pick and The sum of the two is used as the normal operating threshold of the comprehensive health index of the wind turbine at time t.

9. A wind turbine condition monitoring and maintenance decision support system according to claim 8, characterized in that: The method of triggering the alarm mechanism according to the device status signal and generating the abnormality assessment report includes: If an abnormal signal from the equipment is detected, an alarm mechanism is triggered. The alarm time, CHI value, health index of each key component, and the specific values ​​of each indicator are recorded. Through the fault tree analysis method, with the CHI value exceeding the normal threshold as the top event, the sub-events that cause the CHI increase are decomposed, and the basic events that affect the CHI increase are identified. With time as the horizontal axis and the CHI value at each moment as the vertical axis, draw a CHI change trend graph; The CHI change trend chart, the alarm time, the health index of each key component, the specific value of each indicator and the basic events are integrated into an abnormality assessment report, which is sent to maintenance personnel in the form of an alarm notification.

Images