CN118653970A - A method and system for correcting the warning threshold of wind turbine operation status - Google Patents

Abstract

The present invention provides a method and system for correcting the threshold value of early warning of the operation status of a wind turbine set, which relates to the field of wind power generation. The method collects the operation data of the wind turbine set in real time to form time series data; performs mutation detection on the time series data, marks the points exceeding the threshold value as abnormal points, and records the position, amplitude and duration of each abnormal point; extracts the time series features, frequency domain features and statistical features with the abnormal point as the starting point from the historical operation data of the wind turbine set; calculates the mutation amplitude and duration of each abnormal point, uses fast Fourier transform to perform frequency analysis on the data around the abnormal point, and extracts the frequency component; uses the extracted features as input to train a fault prediction model, uses the trained model to predict and evaluate the operation status of the wind turbine set, and identifies the prediction result; and generates an early warning signal. The present invention improves the accuracy and timeliness of early warning by dynamically adjusting the early warning threshold.

Description

A method and system for correcting the warning threshold of wind turbine operation status

Technical Field

The present invention relates to the field of wind power generation, and in particular to a method and system for correcting a warning threshold value of an operating state of a wind turbine set.

Background Art

As an important component of clean energy, wind power generation occupies an increasingly important position in the global energy structure. The safe operation of wind power generation equipment, especially wind turbines, is directly related to the stability and reliability of energy production. Since wind turbines are usually set up in open outdoor environments, the operating status of wind turbines will change during operation due to the influence of various factors such as the environment, load and its own status.

In the prior art, a fixed warning threshold is usually set for monitoring. Traditional wind turbine unit operation monitoring technology is mainly based on fixed threshold settings for fault detection and warning. However, due to the complex and changeable operating environment of the wind turbine unit, the fixed threshold is difficult to adapt to various working conditions, resulting in the traditional method having the problems of weak fixed threshold adaptability, difficulty in feature extraction, and high real-time requirements. The fixed threshold is difficult to adapt to the feature changes of the wind turbine unit under different operating conditions, which is easy to cause false alarms or missed alarms, thereby affecting the safe operation of the wind turbine unit.

Therefore, a method and system are needed to dynamically adjust the warning threshold to improve the accuracy and reliability of the warning.

Summary of the invention

In view of this, in order to solve the above problems, the present invention proposes a method and system for correcting the warning threshold of the operating status of a wind turbine set, which can dynamically adjust the warning threshold according to the actual operating status and environmental conditions of the wind turbine set, thereby improving the accuracy and timeliness of the warning and ensuring the safe operation of the wind turbine set.

The present invention is implemented by the following technical solutions:

In a first aspect, the present invention provides a method for correcting a warning threshold value of an operating state of a wind turbine generator set, the method comprising the following steps:

Use sensors to collect the operating data of the fan unit in real time to form time series data;

Perform mutation detection on the collected time series data, calculate the cumulative sum sequence, dynamically adjust the warning threshold based on historical data and known status, mark the points exceeding the threshold as abnormal points, and record the location, amplitude and duration of each abnormal point;

Extract the time series features, frequency domain features and statistical features starting from the abnormal point from the historical operation data of the wind turbine unit; calculate the mutation amplitude and duration of each abnormal point, use fast Fourier transform to perform frequency analysis on the data around the abnormal point, and extract the frequency components;

The extracted features are used as input to train the fault prediction model, learn the relationship between the features and the operating status of the fan, use the trained model to predict and evaluate the operating status of the fan unit, and identify the prediction results;

An early warning signal is generated based on the key characteristic parameters of the abnormal point's mutation amplitude, duration and frequency component, as well as the prediction results output by the fault prediction model.

As a further solution of the present invention, performing mutation detection on the collected time series data and calculating the cumulative sum sequence includes the following steps:

Get the collected time series data, and calculate the mean and standard deviation of the time series data based on the accumulation;

Select a preset mutation threshold and use the cumulative sum formula to calculate the cumulative sum sequence;

Determine whether the cumulative sum exceeds the mutation threshold to determine whether there is an abnormal point; where:

Time Series Data ,in, is the number of data points for the run data;

The preset mutation threshold is ; The cumulative sum calculation formula is:

;

In the formula, , is the adjustment factor, ; Represents the mean of time series data; represents the cumulative sum at the current moment i; Represents the cumulative sum of the previous moment;

when When a mutation occurs, record the abnormal point , and record the index and corresponding value of the abnormal point.

As a further solution of the present invention, when dynamically adjusting the warning threshold, the collected historical data includes data in normal state and abnormal state, and the statistical characteristics of the mean and standard deviation of the historical data are calculated. The initial threshold is set to the mean plus k times the standard deviation. The mean and standard deviation are recalculated periodically according to the changes in the real-time data stream, and the threshold is adjusted; when calculating the new threshold, a sliding window is used to calculate the mean and standard deviation of the closest N data points, and the threshold is updated; wherein k in k times the standard deviation is a constant, and the value is 2 or 3.

As a further solution of the present invention, points exceeding the threshold are marked as abnormal points, and the position, amplitude and duration of each abnormal point are recorded, including: setting a list as a data structure to store abnormal point information, calculating the abnormal point information, wherein the stored abnormal point information includes the position, amplitude and duration; when an abnormal point is detected, the position, amplitude and duration are recorded, wherein the position is the current timestamp of the abnormal point, and the amplitude is the degree of exceeding the threshold, which is obtained by the difference between the current data point and the current threshold; the duration is the duration of the abnormal state, and if the subsequent data points are still in the abnormal state, the states of multiple abnormal points are continuously recorded and the duration is updated.

As a further solution of the present invention, time series features, frequency domain features and statistical features are extracted with the abnormal point as the starting point. When extracting the time series features, the vibration data within the preset time window is extracted with the abnormal point as the starting point, and the mean, variance and root mean square are calculated based on the time series features of the vibration data; when extracting the frequency domain features, the extracted time series data is subjected to a fast Fourier transform to obtain the frequency domain data and spectrum density of the frequency domain features; when extracting the statistical features, the statistical features are calculated based on the extracted time series and frequency domain features.

As a further solution of the present invention, when calculating the mutation amplitude and duration of each abnormal point, according to the abnormal point found, a number of data points before and after the abnormal point are determined by selecting a window, and the mean and mutation amplitude are calculated, and the duration is calculated for the time period when the abnormal point exceeds a certain threshold in the data, wherein, according to the set threshold of the mutation amplitude, the data points are checked forward and backward from the abnormal point until the data point recovers below the threshold, and the time difference from the time when the abnormal point appears to the time when the data recovers below the threshold is calculated;

When performing frequency analysis, a number of data points before and after the abnormal point are determined by selecting a window, and the frequency resolution and frequency components are calculated using fast Fourier transform to obtain the frequency components;

The formula for calculating frequency resolution is: ;

Among them, the formula for calculating the frequency component is: ;

In the formula, is the sampling frequency, is the number of FFT points, ;

The frequency components are expressed as: ;

in, is the window data, is the frequency component.

As a further solution of the present invention, the extracted features are used as input to train a fault prediction model to learn the relationship between the features and the operating status of the wind turbine, including the following steps:

Based on the collected wind turbine operation data and the extracted time series features, frequency domain features and statistical features starting from the abnormal point, the wind turbine operation data is marked as normal state and fault state;

After data cleaning and standardization of the operating data, the operating data dataset is divided into a training set, a validation set, and a test set;

Use the training set data to train the fault prediction model, select the loss function, use the validation set to tune the hyperparameters, use the test set to evaluate the fault prediction model, use the trained model to monitor and predict the new wind turbine operation data in real time, judge the status of the wind turbine based on the output of the fault prediction model, and identify the prediction results.

As a further solution of the present invention, when generating a warning signal, the fault probability P is predicted according to the identified prediction result, and a threshold judgment is performed, and a warning signal is generated according to the predicted probability; wherein the fault probability P is: ;

in, is the trained fault prediction model. is the mutation amplitude, is the duration, is the frequency component; when judging the threshold: ;in, The preset warning threshold.

The present invention can effectively solve the problems of poor adaptability and high false alarm rate in the monitoring and early warning of the operating status of the fan unit, improve the accuracy and timeliness of the early warning, and ensure the safe operation of the fan unit.

In a second aspect, the present invention further includes a system for correcting a warning threshold value of an operating state of a wind turbine generator set, the system comprising:

A data acquisition module is used to use sensors to collect the operating data of the fan unit in real time to form time series data;

The anomaly monitoring module is used to perform mutation detection on the collected time series data, calculate the cumulative sum series, mark the points exceeding the threshold as anomalies, and record the location, amplitude and duration of each anomaly point;

The feature extraction module is used to extract the time series features, frequency domain features and statistical features starting from the abnormal point from the historical operation data of the wind turbine unit; calculate the mutation amplitude and duration of each abnormal point, use fast Fourier transform to perform frequency analysis on the data around the abnormal point, and extract the frequency component;

A model training module is used to take the extracted features as input, train the fault prediction model, learn the relationship between the features and the operating status of the fan, use the trained model to predict and evaluate the operating status of the fan unit, and identify the prediction results;

The early warning module is used to generate early warning signals based on the key characteristic parameters of the mutation amplitude, duration and frequency component of the abnormal point and the prediction results output by the fault prediction model.

As a further solution of the present invention, the correction system also includes a threshold adjustment module for dynamically adjusting the warning threshold based on historical data and known states.

The correction system combines the advantages of data analysis, feature extraction and machine learning models, and has high accuracy and real-time performance. During specific implementation, it is also necessary to adjust model parameters and algorithm selection according to different devices and application scenarios to improve the adaptability and effectiveness of the system.

The present invention also includes a computer device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor executes the method for correcting the wind turbine operating status warning threshold.

The present invention also includes a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the method for correcting the wind turbine operating status warning threshold.

Compared with the prior art, the method and system for correcting the wind turbine operating status warning threshold provided by the present invention have the following beneficial effects:

1. Real-time monitoring response is achieved, and accurate identification of abnormal points is achieved by dynamically adjusting the warning threshold.

Real-time collection of wind turbine operating data through sensors ensures rapid detection and response to operating status anomalies, improving system safety and reliability. Dynamic adjustment of warning thresholds based on historical data and known status avoids false alarms and missed alarms that may be caused by static thresholds, making the warning mechanism more flexible and accurate. Accurately mark abnormal points using mutation detection and cumulative and sequence analysis, and clearly record the location, magnitude and duration of the anomaly, providing important data support for subsequent analysis.

2. Through multi-dimensional feature extraction and frequency analysis capabilities, the training and application of intelligent fault prediction models can be realized.

Extracting time series features, frequency domain features and statistical features from the abnormal points provides a rich information basis, which helps to fully understand the characteristics of abnormal behavior and its influencing factors; performing frequency analysis on the data around the abnormal points through fast Fourier transform can identify potential periodic problems and improve the depth and breadth of fault prediction; training fault prediction models and learning the relationship between features and wind turbine operating status can predict potential fault risks at an early stage, so that measures can be taken in advance to reduce downtime and maintenance costs.

3. It realizes the generation of early warning signals, reduces maintenance costs, and increases the life of the fan unit.

By combining the key characteristic parameters of mutation amplitude, duration, and frequency components with the output of the prediction model, accurate early warning signals are generated to help managers make timely decisions; by identifying and predicting faults in advance, more scientific maintenance plans can be formulated to reduce the high maintenance costs caused by sudden failures and improve the operating efficiency of the fan; timely early warnings and fault predictions can effectively prevent equipment damage caused by overload or other abnormal conditions, thereby extending the service life of the fan unit.

In summary, the method and system for correcting the wind turbine operating status warning threshold of the present invention greatly improve the operating safety and maintenance efficiency of the wind turbine unit through real-time monitoring, dynamic adjustment, precise identification and intelligent prediction, and have significant economic and social benefits.

These and other aspects of the present invention will become more concise and understandable in the following description of the embodiments. It should be understood that the above general description and the following detailed description are only exemplary and explanatory and cannot limit the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or related technologies, the following briefly introduces the drawings required for use in the exemplary embodiments or related technical descriptions. The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

FIG1 is a flow chart of a method for correcting a warning threshold value of an operating state of a wind turbine generator system according to an embodiment of the present invention.

FIG2 is a flow chart of mutation detection in a method for correcting a warning threshold value of a wind turbine operating state according to an embodiment of the present invention.

FIG3 is a flow chart of training a fault prediction model in a method for correcting a warning threshold value of an operating state of a wind turbine generator system according to an embodiment of the present invention.

DETAILED DESCRIPTION

In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

In some of the processes described in the specification and claims of the present invention and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

The following will be combined with the drawings in the exemplary embodiments of the present invention to clearly and completely describe the technical solutions in the exemplary embodiments of the present invention. Obviously, the described exemplary 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 those skilled in the art without creative work are within the scope of protection of the present invention.

Since the traditional wind turbine operation monitoring technology is mainly based on fixed threshold settings for fault detection and early warning, the wind turbine operation environment is complex and changeable, and the fixed threshold is difficult to adapt to various working conditions, resulting in the traditional method having the problems of weak fixed threshold adaptability, difficulty in feature extraction, and high real-time requirements. The fixed threshold is difficult to adapt to the feature changes of the wind turbine unit under different operating conditions, which easily leads to false alarms or missed alarms, thereby affecting the safe operation of the wind turbine unit. In order to solve this problem, the present invention provides a method and system for correcting the warning threshold of the wind turbine operation state, which can dynamically adjust the warning threshold according to the actual operating state and environmental conditions of the wind turbine unit, improve the accuracy and timeliness of the warning, and ensure the safe operation of the wind turbine unit.

The technical solution of the present invention is further described below in conjunction with specific embodiments:

Referring to FIG. 1 , FIG. 1 is a flow chart of a method for correcting a wind turbine operating status warning threshold provided by the present invention. A method for correcting a wind turbine operating status warning threshold provided in one embodiment of the present invention comprises the following steps:

Step S10: Use sensors to collect operating data of the fan unit in real time to form time series data.

In this step, the operation data of the fan unit is obtained in real time through various sensors installed on the fan unit. These data will be sorted into time series for subsequent analysis and processing. The operation data of the fan unit obtained includes but is not limited to:

1. Mechanical parameters;

Speed (RPM): The actual speed of the fan rotor, providing a direct indicator of the fan's operating status.

Torque (Nm): The torque generated by the rotor, reflecting the load condition of the fan.

Vibration (mm/s): Monitors the vibration levels of various wind turbine components to identify potential mechanical failures.

2. Environmental parameters;

Wind speed (m/s): Wind speed data around the wind turbine, which affects the power generation efficiency and load of the wind turbine.

Wind direction (°): The direction of the wind, which helps determine the optimal working condition of the fan.

Temperature (℃): The temperature inside and outside the fan unit affects the operating status and efficiency of the equipment.

3. Electrical parameters;

Power output (kW): The actual power generated by the wind turbine, reflecting the working status and efficiency of the wind turbine.

Voltage (V): Input and output voltage of the fan motor, monitors the stability of the electrical system.

Current (A): The input current of the motor, which helps identify electrical faults or excessive loads.

4. Hydraulic and lubrication parameters;

Oil temperature (℃): The temperature of the lubricating oil affects the lubrication effect and operation safety of the equipment.

Oil pressure (bar): The pressure of the lubrication system, ensuring normal lubrication of all moving parts.

Liquid level (mm): Liquid level in the oil tank and hydraulic system to avoid malfunctions due to insufficient hydraulic pressure.

5. Condition monitoring data;

Fault Codes: Fault warnings and diagnostic information generated by the fan control system.

Running time (hours): The cumulative running time of the fan, which helps with the maintenance and management of the equipment.

6. Safety monitoring data;

Overload protection status: monitor whether the fan enters the overload state to ensure safe operation.

Emergency stop signal: In the event of an extreme situation, record the time and reason of the emergency stop.

This step can form a comprehensive time series data set through the multi-dimensional operation data collected by the above sensors. These data can not only reflect the operating status of the wind turbine unit in real time, but also provide a basis for subsequent abnormal detection, fault prediction and maintenance decisions. By analyzing these data, potential problems can be discovered in time, the operating efficiency of the wind turbine can be optimized, and the risk of failure can be minimized.

Step S20: perform mutation detection on the collected time series data, calculate the cumulative sum sequence, dynamically adjust the warning threshold based on historical data and known status, mark the points exceeding the threshold as abnormal points, and record the position, amplitude and duration of each abnormal point.

In this step, as shown in FIG2 , mutation detection is performed on the collected time series data, and the calculation of the cumulative sum sequence includes the following steps:

Step S201, obtaining the collected time series data, and calculating the mean and standard deviation of the time series data based on the accumulation;

Step S202: Select a preset mutation threshold and use the cumulative sum formula to calculate the cumulative sum sequence;

Step S203: determine whether the cumulative sum exceeds the mutation threshold to determine whether there is an abnormal point; wherein:

Time Series Data ,in, is the number of data points for the run data;

The preset mutation threshold is ; The cumulative sum calculation formula is:

;

In the formula, , is the adjustment factor, ; is the mean of the time series data, that is, the sample mean calculated before mutation detection, which is used to measure the baseline level of the data. The role is to serve as a reference point and the current observation value The difference between the two is used to determine whether a mutation has occurred;

Specific:

is the cumulative sum of the current moment; It is the cumulative sum of the previous moment; is the observation value at the current moment, which comes from the data point of the time series data; is the mean of the time series data; is the adjustment factor, , used to control the sensitivity of detection; is the preset mutation threshold; by comparing and The difference between Whether it deviates significantly from the mean, thereby identifying potential mutation points;

when When a mutation occurs, record the abnormal point , and record the index and corresponding value of the abnormal point.

Among them, the mean of the time series data is calculated based on the cumulative sum The calculation formulas are:

;

Calculating the standard deviation of time series data based on cumulative sums The calculation formulas are:

.

In this step, when dynamically adjusting the warning threshold, the collected historical data includes data in normal and abnormal states, and the statistical characteristics of the mean and standard deviation of the historical data are calculated. The initial threshold is set to the mean plus k times the standard deviation. The mean and standard deviation are recalculated periodically according to the changes in the real-time data stream, and the threshold is adjusted; when calculating the new threshold, a sliding window is used to calculate the mean and standard deviation of the closest N data points, and the threshold is updated; where k in k times the standard deviation is a constant, and the value is 2 or 3.

Then, update threshold = .

Among them, when marking the points exceeding the threshold as abnormal points and recording the position, amplitude and duration of each abnormal point, it includes: setting a list as a data structure to store abnormal point information, calculating the abnormal point information, wherein the stored abnormal point information includes position, amplitude and duration; when an abnormal point is detected, the position, amplitude and duration are recorded, wherein the position is the current timestamp of the abnormal point, and the amplitude is the degree of exceeding the threshold, which is obtained by the difference between the current data point and the current threshold; the duration is the duration of the abnormal state, if the subsequent data points are still in the abnormal state, the status of multiple abnormal points is continuously recorded and the duration is updated.

For example, suppose there is a set of historical data: [10, 12, 11, 14, 15, 20, 30, 25, 12];

Calculate the mean and standard deviation:

Mean: μ = 17;

Standard deviation: σ ≈ 6.4 (take k=2, threshold=17+2*6.4 = 29.8);

Real-time data inflow: [28, 31, 27, 26, 32, 29];

Record abnormal points:

Data 31: exceeded threshold, position = 1, amplitude = 31-29.8 = 1.2, duration = 1;

Data 32: Exceeding threshold, position = 4, amplitude = 32-29.8 = 2.2, duration = 1 (continuously updated).

Through the above detailed steps, a data structure is provided for storing abnormal point information, including the location, amplitude and duration of the abnormal point, and the duration can be updated after the abnormal state is resolved. The k value can be flexibly adjusted according to the needs of different application scenarios to adapt to different abnormality detection sensitivities.

For example, assume that there are the following historical data records (the unit is the reading of a certain sensor):

10,12,11,14,15,20,30,25,12;

Compute the initial mean and standard deviation:

(1) Mean: μ = (10 + 12 + 11 + 14 + 15 + 20 + 30 + 25 + 12) / 9 ≈ 17;

Standard Deviation: ;

Where n=9, 10, 12, 11, 14, 15, 20, 30, 25, 12;

Initial threshold (set k=2): threshold=17+2×6.4=29.8;

Real-time data inflow (new data): 28, 31, 27, 26, 32, 29;

Monitoring process:

Data 28: does not exceed the threshold;

Data 31: exceeds the threshold, records the abnormal point, position: 1, amplitude: 1.2, duration: 1;

Data 27: does not exceed the threshold;

Data 26: does not exceed the threshold;

Data 32: exceeds the threshold, records abnormal points, position: 4, amplitude: 2.2, duration: 1;

Data 29: does not exceed the threshold;

Update the threshold (set the new sliding window to the last 5 data points):

The new data windows are: 31, 27, 26, 32, 29;

Compute the new mean and standard deviation: ; ;

New threshold = 29 + 2 × 2.5 = 34.

Through the above steps, it is possible to dynamically monitor the data stream and adjust the warning threshold in real time, effectively identify and mark abnormal points. This method not only improves the sensitivity of monitoring, but also adapts to environmental changes and ensures timely response to potential abnormal situations. The above claims provide a framework for possible subsequent technical patents, ensuring the uniqueness and application value of this method.

Step S30, extracting time series features, frequency domain features and statistical features with the abnormal point as the starting point from the historical operation data of the wind turbine unit; and calculating the mutation amplitude and duration of each abnormal point, using fast Fourier transform to perform frequency analysis on the data around the abnormal point, and extracting the frequency component.

In this step, time series features, frequency domain features and statistical features are extracted with the abnormal point as the starting point. When extracting time series features, vibration data within a preset time window (for example, 10 minutes before and after) is extracted with the abnormal point as the starting point, and the mean, variance and root mean square are calculated based on the time series features of the vibration data. When extracting frequency domain features, fast Fourier transform is performed on the extracted time series data to obtain the frequency domain data and spectrum density of the frequency domain features. When extracting statistical features, statistical features are calculated based on the extracted time series and frequency domain features.

When calculating the mutation amplitude and duration of each abnormal point, according to the abnormal point found, a number of data points before and after the abnormal point are determined by selecting a window, and the mean and mutation amplitude are calculated. The duration is calculated for the time period when the abnormal point exceeds a certain threshold in the data, wherein, according to the set threshold of the mutation amplitude, the data points are checked forward and backward from the abnormal point until the data point recovers below the threshold, and the time difference from the time when the abnormal point appears to the time when the data recovers below the threshold is calculated;

When performing frequency analysis, a number of data points before and after the abnormal point are determined by selecting a window, and the frequency resolution and frequency components are calculated using fast Fourier transform to obtain the frequency components;

The formula for calculating frequency resolution is: ;

Among them, the formula for calculating the frequency component is: ;

In the formula, is the sampling frequency, is the number of FFT points, ;

The frequency components are expressed as: ;

in, is the window data, is the frequency component.

Exemplarily, assume that the present application has the following vibration data of a wind turbine unit (unit: mm/s), and the present application detects an abnormal point around timestamp T0;

;

Then, in the time series feature, the mean is 0.585 and the variance is 0.065. In the frequency domain feature, after FFT calculation, the frequency domain data (amplitude spectrum) obtained by this application is:

;

;

←main frequency;

The PSD of the main frequency is calculated as:

;

The main frequency calculated using FFT is 1 Hz, and the peak value of the spectrum density is 0.25;

The sorting data is: 0.4, 0.5, 0.5, 0.5, 0.5, 0.6, 0.6, 0.6, 0.7, 1.5;

The statistical characteristic is about 1.5.

Through the above steps, relevant features can be extracted from the historical operation data of the wind turbine unit, including time series features, frequency domain features and statistical features, which can be used for subsequent tasks such as anomaly detection, fault diagnosis and predictive maintenance.

To calculate the mutation amplitude and duration of each outlier point, and use fast Fourier transform (FFT) to perform frequency analysis on the data around the outlier point, for example, the selected window data is [6,30,2,3], and FFT analysis is performed:

(1) Window data: x[n]=[6,30,2,3];

(2) Apply FFT: Calculate FFT and obtain the frequency domain representation X(f).

(3) Frequency resolution:

Assuming the sampling frequency fs=1Hz and the window size N=4, then:

Δf=fs/N=1/4=0.25Hz;

(4) Frequency component:

The calculated frequency components are:

f0=0,f1=0.25,f2=0.5,f3=0.75Hz;

(5) Result analysis: By analyzing the frequency components, potential periodic behaviors near the anomaly points can be identified.

Through the above steps, we can systematically analyze the abnormal points in the data, calculate their mutation amplitude and duration, and extract the relevant frequency components through FFT. This method is very effective for anomaly detection and feature analysis of time series data, and is applicable to many fields such as financial monitoring, environmental monitoring, medical data analysis, etc.

Step S40: Use the extracted features as input to train a fault prediction model, learn the relationship between the features and the operating status of the wind turbine, use the trained model to predict and evaluate the operating status of the wind turbine unit, and identify the prediction results.

In this step, as shown in FIG3 , the extracted features are used as input to train the fault prediction model, and the relationship between the features and the wind turbine operating status is learned, including the following steps:

Step S401: marking the operation data of the wind turbine as a normal state and a fault state according to the collected operation data of the wind turbine and the extracted time series features, frequency domain features and statistical features with the abnormal point as the starting point;

Step S402: After data cleaning and standardization of the operation data, the data set of the operation data is divided into a training set, a validation set and a test set (70% training set, 15% validation set, and 15% test set);

Step S403: Use the training set data to train the fault prediction model, select the loss function, use the validation set to tune the hyperparameters, use the test set to evaluate the fault prediction model, use the trained model to monitor and predict the new wind turbine operation data in real time, judge the status of the wind turbine according to the output of the fault prediction model, and identify the prediction result.

Through the above steps, a fan fault prediction model can be systematically constructed, and the relationship between the fan operating status and faults can be learned by using the extracted features, so as to realize real-time monitoring and fault warning in practical applications. This method can significantly improve the operating efficiency and safety of the fan and reduce maintenance costs.

Step S50: Generate a warning signal based on the key characteristic parameters of the mutation amplitude, duration and frequency component of the abnormal point and the prediction result output by the fault prediction model.

In this step, when generating a warning signal, the fault probability P is predicted based on the identified prediction results, and a threshold judgment is performed to generate a warning signal based on the predicted probability; wherein the fault probability P is:

;

in, is the trained fault prediction model. is the mutation amplitude, is the duration, is the frequency component; when judging the threshold: ;in, The preset warning threshold.

It is assumed that the sensor data collected by this application on a certain device is as follows:

The current value is 75, and the previous value is 50; the mutation amplitude is:

ΔY=|75−50|=25;

The duration is: from the start time t=0 to the end time t=5, then the duration D=5−0=5;

The frequency component obtained by FFT is the main frequency f=10Hz;

The predicted probability P(f)=0.8 is obtained by training the model;

When generating an early warning signal, if the early warning threshold T=0.7 is set, then =1, triggering an early warning.

Through the above steps, a fault warning system based on abnormal points can be built, which can effectively detect equipment failures and issue warning signals in time. This method combines the advantages of data analysis, feature extraction and machine learning models, and has high accuracy and real-time performance. In specific implementation, it is also necessary to adjust the model parameters and algorithm selection according to different devices and application scenarios to improve the adaptability and effect of the system.

It should be understood that, although described in a certain order, these steps are not necessarily performed in sequence in the above order. Unless there is clear explanation in this article, the execution of these steps does not have strict order restriction, and these steps can be performed in other orders. Moreover, a part of the steps of the present embodiment may include a plurality of steps or a plurality of stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a part of the steps or stages in other steps or other steps.

In one embodiment, the present invention provides a system for correcting a warning threshold value of an operating state of a wind turbine generator set, which is used to execute the method for correcting the warning threshold value of an operating state of a wind turbine generator set, and the system comprises:

A data acquisition module is used to use sensors to collect the operating data of the fan unit in real time to form time series data;

The anomaly monitoring module is used to detect mutations in the collected time series data, calculate the cumulative sum series, mark the points exceeding the threshold as anomalies, and record the location, amplitude and duration of each anomaly point;

Threshold adjustment module, used to dynamically adjust the warning threshold based on historical data and known status;

The feature extraction module is used to extract the time series features, frequency domain features and statistical features starting from the abnormal point from the historical operation data of the wind turbine unit; calculate the mutation amplitude and duration of each abnormal point, use fast Fourier transform to perform frequency analysis on the data around the abnormal point, and extract the frequency component;

A model training module is used to take the extracted features as input, train the fault prediction model, learn the relationship between the features and the operating status of the fan, use the trained model to predict and evaluate the operating status of the fan unit, and identify the prediction results;

The early warning module is used to generate early warning signals based on the key characteristic parameters of the mutation amplitude, duration and frequency component of the abnormal point and the prediction results output by the fault prediction model.

The wind turbine operating status warning threshold correction system of the present invention combines the advantages of data analysis, feature extraction and machine learning models, and has high accuracy and real-time performance. In specific implementation, it is also necessary to adjust the model parameters and algorithm selection according to different devices and application scenarios to improve the adaptability and effect of the system.

In this embodiment, the wind turbine operating status warning threshold correction system adopts the steps of a wind turbine operating status warning threshold correction method as described above during execution. Therefore, the operation process of the wind turbine operating status warning threshold correction system in this embodiment will not be introduced in detail.

The method and system for correcting the warning threshold of the operating status of a wind turbine unit of the present invention collects the operating data of the wind turbine unit in real time through sensors, ensuring that the abnormality of the operating status can be quickly discovered and responded to, thereby improving the safety and reliability of the system; dynamically adjusting the warning threshold based on historical data and known status avoids false alarms and missed alarms that may be caused by static thresholds, making the warning mechanism more flexible and accurate; using mutation detection and accumulation and sequence analysis to accurately mark abnormal points, the location, amplitude and duration of the abnormality can be clearly recorded, providing important data support for subsequent analysis. Extracting time series features, frequency domain features and statistical features from the abnormal points provides a rich information basis, which helps to fully understand the characteristics of abnormal behavior and its influencing factors; performing frequency analysis on the data around the abnormal points through fast Fourier transform can identify potential periodic problems and improve the depth and breadth of fault prediction; training the fault prediction model and learning the relationship between the features and the operating status of the wind turbine can predict potential fault risks at an early stage, thereby taking measures in advance to reduce downtime and maintenance costs. By combining the key characteristic parameters of mutation amplitude, duration, and frequency components with the output of the prediction model, accurate early warning signals are generated to help managers make timely decisions; by identifying and predicting faults in advance, more scientific maintenance plans can be formulated to reduce the high maintenance costs caused by sudden failures and improve the operating efficiency of the fan; timely early warnings and fault predictions can effectively prevent equipment damage caused by overload or other abnormal conditions, thereby extending the service life of the fan unit.

In summary, the method and system for correcting the wind turbine operating status warning threshold of the present invention greatly improve the operating safety and maintenance efficiency of the wind turbine unit through real-time monitoring, dynamic adjustment, precise identification and intelligent prediction, and have significant economic and social benefits.

In one embodiment, a computer device is also provided in an embodiment of the present invention, comprising at least one processor and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor executes the steps of the method for correcting the wind turbine operating status warning threshold.

In one embodiment, the present invention further provides a computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the steps of the method for correcting the wind turbine operating status warning threshold.

A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be realized by instructing the relevant hardware through a computer program represented by computer instructions, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory.

Non-volatile memory may include read-only memory, magnetic tape, floppy disk, flash memory or optical storage, etc. Volatile memory may include random access memory or external cache memory. As an illustration and not limitation, RAM may be in various forms, such as static random access memory or dynamic random access memory, etc.

The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims (10)

1. A correction method of a wind turbine running state early warning threshold value is characterized by comprising the following steps:

Collecting operation data of the fan unit in real time by using a sensor to form time sequence data;

Performing mutation detection on the collected time sequence data, calculating accumulation sum sequence, dynamically adjusting an early warning threshold value based on historical data and known states, marking points exceeding the threshold value as abnormal points, and recording the position, amplitude and duration of each abnormal point;

Extracting time sequence features, frequency domain features and statistical features taking abnormal points as starting points from historical operation data of the fan unit; calculating the mutation amplitude and duration of each abnormal point, and carrying out frequency analysis on data around the abnormal point by using fast Fourier transform to extract frequency components;

Taking the extracted features as input, training a fault prediction model, learning the relation between the features and the running state of the fan, and predicting and evaluating the running state of the fan unit by using the trained model to identify a prediction result;

and generating an early warning signal based on key characteristic parameters of mutation amplitude, duration and frequency components of the abnormal point and a prediction result output by the fault prediction model.

2. The method for correcting the early warning threshold value of the running state of the wind turbine generator set according to claim 1, wherein the method for correcting the early warning threshold value of the running state of the wind turbine generator set is characterized by comprising the following steps when mutation detection is carried out on collected time series data and accumulation and sequence are calculated:

acquiring collected time sequence data, and calculating the mean value and standard deviation of the time sequence data based on accumulation;

selecting a preset mutation threshold value, and calculating an accumulation sum sequence by using an accumulation sum formula;

Judging whether the cumulative sum exceeds a mutation threshold value to determine whether an abnormal point exists; wherein:

time series data Wherein, the method comprises the steps of, wherein,The number of data points that are operational data;

the preset mutation threshold value is ; The cumulative sum is calculated as:

；

In the method, in the process of the invention, ，Is an adjustment factor, take；Representing a mean value of the time series data; representing the cumulative sum of the current time i; Representing the cumulative sum of the previous time instant;

When (when) When mutation occurs, record abnormal pointAnd records the index and corresponding value of the outlier.

3. The method for correcting the early warning threshold value of the running state of the wind turbine generator set according to claim 2, wherein when the early warning threshold value is dynamically adjusted, collected historical data comprise data in a normal state and an abnormal state, statistical characteristics of a mean value and a standard deviation of the historical data are calculated, an initial threshold value is set to be the mean value plus k times the standard deviation, the mean value and the standard deviation are recalculated according to the change of a real-time data stream in a periodic manner, and the threshold value is adjusted; when calculating a new threshold value, calculating the mean value and standard deviation of the nearest N data points by using a sliding window, and updating the threshold value; wherein k is a constant in k times of standard deviation, and the value is 2 or 3.

4. The method for correcting the early warning threshold value of the running state of the wind turbine according to claim 3, wherein the marking the points exceeding the threshold value as abnormal points and recording the position, the amplitude and the duration of each abnormal point comprises the following steps:

The method comprises the steps of setting a list, storing abnormal point information as a data structure, and calculating the abnormal point information, wherein the stored abnormal point information comprises positions, amplitudes and duration; recording a position, an amplitude and a duration when the abnormal point is detected, wherein the position is a current timestamp of the abnormal point, the amplitude is the degree exceeding a threshold value, and the amplitude is obtained by the difference between the current data point and the current threshold value; the duration is the duration of the abnormal state, and if the subsequent data point is still in the abnormal state, the states of a plurality of abnormal points are continuously recorded and the duration is updated.

5. The method for correcting the early warning threshold of the running state of the wind turbine according to claim 4, wherein the time series features, the frequency domain features and the statistical features taking the abnormal point as a starting point are extracted, and wherein:

When the time sequence features are extracted, vibration data in a preset time window are extracted by taking an abnormal point as a starting point, and the mean value, the variance and the root mean square are calculated based on the time sequence features of the vibration data;

when the frequency domain features are extracted, performing fast Fourier transform on the extracted time sequence data to obtain frequency domain data and spectrum density of the frequency domain features;

And when the statistical features are extracted, calculating to obtain the statistical features based on the extracted time sequence and frequency domain features.

6. The method for correcting the early warning threshold value of the running state of the wind turbine generator according to claim 5, wherein when the mutation amplitude and the duration of each abnormal point are calculated, a plurality of data points before and after the abnormal point are determined through a selection window according to the found abnormal point, the average value and the mutation amplitude are calculated, the duration of the abnormal point is calculated in a time period when the abnormal point exceeds a certain threshold value in the data, the data points are checked forwards and backwards according to the threshold value of the set mutation amplitude, and the time difference from the occurrence time of the abnormal point to the time when the data is restored below the threshold value is calculated.

7. The method for correcting the early warning threshold value of the running state of the wind turbine generator set according to claim 6, wherein when frequency analysis is carried out, a plurality of data points before and after an abnormal point are determined through a selection window, and frequency resolution and frequency components are calculated through fast Fourier transform to obtain frequency components;

Wherein, the formula for calculating the frequency resolution is: ；

Wherein, the formula for calculating the frequency component is: ；

In the method, in the process of the invention, Is the sampling frequency at which the sample is to be taken,Is the number of FFT points,；

The frequency components are expressed as:；

Wherein, Is the window data of the window and,Is a frequency component.

8. The method for correcting the early warning threshold value of the running state of the wind turbine according to claim 7, wherein the extracted features are used as input to train a fault prediction model, learn the relation between the features and the running state of the wind turbine, and the method comprises the following steps:

marking the operation data of the fan as a normal state and a fault state according to the collected operation data of the fan and the extracted time sequence features, frequency domain features and statistical features taking the abnormal point as a starting point;

After data cleaning and standardization processing are carried out on the operation data, dividing a data set of the operation data into a training set, a verification set and a test set;

Training a failure prediction model by using training set data, selecting a loss function, performing super-parameter tuning by using a verification set, evaluating the failure prediction model by using a test set, performing real-time monitoring and prediction on new fan operation data by using the trained model, judging the state of a fan according to the output of the failure prediction model, and identifying a prediction result.

9. The method for correcting the early warning threshold value of the running state of the wind turbine generator set according to claim 8 is characterized in that when an early warning signal is generated, the fault probability P is predicted according to the recognized prediction result, the threshold value is judged, and the early warning signal is generated according to the prediction probability; wherein, the fault probability P is:

；

Wherein, In order to train a good model of the fault prediction,In order for the magnitude of the abrupt change to be,For the duration of time it is possible,Is a frequency component; threshold judgment: ; wherein, And the early warning threshold value is preset.

10. A system for correcting an early warning threshold of an operation state of a wind turbine, characterized by performing the steps of the method for correcting an early warning threshold of an operation state of a wind turbine according to any one of claims 1 to 9, the system for correcting an early warning threshold of an operation state of a wind turbine comprising:

The data acquisition module is used for acquiring the operation data of the fan unit in real time by using the sensor to form time sequence data;

the anomaly monitoring module is used for carrying out mutation detection on the collected time sequence data, calculating a cumulative sum sequence, marking points exceeding a threshold value as abnormal points, and recording the position, the amplitude and the duration of each abnormal point;

The feature extraction module is used for extracting time sequence features, frequency domain features and statistical features taking abnormal points as starting points from historical operation data of the fan unit; calculating the mutation amplitude and duration of each abnormal point, and carrying out frequency analysis on data around the abnormal point by using fast Fourier transform to extract frequency components;

The model training module is used for taking the extracted characteristics as input, training a fault prediction model, learning the relation between the characteristics and the running state of the fan, predicting and evaluating the running state of the fan unit by using the trained model, and identifying a prediction result;

The early warning module is used for generating an early warning signal based on key characteristic parameters of mutation amplitude, duration and frequency components of the abnormal point and a prediction result output by the fault prediction model.