A fan yaw error identification method and system
By analyzing the scatter density distribution characteristics and time-series evolution of wind speed and power, the problem of dynamic monitoring and early warning of yaw error identification of wind turbine units was solved, realizing efficient and accurate yaw error identification and operation and maintenance optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING BRANCH OF BEIJING JINGNENG CLEAN ENERGY POWER CO LTD
- Filing Date
- 2026-02-26
- Publication Date
- 2026-06-16
Smart Images
Figure CN122216017A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the technical field of wind power generation, and more specifically, to a method for identifying yaw error of wind turbine generators based on the time-series evolution of wind speed-power scatter point density. Background Technology
[0002] In the wind power industry, the power generation efficiency of wind turbines directly determines the economic benefits of wind power projects. The yaw system, as a core component of wind turbines, functions to precisely align the nacelle with the wind direction, ensuring the rotor captures maximum wind energy. Yaw error is defined as the angle between the actual orientation of the nacelle and the true wind direction. Related research data shows that for every 1-degree increase in yaw error, the wind turbine's power generation will decrease by approximately 0.15%. Therefore, timely and accurate identification of yaw error and implementation of corrective measures are crucial for ensuring the efficient and stable operation of wind turbines.
[0003] In existing technologies, methods for identifying yaw error in wind turbines are mainly divided into three categories: First, the method based on direct measurement using wind vanes. This method collects wind direction data using wind vanes and directly calculates the yaw error. However, the installation position of the wind vanes is easily affected by environmental factors such as wake and turbulence, limiting measurement accuracy and failing to reflect the true yaw error of the wind turbine. Second, the method based on power curve comparison. This method compares the actual power curve of the wind turbine with a standard power curve and identifies the yaw error based on the curve differences. However, this method cannot effectively distinguish yaw error from other factors that cause performance degradation, such as blade wear, transmission system failure, and pitch system abnormalities, and is prone to misjudgment. Third, the method based on static analysis of wind speed-power scatter plots. This method judges the yaw state of the turbine by analyzing the distribution characteristics of wind speed-power scatter plots at a certain moment. However, static analysis can only reflect the instantaneous scatter plot distribution and cannot capture the dynamic evolution of yaw error. Lacking time-series information support, it is difficult to achieve early warning and trend prediction of yaw error.
[0004] In summary, existing yaw error identification methods generally suffer from the following technical defects: static analysis modes cannot capture the dynamic changes in yaw error and are difficult to reflect the generation, development, and evolution of the error; abnormal data and measurement noise can easily interfere with the identification results, reducing the accuracy and robustness of the identification; there is a lack of analysis of the temporal evolution information of yaw error, resulting in low identification sensitivity and an inability to achieve early warning and trend prediction; it is difficult to distinguish between persistent yaw error and temporary wind direction deviation caused by wind direction fluctuations, which can easily lead to false alarms and increase unnecessary operation and maintenance costs. Summary of the Invention
[0005] To address the aforementioned technical problems in related technologies, this invention provides a method and system for identifying wind turbine yaw error, which can solve the above problems.
[0006] To achieve the above-mentioned technical objectives, the technical solution of the present invention is implemented as follows: A method for identifying wind turbine yaw error includes the following steps: S1: Data Acquisition and Preprocessing: Collect SCADA operation data of wind turbine units, which includes wind speed, active power, wind direction angle and timestamp; preprocess the SCADA operation data to remove shutdown status data, fault status data, power-limited operation data and data with pitch angle greater than 2 degrees, and construct a time-series scatter dataset. S2: Time window division and scatter point density distribution calculation: The time series scatter point dataset is divided into multiple continuous time windows according to the set time window length and sliding step size, and the density distribution of wind speed-power scatter points in each time window is calculated by the kernel density estimation method. S3: Density distribution feature extraction: Extract density distribution features from the scatter density distribution of each time window. The density distribution features include density peak location, distribution concentration, and distribution skewness. S4: Temporal evolution feature analysis: Based on the density distribution features extracted from each time window, the temporal evolution features are calculated, including the density center offset trajectory, the rate of change of distribution morphology, and the evolution trend index; S5: Yaw error state identification: Based on the degree of deviation between the time-series evolution characteristics and the reference mode, identify the yaw error state of the wind turbine. S6: Output identification results: Output the yaw error status identification results of the wind turbine.
[0007] Furthermore, in step S2, the kernel density estimation method employs an adaptive bandwidth Gaussian kernel function; the time window length is 12 to 48 hours, and the sliding step size is 50% of the time window length.
[0008] Furthermore, in step S3, the density peak position is the position of the maximum power density within the wind speed range of 6~8m / s; the distribution concentration is represented by the full width at half maximum (FWHM) or the area of an equivalent ellipse; and the distribution skewness is the ratio of peak power to standard power.
[0009] Furthermore, in step S4, the density center offset trajectory is the movement trajectory of the density peak position over time; the distribution morphology change rate is the rate of change of distribution concentration over time; and the evolution trend index is the slope obtained by linear fitting using a sliding window.
[0010] Furthermore, in step S5, the yaw error state includes persistent yaw error, aggravated yaw error, and severe yaw error. The specific identification rules are as follows: when the density center offset trajectory shows a persistent unidirectional offset and the offset of multiple consecutive windows exceeds the first threshold, it is determined to be persistent yaw error; when the distribution morphology change rate exceeds the second threshold and the evolution trend index is positive, it is determined to be aggravated yaw error; when the distribution concentration decreases significantly and the peak power is lower than the third threshold, it is determined to be severe yaw error.
[0011] Furthermore, step S5 also includes the identification of temporary wind direction deviation: extracting the multi-peak features of the wind speed-power scatter density distribution, analyzing the ratio of the main peak value to the secondary peak value, and when a multi-peak distribution is detected and the ratio of the main peak value to the secondary peak value is close, it is determined to be a temporary wind direction deviation caused by wind direction fluctuation, rather than a continuous yaw error.
[0012] Furthermore, in step S6, the yaw error state identification result includes the yaw error state, error level, and targeted correction suggestions.
[0013] A wind turbine yaw error identification system, the system comprising a data acquisition layer, a data processing layer, a core processing layer, and an application layer; The data acquisition layer includes a wind direction sensor, a wind speed sensor, a power sensor, and a SCADA system, which are used to collect SCADA operation data of the wind turbine and transmit it to the data processing layer. The data processing layer includes a data preprocessing module, which is used to preprocess SCADA operation data and construct a time-series scatter dataset; The core processing layer includes a time window partitioning module, a density distribution calculation module, a feature extraction module, a temporal evolution analysis module, and a yaw error identification module. The time window partitioning module is used to partition continuous time windows, the density distribution calculation module is used to calculate scattered density distribution, the feature extraction module is used to extract density distribution features, the temporal evolution analysis module is used to calculate temporal evolution features, and the yaw error identification module is used to identify the yaw error state based on the temporal evolution features. The application layer includes a report generation module, an alarm system, and a monitoring interface, which are used to output yaw error identification results, abnormal alarms, and real-time display of the unit's yaw status.
[0014] Furthermore, the preprocessing function of the data preprocessing module includes invalid data removal, data quality check and data format conversion. The invalid data includes shutdown status data, fault status data, power-limited operation data and data with a pitch angle greater than 2 degrees.
[0015] The beneficial effects of this invention are: Achieving dynamic monitoring of yaw error throughout the entire process: This invention introduces time window division and time series evolution analysis methods to transform static wind speed-power scatter plots into dynamic time series evolution characteristics. It can accurately capture the generation, development and evolution of yaw error, overcome the shortcomings of traditional static analysis methods that cannot reflect the dynamic changes of yaw error, and achieve dynamic monitoring of yaw error throughout the entire process.
[0016] Significantly enhances the anti-interference capability of the identification results: The present invention uses an adaptive bandwidth Gaussian kernel density estimation method to calculate the scatter point density distribution. The adaptive bandwidth can be automatically adjusted according to the actual distribution characteristics of the scatter points, which can effectively smooth the influence of abnormal data, random noise and measurement errors, reduce their interference on the identification results, and improve the robustness and accuracy of yaw error identification.
[0017] Improving identification sensitivity and enabling early warning of yaw error: This invention analyzes subtle changes in the density distribution of wind speed-power scatter points, extracts characteristic parameters such as the location of density peaks and the concentration of distribution, and analyzes their temporal evolution. It can identify slight yaw errors that are difficult to detect by traditional methods, capture early signs of yaw errors, and provide maintenance personnel with sufficient time to take timely maintenance measures.
[0018] This invention enables graded early warning and trend prediction of yaw error: Based on time-series evolution characteristics, yaw error is classified into states such as persistent yaw error, aggravated yaw error, and severe yaw error, and corresponding error levels are assigned, realizing graded early warning from no yaw, slight yaw to severe yaw. At the same time, by calculating the evolution trend index, the development trend of yaw error is quantified, enabling trend prediction of yaw error and providing data support for the refined operation and maintenance of wind farms.
[0019] Accurately distinguish error types, avoid false alarms and reduce operation and maintenance costs: By analyzing the multi-peak characteristics of the wind speed-power scatter density distribution and the ratio of the primary and secondary peaks, this invention can effectively distinguish between persistent yaw errors and temporary wind direction deviations caused by wind direction fluctuations. This solves the problem of false alarms in existing methods, avoids unnecessary operation and maintenance operations caused by misjudgment, and significantly reduces the operation and maintenance costs of wind farms. Attached Figure Description
[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0021] Figure 1 This is a flowchart of a wind turbine yaw error identification method according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the calculation of wind speed-power scatter point density distribution according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the temporal evolution analysis of density distribution as described in an embodiment of the present invention; Figure 4 This is a schematic diagram of the yaw error identification result output according to an embodiment of the present invention; Figure 5 This is a structural block diagram of a wind turbine yaw error identification system according to an embodiment of the present invention. Detailed Implementation
[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the scope of protection of the present invention.
[0023] like Figures 1-5As shown, the present invention discloses a method for identifying wind turbine yaw error, comprising the following steps: S1: Data acquisition and preprocessing: acquiring SCADA operating data of the wind turbine, the SCADA operating data including wind speed, active power, wind direction angle and timestamp; preprocessing the SCADA operating data to remove shutdown status data, fault status data, power-limited operation data and data with pitch angle greater than 2 degrees, and constructing a time-series scatter dataset; S2: Time window division and scatter density distribution calculation: dividing the time-series scatter dataset into multiple continuous time windows according to a set time window length and sliding step size, and calculating the yaw error of each time window using the kernel density estimation method. S3: Density distribution of scatter points of wind speed-power within the airway; S4: Density distribution feature extraction: Extract density distribution features from the scatter point density distribution of each time window, the density distribution features including density peak position, distribution concentration and distribution skewness; S5: Temporal evolution feature analysis: Calculate temporal evolution features based on the density distribution features extracted from each time window, the temporal evolution features including density center offset trajectory, distribution shape change rate and evolution trend index; S6: Yaw error state identification: Identify the yaw error state of the wind turbine according to the degree of deviation between the temporal evolution features and the baseline model; S7: Output identification result: Output the yaw error state identification result of the wind turbine. A wind turbine yaw error identification system is also disclosed. The system includes a data acquisition layer, a data processing layer, a core processing layer, and an application layer. The data acquisition layer includes wind direction sensors, wind speed sensors, power sensors, condition monitoring sensors, and a SCADA system, used to collect SCADA operating data of the wind turbine and transmit it to the data processing layer. The data processing layer includes a data preprocessing module, used to preprocess the SCADA operating data and construct a time-series scatter dataset. The core processing layer includes a time window partitioning module, a density distribution calculation module, a feature extraction module, a time-series evolution analysis module, and a yaw error identification module. The time window partitioning module is used to partition continuous time windows; the density distribution calculation module is used to calculate the scatter density distribution; the feature extraction module is used to extract density distribution features; the time-series evolution analysis module is used to calculate time-series evolution features; and the yaw error identification module is used to identify the yaw error state based on the time-series evolution features. The application layer includes a report generation module, an alarm system, and a monitoring interface, used to output yaw error identification results, abnormal alarms, and real-time display of the turbine's yaw status.
[0024] In a specific embodiment of this application, taking a 1.5MW wind turbine as an example, the specific implementation process of the yaw error identification method of the present invention is described as follows: Data Acquisition and Preprocessing: SCADA operation data of the unit was continuously collected for 60 days, with a sampling period of 1 minute. The data range included wind speed of 3~15 m / s and active power of 0~1500 kW. The collected data was preprocessed to remove data from shutdown, fault, and power-limited operation states, as well as data with a pitch angle greater than 2 degrees, resulting in approximately 25,000 valid data points, which were used to construct a time-series scatter dataset.
[0025] Time window division and scatter point density distribution calculation: The time window length was set to 24 hours, and the sliding step size was 12 hours. The time series scatter point dataset was divided into 59 consecutive time windows. The adaptive bandwidth Gaussian kernel density estimation method was used to calculate the density distribution of wind speed-power scatter points in each time window, and the density distribution heatmap and three-dimensional surface of each window were obtained.
[0026] Density distribution feature extraction: From the density distribution results of each time window, extract the density peak location in the 6~8m / s wind speed range, the distribution concentration expressed as half-width at half-height, and the distribution skewness expressed as the ratio of peak power to standard power.
[0027] Temporal evolution feature analysis: Based on the extracted density distribution features, the movement trajectory of the density peak position over time is plotted, the daily rate of change of distribution concentration is calculated as the rate of change of distribution morphology, and the evolution trend index is calculated by linear fitting using a 7-day sliding window.
[0028] Yaw error status identification: Determination is based on the degree of deviation between the temporal evolution characteristics and the baseline model. From days 1 to 20, the density center position is stable, and the rate of change in distribution morphology is less than 5% / day, indicating no yaw error. From days 21 to 35, a persistent unidirectional shift of the density center occurs, with the shift exceeding 0.3 m / s for five consecutive windows, indicating a slight persistent yaw error. From days 36 to 50, the rate of change in distribution morphology exceeds 5% / day, and the evolution trend index is positive, indicating an aggravated yaw error. From days 51 to 60, the distribution concentration decreases significantly, and the peak power is below 90% of the standard power, indicating a severe yaw error (Level III).
[0029] Output identification results: Output a yaw error identification report for the 1.5MW wind turbine, clarify the yaw error status and error level at each stage, and propose correction suggestions for severe yaw errors, such as immediate inspection of the mechanical structure of the yaw system, correction of the wind direction sensor, and zeroing of the yaw angle.
[0030] In a specific embodiment of this application, taking a 2.0MW wind turbine as an example, the ability of the method of the present invention to identify slight yaw errors is verified as follows: Data acquisition and preprocessing: SCADA operation data of the unit for 30 days were collected, with a sampling period of 1 minute. After preprocessing to remove invalid data, approximately 18,000 valid data points were obtained, and a time-series scatter dataset was constructed.
[0031] Time window division and scatter density distribution calculation: To meet the monitoring requirements of higher time resolution, the time window length is set to 12 hours and the sliding step is 6 hours. The time series scatter dataset is divided into 59 continuous time windows, and the scatter density distribution of each window is calculated using the adaptive bandwidth Gaussian kernel density estimation method.
[0032] Feature extraction and temporal evolution analysis: Extract the density distribution feature parameters of each window, calculate the temporal evolution features and analyze their variation patterns.
[0033] Yaw error status identification: This embodiment successfully identified the slight yaw error that occurred on the 15th day of the unit. At this time, the yaw angle of the unit was about 3 degrees. This slight yaw error could not be detected by the traditional power curve comparison method. However, the method of this invention accurately captured the early signs of yaw error by analyzing the small shift in the position of the density peak and the subtle changes in the distribution concentration.
[0034] Output identification results: Output identification report, which is determined to be a slight yaw error (Level I), and proposes suggestions for inspecting the yaw system and calibrating the wind direction sensor, thus realizing early warning and timely intervention of yaw error.
[0035] In a specific embodiment of this application, taking a 3.0MW wind turbine as an example, the ability of the method of the present invention to identify complex yaw states and to distinguish between persistent yaw errors and temporary wind direction deviations are verified as follows: Data Acquisition and Preprocessing: SCADA operation data of the unit for 45 days was collected. The unit's operating environment has obvious wind direction fluctuations and is prone to temporary wind direction deviations. After preprocessing, a time-series scatter dataset was constructed.
[0036] Time window division and density distribution calculation: The time window length is set to 48 hours and the sliding step size is 24 hours. The scatter density distribution of each window is calculated by the adaptive bandwidth Gaussian kernel density estimation method.
[0037] Feature extraction and evolution analysis: In addition to extracting the density peak position, distribution concentration, and distribution skewness, multi-peak distribution feature extraction is added to analyze the ratio of the main peak and the secondary peak of the density distribution in each window.
[0038] Yaw error status identification: From day 1 to day 10, the density distribution of the aircraft is not significantly shifted, and it is determined that there is no yaw error. From day 11 to day 18, a multi-peak feature is detected in the density distribution, and the proportions of the primary and secondary peaks are close. It is determined that the wind direction deviation is caused by wind direction fluctuations and is not a continuous yaw error. No correction measures are required. From day 19 to day 45, a continuous unidirectional shift of the density center occurs, and there is no obvious multi-peak feature. It is determined to be a continuous yaw error. Moreover, the rate of change of the distribution pattern continues to increase over time, and it is determined that the yaw error is aggravated.
[0039] Output identification results: Output an identification report that clearly distinguishes between temporary wind direction deviation and persistent yaw error. For situations where persistent yaw error worsens, it proposes correction suggestions such as checking the yaw drive device and adjusting the yaw control strategy, thus avoiding unnecessary maintenance operations caused by misjudgment and reducing maintenance costs.
[0040] 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, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for identifying yaw error in wind turbines, characterized in that, Includes the following steps: S1: Data Acquisition and Preprocessing: Acquire SCADA operating data of wind turbine units, which includes wind speed, active power, wind direction angle and timestamp; The SCADA operation data is preprocessed to remove shutdown status data, fault status data, power-limited operation data, and data with a pitch angle greater than 2 degrees, and a time-series scatter dataset is constructed. S2: Time window division and scatter point density distribution calculation: The time series scatter point dataset is divided into multiple continuous time windows according to the set time window length and sliding step size, and the density distribution of wind speed-power scatter points in each time window is calculated by the kernel density estimation method. S3: Density distribution feature extraction: Extract density distribution features from the scatter density distribution of each time window. The density distribution features include density peak location, distribution concentration, and distribution skewness. S4: Temporal evolution feature analysis: Based on the density distribution features extracted from each time window, the temporal evolution features are calculated, including the density center offset trajectory, the rate of change of distribution morphology, and the evolution trend index; S5: Yaw error state identification: Based on the degree of deviation between the time-series evolution characteristics and the reference mode, identify the yaw error state of the wind turbine. S6: Output identification results: Output the yaw error status identification results of the wind turbine.
2. The wind turbine yaw error identification method according to claim 1, characterized in that, In step S2, the kernel density estimation method uses an adaptive bandwidth Gaussian kernel function; the time window length is 12 to 48 hours, and the sliding step size is 50% of the time window length.
3. The wind turbine yaw error identification method according to claim 1, characterized in that, In step S3, the density peak position is the position of the maximum power density within the wind speed range of 6~8m / s; the distribution concentration is represented by the full width at half maximum (FWHM) or the area of an equivalent ellipse; and the distribution skewness is the ratio of peak power to standard power.
4. The wind turbine yaw error identification method according to claim 1, characterized in that, In step S4, the density center offset trajectory is the movement trajectory of the density peak position over time; the distribution morphology change rate is the rate of change of distribution concentration over time; and the evolution trend index is the slope obtained by linear fitting using a sliding window.
5. The wind turbine yaw error identification method according to claim 1, characterized in that, In step S5, the yaw error state includes persistent yaw error, aggravated yaw error, and severe yaw error. The specific identification rules are as follows: when the density center offset trajectory shows a persistent unidirectional offset and the offset of multiple consecutive windows exceeds the first threshold, it is determined to be persistent yaw error; when the distribution morphology change rate exceeds the second threshold and the evolution trend index is positive, it is determined to be aggravated yaw error; when the distribution concentration decreases significantly and the peak power is lower than the third threshold, it is determined to be severe yaw error.
6. The wind turbine yaw error identification method according to claim 5, characterized in that, Step S5 also includes the identification of temporary wind direction deviation: extract the multi-peak features of the wind speed-power scatter density distribution, analyze the ratio of the main peak value to the secondary peak value, and when a multi-peak distribution is detected and the ratio of the main peak value to the secondary peak value is close, it is determined to be a temporary wind direction deviation caused by wind direction fluctuation, rather than a continuous yaw error.
7. The wind turbine yaw error identification method according to claim 1, characterized in that, In step S6, the yaw error status identification result includes the yaw error status, error level, and targeted correction suggestions.
8. A wind turbine yaw error identification system, characterized in that, To implement the yaw error identification method as described in any one of claims 1 to 7, the system includes a data acquisition layer, a data processing layer, a core processing layer, and an application layer; The data acquisition layer includes a wind direction sensor, a wind speed sensor, a power sensor, and a SCADA system, which are used to collect SCADA operation data of the wind turbine and transmit it to the data processing layer. The data processing layer includes a data preprocessing module, which is used to preprocess SCADA operation data and construct a time-series scatter dataset; The core processing layer includes a time window partitioning module, a density distribution calculation module, a feature extraction module, a temporal evolution analysis module, and a yaw error identification module. The time window partitioning module is used to partition continuous time windows, the density distribution calculation module is used to calculate scattered density distribution, the feature extraction module is used to extract density distribution features, the temporal evolution analysis module is used to calculate temporal evolution features, and the yaw error identification module is used to identify the yaw error state based on the temporal evolution features. The application layer includes a report generation module, an alarm system, and a monitoring interface, which are used to output yaw error identification results, abnormal alarms, and real-time display of the unit's yaw status.
9. A wind turbine yaw error identification system according to claim 8, characterized in that, The preprocessing functions of the data preprocessing module include invalid data removal, data quality checking, and data format conversion. The invalid data includes shutdown status data, fault status data, power-limited operation data, and data with a pitch angle greater than 2 degrees.