阅读说明：本技术 一种风电场风速传感器故障诊断方法、装置 (Fault diagnosis method and device for wind speed sensor of wind power plant ) 是由 沈小军 于 2021-08-26 设计创作，主要内容包括：本发明涉及一种风电场风速传感器故障诊断方法、装置,该方法包括：对风电场机组进行故障互诊群落划分；在故障互诊群落中分别对机组对应的风速传感器进行故障诊断,包括：获取故障互诊群落中风速传感器的风速时序流,确定群落内存在风速异常数据的时间区间,定位存在风速异常数据的时间区间内发生故障的风速传感器。与现有技术相比,本发明具有故障诊断精确度高、误检和漏检率低等优点。(The invention relates to a method and a device for diagnosing faults of a wind speed sensor of a wind power plant, wherein the method comprises the following steps: carrying out fault mutual diagnosis community division on the wind power plant set; respectively carrying out fault diagnosis on the wind speed sensors corresponding to the unit in the fault mutual diagnosis community, and the fault diagnosis method comprises the following steps: acquiring a wind speed time sequence flow of a wind speed sensor in a fault mutual diagnosis community, determining a time interval in which wind speed abnormal data exist in the community, and positioning the wind speed sensor which has a fault in the time interval in which the wind speed abnormal data exist. Compared with the prior art, the invention has the advantages of high fault diagnosis accuracy, low false detection and missed detection rate and the like.)

1. A fault diagnosis method for a wind speed sensor of a wind power plant is characterized by comprising the following steps:

carrying out fault mutual diagnosis community division on the wind power plant set;

respectively carrying out fault diagnosis on the wind speed sensors corresponding to the unit in the fault mutual diagnosis community, and the fault diagnosis method comprises the following steps:

acquiring the wind speed time sequence flow of a wind speed sensor in the fault mutual diagnosis community,

determining a time interval in which wind speed abnormal data exist in the fault inter-diagnosis community,

and positioning the wind speed sensor which has faults in the time interval with the wind speed abnormal data.

2. The wind power plant wind speed sensor fault diagnosis method according to claim 1, characterized in that the manner of dividing fault inter-diagnosis communities comprises:

and screening the spatial correlation of the wind speed of the units on the basis of the yaw data of the units, and screening a plurality of units with higher spatial correlation of the wind speed as associated units and dividing the associated units into a fault mutual diagnosis community.

3. The wind power plant wind speed sensor fault diagnosis method according to claim 1, wherein the specific manner of determining the time interval in which wind speed abnormal data exist in the fault mutual diagnosis community comprises the following steps:

acquiring index parameters describing the separation degree of wind speed data measured by each wind speed sensor in the fault mutual diagnosis community at each sampling moment;

determining a variable point index parameter threshold value by taking the index parameter as a variable point index for determining abnormity;

and determining the time interval with the index parameter larger than the variable point index parameter threshold value as the time interval with the wind speed abnormal data.

4. A wind farm wind speed sensor fault diagnosis method according to claim 3, characterized in that the index parameters comprise the variance of the wind speed data measured by each wind speed sensor in the fault inter-diagnosis community at each time point.

5. A wind farm wind speed sensor fault diagnosis method according to claim 3, characterized in that the obtaining manner of the variable point index parameter threshold value comprises:

the index parameters at each sampling moment are arranged according to an ascending order, the arranged index parameters are averagely divided into a plurality of parts according to a quantile algorithm, and the index parameters corresponding to the set quantile points are selected as variable point index parameter thresholds.

6. The wind farm wind speed sensor fault diagnosis method according to claim 1, characterized in that the manner of locating a faulty wind speed sensor comprises:

and carrying out wind speed correlation cross comparison on each wind speed sensor in the fault mutual diagnosis community, and determining the unit sensor causing the reduction of the wind speed correlation as a fault wind speed sensor.

7. A wind farm wind speed sensor fault diagnosis method according to claim 6, characterized in that the specific way of locating a faulty wind speed sensor is: and rolling to calculate the correlation of the wind speeds measured by any two wind speed sensors, selecting unit data with the correlation coefficient smaller than a set threshold value, counting the occurrence frequency of the associated wind speed sensors, and determining the wind speed sensor with the maximum occurrence frequency as a fault wind speed sensor.

8. A wind farm wind speed sensor fault diagnosis method according to claim 6 or 7, characterized in that the wind speed correlation is described by Pearson coefficient quantization, expressed as:

wherein r is_iTo quantify the correlation, X, of the wind speed data measured by two wind speed sensors in time series i_i,t、Y_i,tCorresponding to the wind speed data X of two wind speed sensors at the sampling time t in the quantitative time sequence i_i,avg、Y_i,avgCorresponding to the average value of the wind speed data of two wind speed sensors in a quantitative time sequence i, and n is a quantityThe total number of sampling instants within the time series i is quantified.

9. The wind farm wind speed sensor fault diagnosis method according to claim 1, characterized in that during fault diagnosis, the fault co-diagnosis community is re-classified when the wind direction changes over a threshold value.

10. A wind farm wind speed sensor fault diagnosis device, characterized by comprising a memory for storing a computer program and a processor for implementing the wind farm wind speed sensor fault diagnosis method according to any one of claims 1 to 9 when executing the computer program.

Technical Field

The invention relates to the technical field of wind power generation, in particular to a method and a device for diagnosing faults of a wind speed sensor of a wind power plant.

Background

Under the drive of a double-carbon target, renewable energy represented by wind energy gradually becomes an important means for developing a novel energy industry and reducing carbon emission in China due to the advantages of cleanness, no pollution, abundant reserves, mature development technology, low cost and the like. Wind speed is very important for fan optimal control and grid-connected scheduling, and misalignment of measured wind speed not only affects the generating efficiency of the wind turbine generator, but also can cause the generator to be damaged in severe cases. Typical failures of wind speed sensors can be summarized in the following aspects:

(1) destructive failure: the sensor quits operation or the system has no wind speed signal input, and the wind speed data is lack of measurement and missing measurement, for example: the power supply line of the wind speed sensor is damaged, so that the power supply voltage of the wind speed sensor is unstable, and wind speed data is missed;

(2) non-destructive failure: sensor malfunction, measurement data offset from normal values, for example: due to the influence of factors such as thunderstorms, bird impact and the like, the wind cup or the rotating shaft of the wind speed sensor deforms, so that the problems of rotation clamping stagnation, inflexibility and the like of the sensor are caused, and the measured data is misaligned for a long time;

(3) an evolving fault: the wind speed sensor operates in severe environment and complex weather conditions for a long time, the severity of the fault is gradually increased along with the external influence, and the fault state is in a long-term development process. For example: the weather such as storm, rainstorm, hail and the like can affect the performance of lubricating oil of the rotating bearing, and gradually attenuate along with the accumulation of time, if the lubricating oil is not uniformly coated or is not supplemented for a long time, the rotation of the bearing is easy to be inflexible, and the wind measurement data is gradually misaligned.

The technology for monitoring the running state of the wind speed sensor in real time is researched, and the method has important significance for ensuring the running safety, stability and economy of a unit. Typical wind speed sensor condition monitoring has been studied primarily from three directions: (1) a multi-variable correlation analysis method has the basic principle that a multi-dimensional nonlinear mapping model of operating parameters such as the rotating speed of a main shaft of a wind turbine generator, the power of a generator, the pitch angle and the like and the wind speed is established to calculate the effective input wind speed. (2) The sensor redundancy method is characterized in that a plurality of wind speed sensors are simultaneously installed on a single wind turbine generator, and fault sensors are detected by carrying out statistics, comparison and analysis on measurement data of the plurality of sensors. (3) The historical data statistical method is combined with a mathematical statistical method to calculate the expected wind speed of a specific position, fault diagnosis of the sensor is completed through transverse comparison of the expected wind speed and the actually measured wind speed, the statistical-based method has certain engineering application value due to high economical efficiency and feasibility, however, the statistical method based on the historical data has relatively large errors and is easy to generate false detection and missed detection.

Disclosure of Invention

The invention aims to overcome the defects in the prior art and provide a method and a device for diagnosing faults of a wind speed sensor of a wind power plant.

The purpose of the invention can be realized by the following technical scheme:

a wind power plant wind speed sensor fault diagnosis method comprises the following steps:

carrying out fault mutual diagnosis community division on the wind power plant set;

respectively carrying out fault diagnosis on the wind speed sensors corresponding to the unit in the fault mutual diagnosis community, and the fault diagnosis method comprises the following steps:

acquiring the wind speed time sequence flow of a wind speed sensor in the fault mutual diagnosis community,

determining a time interval in which wind speed abnormal data exist in the fault inter-diagnosis community,

and positioning the wind speed sensor which has faults in the time interval with the wind speed abnormal data.

Preferably, the method for dividing the fault inter-diagnosis community comprises the following steps:

and screening the spatial correlation of the wind speed of the units on the basis of the yaw data of the units, and screening a plurality of units with higher spatial correlation of the wind speed as associated units and dividing the associated units into a fault mutual diagnosis community.

Preferably, the specific manner for determining the time interval in which the wind speed abnormal data exists in the fault inter-diagnosis community includes:

acquiring index parameters describing the separation degree of wind speed data measured by each wind speed sensor in the fault mutual diagnosis community at each sampling moment;

determining a variable point index parameter threshold value by taking the index parameter as a variable point index for determining abnormity;

and determining the time interval with the index parameter larger than the variable point index parameter threshold value as the time interval with the wind speed abnormal data.

Preferably, the index parameter includes a variance of wind speed data measured by each wind speed sensor in the fault mutual diagnosis community at each time point.

Preferably, the obtaining method of the variable point index parameter threshold includes:

the index parameters at each sampling moment are arranged according to an ascending order, the arranged index parameters are averagely divided into a plurality of parts according to a quantile algorithm, and the index parameters corresponding to the set quantile points are selected as variable point index parameter thresholds.

Preferably, the means for locating the malfunctioning wind speed sensor comprises:

and carrying out wind speed correlation cross comparison on each wind speed sensor in the fault mutual diagnosis community, and determining the unit sensor causing the reduction of the wind speed correlation as a fault wind speed sensor.

Preferably, the specific way of locating the wind speed sensor with the fault is as follows: and rolling to calculate the correlation of the wind speeds measured by any two wind speed sensors, selecting unit data with the correlation coefficient smaller than a set threshold value, counting the occurrence frequency of the associated wind speed sensors, and determining the wind speed sensor with the maximum occurrence frequency as a fault wind speed sensor.

Preferably, the wind speed dependence is described by pearson coefficient quantization, expressed as:

wherein r is_iTo quantify the correlation, X, of the wind speed data measured by two wind speed sensors in time series i_i,t、Y_i,tCorresponding to the wind speed data X of two wind speed sensors at the sampling time t in the quantitative time sequence i_i,avg、Y_i,avgThe average value of the wind speed data of the two wind speed sensors in the quantization time sequence i corresponds to, and n is the total number of sampling moments in the quantization time sequence i.

Preferably, during fault diagnosis, the mutual fault diagnosis community is re-classified when the wind direction changes beyond a threshold value.

A wind farm wind speed sensor fault diagnosis device comprises a memory and a processor, wherein the memory is used for storing a computer program, and the processor is used for realizing the wind farm wind speed sensor fault diagnosis method when the computer program is executed.

Compared with the prior art, the invention has the following advantages:

(1) according to the method, the wind speed sensor array of the wind power plant is divided into a plurality of fault mutual diagnosis communities by taking the spatial correlation of the wind speed as a measure, the distribution characteristics of the fault wind speed are mined in the fault mutual diagnosis communities, the state of the wind speed sensor is diagnosed in time and space, and the abnormal state of the wind speed sensor can be quickly and accurately positioned and sensed;

(2) the method utilizes the characteristic that the wind speed data separation degree between the fault inter-diagnosis intra-cluster units is increased due to the fault wind speed data to formulate a criterion, analyzes the distribution condition of the variance statistics of the wind speed data, realizes accurate positioning of the time sequence interval of the fault data, and effectively avoids the phenomena of false detection and missed detection;

(3) according to the invention, a multi-machine correlation cross comparison mode is adopted to position the faulty sensor, so that the faulty wind speed sensor can be accurately positioned, and the accuracy of a fault diagnosis result is improved;

(4) according to the method, the wind direction change threshold value is set, when the wind direction change exceeds the threshold value, the fault inter-diagnosis community needs to be divided again, high correlation among the fault inter-diagnosis community units is ensured all the time, and therefore the fault diagnosis result is accurate.

Drawings

FIG. 1 is a flow chart of a wind farm wind speed sensor fault diagnosis method of the present invention.

Detailed Description

The invention is described in detail below with reference to the figures and specific embodiments. Note that the following description of the embodiments is merely a substantial example, and the present invention is not intended to be limited to the application or the use thereof, and is not limited to the following embodiments.

Example 1

As shown in fig. 1, the present embodiment provides a method for diagnosing a fault of a wind speed sensor in a wind farm, where the method includes:

carrying out fault mutual diagnosis community division on the wind power plant set;

respectively carrying out fault diagnosis on the wind speed sensors corresponding to the unit in the fault mutual diagnosis community, and the fault diagnosis method comprises the following steps:

acquiring the wind speed time sequence flow of a wind speed sensor in the fault mutual diagnosis community,

determining a time interval in which wind speed abnormal data exist in the fault inter-diagnosis community,

and positioning the wind speed sensor which has faults in the time interval with the wind speed abnormal data.

The method comprises the following main contents: the classification of the fault mutual diagnosis community and the fault diagnosis and positioning of the sensor are specifically described as follows:

1. fault inter-diagnosis community division

The wind speed has spatial correlation, the wind speed spatial correlation among the units is used as a division standard, the wind generation units with the strong wind speed spatial correlation are divided into a fault mutual diagnosis community, and the same wind power plant can be divided into a plurality of fault mutual diagnosis communities to realize group cooperative mutual diagnosis. For the ultra-short-term wind speed mapping model, factors influencing the spatial correlation of wind speed are mainly caused by the change of atmospheric pressure, and random fluctuation of wind speed and wind direction is caused. The wind speed correlation strength among the wind turbine generators directly determines the accuracy of a wind speed prediction result based on the traditional spatial correlation, so that the seeds of the associated wind turbine generators are preferably the core part of a real-time prediction system of the wind turbine generators. Aiming at the real-time perception and prediction of the wind speed, the number of the associated seed units of the target unit is generally more than two, so that the reliability and the anti-interference performance of the wind speed prediction value can be ensured. The screening of the associated wind turbine generator can be carried out according to the following steps: (1) and (3) wind speed correlation time sequence analysis: because the wind speed correlation has strong time-varying property, the time sequence change of the wind speed correlation of the associated units is considered, and the change characteristics of the wind speed correlation coefficient of each alternative seed unit and the target unit within one day along with the time are analyzed according to the historical wind power operation data, so that a basis is provided for the subsequent dynamic screening of the seeds of the associated units; (2) screening a dynamic association seed set: and setting a threshold value of the associated unit according to the correlation coefficient of each wind turbine generator to be selected and the target unit at different time, and preferably selecting the associated seed unit when the wind speed correlation coefficient of the alternative unit reaches the threshold value range. According to the principle, the optimal results of the dynamic association seed set under different time of day can be obtained. Because the wind speed and the wind direction have strong time-varying characteristics, the fault inter-diagnosis community needs to be dynamically updated according to the measured data, a wind direction change threshold value is set, when the wind direction change exceeds the threshold value, the fault inter-diagnosis community needs to be divided again, and high correlation of the inter-diagnosis community is ensured to be kept all the time.

2. Sensor fault diagnosis positioning

According to the invention, the abnormal diagnosis of the sensor is researched from two application scenes of the fault which can be identified and the fault which cannot be identified by a wind power plant diagnosis system (SCADA system), wherein for the fault data which can be identified by the SCADA system, for example: the system can alarm and mark the fault state after identifying the abnormal state, so that the abnormal state can be judged based on the state mark value of the data. For the fault state which can not be identified by the system, the invention carries out the spread analysis of the distribution characteristic dissimilarity of the fault data and the normal data, identifies the statistical characteristic of the abnormal data through the variable point algorithm and locates the starting moment of the fault.

The point change problem refers to a point where some quantities of a certain subsequence in a time sequence suddenly change, and in mathematical statistics, the direct reaction of the point change problem is that the mean value, variance and other statistical characteristics of the sequence suddenly change, and data may change in some germplasm. Therefore, the method has theoretical feasibility for the abnormity diagnosis of the actually measured wind speed data by adopting the variable point analysis. When the wind parameter data state diagnosis problem is researched, the position parameter of sudden change of wind speed data is mainly concerned, and the research result shows that if abnormal data exists in a wind speed interval, the change rate, the mean value and the variance of the wind speed all have sudden change, the variance serves as the basis of grouping of the variable points, and the separation degree of the wind speed data measured by a plurality of associated units can be described better, so that the wind speed variance is selected as the grouping basis of the variable points.

Therefore, the specific manner for determining the time interval in which the wind speed abnormal data exists in the fault inter-diagnosis community includes:

acquiring index parameters describing the separation degree of the wind speed data measured by each wind speed sensor in the fault mutual diagnosis community at each sampling moment, wherein the index parameters comprise the variance of the wind speed data measured by each wind speed sensor in the fault mutual diagnosis community at each moment;

the index parameter is used as a variable point index for determining abnormity, a variable point index parameter threshold value is determined, and the acquisition mode of the variable point index parameter threshold value comprises the following steps: the index parameters at each sampling moment are arranged according to an ascending order, the arranged index parameters are averagely divided into a plurality of parts according to a quantile algorithm, and the index parameters corresponding to the set quantile points are selected as variable point index parameter thresholds;

and determining the time interval with the index parameter larger than the variable point index parameter threshold value as the time interval with the wind speed abnormal data.

The specific steps of the variable point grouping method are as follows:

the wind speed sequence at a certain moment is recorded as follows:

V_t＝(v_1,t,v_2,t,......,v_n,t)

in the formula, v_n,tRepresenting the wind speed of the nth unit at time t.

The variance of each wind speed at the moment t is calculated as:

in the formula, S_tRepresents the sequence V_tVariance of v_t ^meanRepresents the sequence V_tThe average wind speed of (2).

Further, the variance of the wind speed sequence in the time sequence to be diagnosed is respectively solved to obtain S_t1、S_t2……S_tNThe subscripts t1, t2 … … tN denote the sample times within the time series to be diagnosed, S_t1、S_t2……S_tNAccording to ascending order, the embodiment adopts a quartile algorithm to determine the variable point index parameter threshold, wherein the quartile is that one ordered data sample is averagely divided into four parts and three division pointsThe numerical values of the positions are an upper quartile, a median and a lower quartile, which are respectively marked as Q₁、Q₂、Q₃. For a sequence sample containing abnormal data, the variance of the abnormal data sequence is far larger than that of normal sequence data, so that the upper quartile is selected as a boundary value for judging the abnormal data, and the data in a time interval with the variance value larger than the upper quartile is judged as the abnormal data.

The upper quartile is selected as a boundary value of the fault, actually, a criterion is formulated by utilizing the characteristic that the separation degree of the wind speed data among the fault inter-diagnosis intra-group units is increased due to fault wind speed data, and the time sequence interval of the fault data can be positioned by analyzing the variance statistic subsection condition of the wind speed data. In practical application, different services have different requirements on data quality, data scale, transmission speed and the like, and the fault boundary value formulation needs to be further perfected and refined by combining with the actual service category requirements.

Further, to determine the spatial location of the failed wind speed sensor, the manner of locating the failed wind speed sensor includes:

and carrying out wind speed correlation cross comparison on each wind speed sensor in the fault mutual diagnosis community, and determining the unit sensor causing the reduction of the wind speed correlation as a fault wind speed sensor.

Specifically, in the fault mutual diagnosis community, the unit has high wind speed correlation, and the unit sensor with reduced wind speed correlation can be determined as the fault wind speed sensor. And rolling the wind speed correlation coefficients among the computer groups, selecting a plurality of groups of data with the minimum correlation coefficients, and counting the occurrence times of the associated units, wherein the wind speed sensor with the maximum occurrence times is the fault wind speed sensor. The wind speed correlation is determined by a rolling statistical method, namely, the rolling statistical method is to select a continuous wind speed sequence consisting of the point and n-1 wind speed sequences positioned in front of the point in time sequence, and further adopt Pearson coefficient quantization description in a quantization time sequence, wherein the description is as follows:

wherein r is_iTo quantify the correlation, X, of the wind speed data measured by two wind speed sensors in time series i_i,t、Y_i,tCorresponding to the wind speed data X of two wind speed sensors at the sampling time t in the quantitative time sequence i_i,avg、Y_i,avgThe average value of the wind speed data of the two wind speed sensors in the quantization time sequence i corresponds to, and n is the total number of sampling moments in the quantization time sequence i.

According to the method, the wind speed sensor array of the wind power plant is divided into a plurality of fault mutual diagnosis communities by taking the spatial correlation of the wind speed as a measure, the distribution characteristics of the fault wind speed are mined in the fault mutual diagnosis communities, the state of the wind speed sensor is diagnosed in time and space, and the abnormal state of the wind speed sensor can be quickly and accurately positioned and sensed. And a wind direction change threshold value is set, when the wind direction change exceeds the threshold value, the fault mutual diagnosis community needs to be divided again, and high correlation is kept among the fault mutual diagnosis community units all the time, so that the fault diagnosis result is more accurate.

Example 2

The embodiment provides a fault diagnosis device for a wind power plant wind speed sensor, which includes a memory and a processor, where the memory is used to store a computer program, and the processor is used to implement the fault diagnosis method for the wind power plant wind speed sensor in embodiment 1 when executing the computer program, where the fault diagnosis method for the wind power plant wind speed sensor is described in embodiment 1 specifically, and is not described again in this embodiment.

The above embodiments are merely examples and do not limit the scope of the present invention. These embodiments may be implemented in other various manners, and various omissions, substitutions, and changes may be made without departing from the technical spirit of the present invention.