JP2009243428A - Monitoring device, method and program of wind mill - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To automatically monitor the state of a wind mill and to conduct its state assessment based on an appropriate standard quantitatively. <P>SOLUTION: A monitoring device of the wind mill is provided with a first storage part 25 for storing diagnosed data files, a second storage part 28 for storing normal data files, a diagnosis setting part 29 for extracting a plurality of data sets used for diagnosis from the first storage part 25 to set them and for extracting a plurality of data sets used for diagnosis from the second storage part to set them, an index value calculation part 30 for calculating a state index value representing the state of the mill by using a statistical calculation method based on the data sets of diagnosed data files which have been set and data sets of reference data files which have been set, an abnormality judging part 31 for assessing the state of the wind mill based on the state index value, and a notification part 32 for notifying an assessment result. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a windmill monitoring apparatus and method suitable for application to a windmill that converts wind energy into electric power, and a program.

Conventionally, in wind turbine monitoring technology, for example, measurement data corresponding to operating conditions is automatically acquired from sensors attached to various locations of the wind turbine, and the measurement data related to each measurement item is compared with a pre-registered threshold value, respectively. Therefore, it is generally known to monitor the operation state (normal, attention required, failure, etc.).
JP 2006-342766 A

By the way, it is very difficult even for an expert to determine whether or not the state quantity related to each measurement item obtained from the measurement data measured by each sensor indicates an abnormality. As described above, the measurement data is registered in advance. With a method of determining an abnormality by comparing with a threshold value, a highly accurate determination result cannot be expected.
In addition, when there are a plurality of combinations of correlations among a plurality of measurement items, it is preferable to determine the driving state in consideration of these correlations, but it is difficult to automatically determine such complicated processing. there were.
Further, in order to improve diagnosis accuracy, it may be possible to make a diagnosis by an expert, but it is not preferable from the viewpoint of labor and time to process a large amount of measurement data. In the case of diagnosis by an expert, the accuracy and reproducibility of abnormality detection during driving and factor analysis depends on the knowledge and skill level of the individual expert. Therefore, there arises a problem that a large variation occurs in the diagnosis level of wind turbine equipment abnormality detection and factor analysis.

  The present invention has been made to solve the above-described problem. A wind turbine monitoring apparatus, method, and program capable of automatically monitoring a wind turbine state and quantitatively evaluating the state according to an appropriate standard. The purpose is to provide.

In order to solve the above problems, the present invention employs the following means.
The present invention is a windmill monitoring device that monitors the state of the windmill using characteristic values created based on measurement data measured by a plurality of sensors provided in the windmill, and is associated with a measurement time. When a plurality of characteristic values are stored for each characteristic item and the characteristic values associated with the same measurement time are set as one data set, the data set corresponds to the characteristic value of a predetermined characteristic item. The first storage means storing the identification information indicating the class classification determined in this manner and a plurality of characteristic values associated with the measurement time are stored for each characteristic item and associated with the same measurement time If the characteristic value is a single data set, identification information indicating the class classification determined according to the characteristic value of a predetermined characteristic item is assigned to the data set. And a second storage means belonging to a predetermined reference range in which characteristic values of the specific characteristic items constituting the data set are defined in advance, and a plurality of the data used for diagnosis from the first storage means A diagnostic setting means for extracting and setting a set and extracting and setting a plurality of the data sets used for the diagnosis from the second storage means; and data of the diagnostic data file set by the diagnostic setting means Based on the set and the data set of the reference data file, using a statistical calculation method, an index value calculating unit that calculates a state index value representing the state of the windmill, and a state index calculated by the index value calculating unit A wind turbine monitoring device comprising: an evaluation unit that evaluates the state of the wind turbine based on a value; and a notification unit that notifies an evaluation result by the evaluation unit. To.

  As described above, since the state index value representing the state of the windmill is calculated using the diagnosis data file and the reference data file, quantitative evaluation is realized instead of qualitative evaluation based on experience and knowledge. It becomes possible. Further, since the state index value is a value in consideration of the class classification given to each data set, it is possible to compare data acquired under the same situation. Thereby, it becomes possible to evaluate the state of a windmill more appropriately.

  In the wind turbine monitoring device, the characteristic items include an environmental classification related to an environment surrounding the wind turbine, a performance classification related to the performance and power generation conditions of the wind turbine, and a characteristic classification related to diagnosis of an operation state related to various monitoring parts set in the wind turbine. It is good also as being roughly divided into three.

  In the wind turbine monitoring apparatus, the class classification may be determined according to a characteristic value of a predetermined characteristic item classified into at least one of the environmental classification and the performance classification.

  Since the classification is determined based on the environmental classification and performance classification in this way, it is possible to classify the data set according to the environmental situation and power generation situation when the characteristic values classified in the characteristic classification are acquired. It becomes.

  In the wind turbine monitoring device, the plurality of data sets stored in the second storage unit are the specific characteristic items among the plurality of data sets generated from the plurality of measurement data collected from the wind turbine. Only a data set belonging to a reference range in which the characteristic values related to are set in advance may be extracted.

  Since the reference data file is generated based on the measurement data measured in the same windmill, it is possible to evaluate the state of the windmill using the reference data file reflecting the characteristics of the solid. Thereby, it becomes possible to judge the state of a windmill more correctly.

In the wind turbine monitoring apparatus, the reference range may be set for each class classification.
By doing in this way, it becomes possible to create a reference data file using an appropriate reference range set for each class classification.

  In the wind turbine monitoring apparatus, the index value calculation means obtains the characteristic distribution of the reference data set by the diagnosis setting means, obtains the characteristic distribution of the diagnosis data, and the characteristic distributions are different from each other. The state index value may be calculated by quantitatively obtaining the distance.

  In this way, the characteristic distributions of each other are obtained, and the distance at which these distributions are separated is obtained quantitatively, so it is possible to quantitatively determine how far the diagnosis data is from the characteristic distribution of the reference data. It becomes possible to evaluate.

  In the wind turbine monitoring apparatus, the state index value calculated by the index value calculation means is, for example, a Mahalanobis distance calculated using a Mahalanobis Taguchi method.

  The wind turbine monitoring device may further include a factor analysis unit that performs a factor analysis of the abnormality when the evaluation unit evaluates that an abnormality has occurred.

  In this way, by performing factor analysis, it is possible to quickly grasp which part is determined to be abnormal due to the cause. This makes it possible to respond quickly.

  The present invention is a wind turbine group monitoring system for monitoring the state of a part or the whole of a wind farm including a plurality of wind turbines, comprising any one of the wind turbine monitoring devices described above, and obtained by the wind turbine monitoring device. Provided is a wind turbine group monitoring system that monitors the state of a part or the whole of a wind farm based on the state index value of each wind turbine and the operation performance of the part or the whole of the wind farm.

  According to such a configuration, it is possible to relatively evaluate a part (segment) or the whole of the wind farm and identify a wind turbine that exhibits different characteristics compared to other wind turbines. . In this way, by evaluating the state between wind turbines, it is possible to monitor the state of each wind turbine using a more appropriate criterion.

  The present invention is a windmill monitoring method for monitoring the state of the windmill using characteristic values created based on measurement data measured by a plurality of sensors provided in the windmill, and is associated with a measurement time. In the process of creating a diagnosis data file in which a plurality of characteristic values are stored for each characteristic item, and in the diagnosis data file, when the characteristic values associated with the same measurement time are set as one data set, The process of assigning identification information indicating the class classification determined according to the characteristic value of the predetermined characteristic item to the data set, and the characteristic value related to the specific characteristic item belong to a predetermined reference range. In addition, the process of creating a reference data file in which the characteristic values of each characteristic item are associated with the measurement time, and the same measurement in the reference data file When the characteristic value associated with time is set as one data set, the process of giving identification information indicating a class classification determined according to the characteristic value of a predetermined characteristic item to the data set; Extracting and setting a plurality of the data sets used for diagnosis from the diagnosis data file, extracting and setting the plurality of data sets used for the diagnosis from the reference data file, and the set diagnosis data Based on the data set of the file and the data set of the reference data file, a process of calculating a state index value representing the state of the windmill using a statistical calculation method, and based on the state index value, There is provided a windmill monitoring method including a process of evaluating a state and a process of notifying a result of the evaluation.

  The present invention is a windmill monitoring program used to monitor the state of the windmill using characteristic values created based on measurement data measured by a plurality of sensors provided in the windmill, A process for creating a diagnosis data file in which a plurality of characteristic values associated with measurement times are stored for each characteristic item, and the characteristic values associated with the same measurement time in the diagnosis data file are stored in one data set. In this case, a process for giving identification information indicating a class classification determined according to the characteristic value of a predetermined characteristic item to the data set, and a characteristic value related to the specific characteristic item are defined in advance. And a process for creating a reference data file in which the characteristic value of each characteristic item is associated with the measurement time, and the reference data file In this case, when the characteristic values associated with the same measurement time are set as one data set, identification information indicating class classification determined according to the characteristic value of a predetermined characteristic item is given to the data set. Extracting and setting a plurality of the data sets used for diagnosis from the diagnosis data file, extracting and setting the plurality of data sets used for the diagnosis from the reference data file, and setting Based on the data set of the diagnosed data file and the data set of the reference data file, a process for calculating a state index value representing the state of the windmill using a statistical calculation method, and based on the state index value And monitoring the windmill for causing the computer to execute a process for evaluating the state of the windmill and a process for notifying the result of the evaluation. To provide a program.

  According to the present invention, it is possible to automatically monitor the state of the windmill, and it is possible to quantitatively perform the state evaluation based on an appropriate standard.

  DESCRIPTION OF EMBODIMENTS Embodiments of a windmill monitoring apparatus and method and a program according to the present invention will be described below with reference to the drawings.

[First Embodiment]
FIG. 1 is a diagram showing a schematic configuration of a windmill. As shown in FIG. 1, the windmill 1 is provided in the nacelle 3 so as to be rotatable around a substantially horizontal axis line, with a support column 2 standing on the foundation 6, a nacelle 3 installed at the upper end of the support column 2, and The rotor head 4 is provided. A plurality of wind turbine blades 5 are attached to the rotor head 4 radially around its rotational axis. As a result, the force of the wind striking the wind turbine blade 5 from the direction of the rotation axis of the rotor head 4 is converted into power for rotating the rotor head 4 around the rotation axis, and this power is converted into electric energy by the generator. It has become.

FIG. 2 is a block diagram showing a schematic configuration of a wind turbine monitoring device (hereinafter referred to as “monitoring device”) according to the present embodiment. The monitoring device according to the present embodiment shown in FIG. 2 may be installed at any place inside or outside the wind turbine configuration. As shown in FIG. 2, the monitoring device 10 is a computer system (computer system), and includes a CPU (Central Processing Unit) 11, a RAM (Random Access).
Main memory 12, such as Memory, ROM (Read Only Memory), HDD (Hard Disk)
Auxiliary storage device 13 such as a drive), an input device 14 such as a keyboard and a mouse, an output device 15 such as a monitor and a printer, and a communication device 16 that exchanges information by communicating with external devices. Yes.
Various programs (for example, a monitoring program) are stored in the auxiliary storage device 13, and the CPU 11 reads out the programs from the auxiliary storage device 13 to the main storage device 12 and executes them to implement various processes.

  FIG. 3 is a functional block diagram in which the functions of the monitoring device 10 are expanded. As shown in FIG. 3, the monitoring device 10 includes a measurement information storage unit 21, a data generation unit 22, a class classification definition unit 23, a class classification unit 24, and a first storage unit (first storage unit) 25. A normal data condition definition unit 26, a normal data extraction unit 27, a second storage unit (second storage unit) 28, a diagnosis setting unit (diagnosis setting unit) 29, and an index value calculation unit (index value calculation unit) ) 30, an abnormality determination unit (evaluation unit) 31, and a notification unit (notification unit) 32.

The measurement information storage unit 21 stores a plurality of data files including a plurality of measurement data for each sensor.
Here, each measurement data of each data file is associated with a measurement time during which the measurement data was measured. This measurement time functions as a linking parameter for associating various measurement data between data files with each other in a diagnostic data creation process performed in a class classification unit 24 described later.

The data generation unit 22 mainly executes the following two processes.
[Unification processing of sampling time]
The time interval (hereinafter referred to as “sampling time”) of the measurement time of each measurement data related to various data files stored in the measurement information storage unit 21 described above is not unified. Therefore, the data generation unit 22 first performs processing for unifying these sampling times. In this embodiment, each data file is reconstructed so as to be measurement data at 1-minute intervals.

For example, when the sampling time is sufficiently faster than 1 minute, a representative value for 1 minute is selected by a statistical method using all measurement data acquired in 1 minute. For example, the representative value is represented by an average value and a standard deviation.
By doing in this way, the measurement data of all the data files can be synchronously related at a common time interval.

[Calculation of diagnostic physical quantity]
Next, the data generation unit 22 extracts “diagnosis physical quantity” for a predetermined data file among various data files in which the measurement time is unified.
That is, as described above, raw data measured by various sensors is stored in the measurement information storage unit 21. In order to diagnose the driving situation of each monitored part, these raw data are stored. It is necessary to generate and extract a diagnostic physical quantity suitable for diagnosis from the above data.

  For example, in a bearing / speed increaser, eight acceleration sensors are attached to each measurement location in order to monitor the operation status. A time-series waveform, which is measurement data measured by each sensor, is stored in the measurement information storage unit 21 for each sensor.

  By the way, in the diagnosis of the abnormality of the gear stage with the speed increaser in the bearing / speed increaser, the vibration acceleration of the meshing frequency (eigenvalue) of a plurality of gears constituting the speed change stage is calculated, and this vibration acceleration is The shift stage is diagnosed. For this reason, the data generation unit 22 performs signal processing called frequency conversion (for example, Fast Fourier Transfer means) on the time-series waveforms respectively measured by the eight sensors CH1 to CH8, and FIG. A frequency spectrum as shown in FIG. 4 is obtained, and amplitude accelerations at a plurality of natural frequencies indicated by arrows in FIG. 4 are extracted from the frequency spectrum. Then, a new data file is created by storing the extracted amplitude acceleration in a file identified by each channel and natural frequency.

FIG. 5 shows an example of a newly created data file. As shown in FIG. 5, each measurement is associated with a channel (sensor) expressed as CH1, CH2, etc., and a natural frequency expressed as “AZi1,” “AZi2,” “AZi3,” etc. The vibration acceleration in time is stored as a diagnostic physical quantity. Here, as described above, the measurement time of each diagnostic acceleration is unified with the measurement time of other data files.
In this way, a diagnostic physical quantity is calculated in a predetermined data file, and a new database is created.

  The calculation of the diagnostic physical quantity is performed mainly on measurement data measured by sensors attached to various monitoring parts in order to diagnose the operating state of the wind turbine 1. What signal processing is performed on the measurement data of which data file and what diagnostic physical quantity is calculated are registered in the data generation unit 22 in advance.

  Various data files whose sampling times are unified by the data generation unit 22 and data files for newly created diagnostic physical quantities are output to the class classification unit 24. Note that the original data file used to calculate the diagnostic physical quantity is not particularly required for the subsequent processing, and is therefore not output to the class classification unit 24.

The class classification unit 24 first creates one diagnostic data file by integrating various data files input from the data generation unit 22.
FIG. 6 shows an example of a diagnosis data file. As shown in FIG. 6, the measurement data or the diagnostic physical quantity at each measurement time is associated with each sensor. In the present embodiment, “normalized wind speed”, “MET wind speed turbulence”, “power transmission end output”, “AZi1”, “attribute” indicating the attributes of each measurement data and diagnostic physical quantity described at the top of the table of FIG. A heading such as “AZi2” is called “characteristic item”, and each data of each characteristic item is defined as “characteristic value”.

The characteristic items are classified into “environment”, “performance”, and “characteristic” according to their attributes. “Environment” includes items related to the environment surrounding the wind turbine such as “wind speed” and “wind speed turbulence”, and “performance” includes the power generation conditions, generator speed, command values related to power generation control, etc. Characteristic items related to performance and power generation conditions are classified into “characteristics” such as “AZi1”, “AZi2”, etc. Here, the characteristic item classified as “characteristic” corresponds to a data file newly generated in the data generation unit 22 described above.
In this embodiment, one diagnostic data file is configured by integrating characteristic values in increments of 1 minute from 0:00 to 23:59 on a certain day.

Subsequently, the class classification unit 24 sets characteristic values associated with the same measurement time as one data set in the diagnosis data file, and adds identification information indicating the class classification to each data set.
Specifically, the class classification unit 24 selects which data set based on the class definition defined in the class classification definition unit 23, in other words, for each line of the diagnostic data file shown in FIG. Whether the data belongs to the class classification is classified, and a flag indicating the class classification is set in each data set.

  “Class classification” as used herein refers to data that matches the reference ranges of a plurality of characteristic items defined by the class classification definition unit 23 in a statistical diagnosis method such as the Mahalanobis Taguchi method (hereinafter referred to as “MT method”). This is the separation of groups. As described above, the statistical diagnosis for identifying normality or abnormality between the same “class classification” has higher identification accuracy than the case of the whole data not classified.

“Class classification”, which is a feature of the present invention in a windmill, will be described with reference to a schematic diagram of class classification in FIG.
In the present embodiment, the characteristic item that is an index amount for class classification is “wind speed”. “Wind speed” is advantageous because it has a strong correlation with the power generation performance of the wind turbine. Here, boundary values (conditions) are set with physical quantities for each stage of “wind speed”, and classes F0, F1, F2, and F3 and class classification are defined.

As specific implications for windmill performance, F0 is the wind speed range where the windmill does not contribute to power generation, F1 is the low speed range where power generation starts, F2 is the medium speed range where power generation begins to become full-scale, and F3 is the rating at which rated power generation begins It is considered as an area.
Since the class F0 is a wind speed region where the windmill does not contribute to power generation, the class F0 is defined as a data group that is not a diagnosis target, and the three classes F1 to F3 are defined as classes to be diagnosed.

Instead of the class classification, each measurement time may be classified into a plurality of classes based on two parameters, a wind speed turbulence rate and a wind speed. FIG. 8 shows a case in which the wind turbulence rate (%) is divided into a case where the wind turbulence rate (%) is small and a case where the wind turbulence rate (%) is large and is classified into a total of six classes F1 ′ to F6.
In the above example, the class classification is performed based on the wind speed or the like. However, as indicated in parentheses in FIG. 7, the class classification is performed based on the “rotation speed” indicating a strong correlation with the power generation performance of the windmill. It is good.

  Information defining the class classification as described above is stored in the class classification definition unit 23. The class classification unit 24 assigns a class classification for each data set (for each row) of the diagnosis data file with reference to the condition value of the class classification stored in the class classification definition unit 23.

  The first storage unit 25 stores the diagnosis data file classified (flagged) by the class classification unit 24. The diagnosed data file with classification classified stored in the first storage unit 25 is handled as a “signal space” in the calculation processing in the index calculation unit 30.

On the other hand, in the second storage unit 28, the condition of the normal definition unit 26 is selected from the “past diagnosed data file” through the processing of the normal data extraction unit 27 from the “past diagnostic data file” that has already been acquired. Only data sets that are determined to be “normal” are extracted and stored in accordance with instructions for defining expressions, values, and the like. The stored data is called “normal data”.
Of course, at this stage, a class classification flag is added to each data set of the normal data file stored in the second storage unit 28 in the class classification unit 24 of the preceding process.

  The normal data file stored in the second storage unit 28 is automatically generated by the normal data extraction unit 27, and is automatically updated and accumulated daily from past normal data files. Here, a processing function for automatically generating a “normal data file” from the “diagnostic data file” stored in the first storage unit 25 will be described with reference to the drawings.

  The normal data extraction unit 27 refers to the normal range (reference range) defined in the normal data condition definition unit 26 to extract only the data set of the diagnosis data file that matches the definition, and the second storage unit Transfer to 28 and store.

Preferred examples of the definition conditions of the normal data condition definition unit 26 according to the present invention are shown in FIGS.
FIG. 9 is a normal definition in terms of performance that defines the fluctuation range of the power generation amount.
In FIG. 9, the horizontal axis represents the wind speed, and the vertical axis represents the power generation amount P. The performance curve of the power generation amount can be expressed as a function of P (V, r). As for the normal range of the power generation amount, the normal ranges (I), (II), and (III) are defined for each of the class classifications F1 to F3 based on the wind speed, and the condition that defines the normal performance range is The fluctuation range is ± ΔP / 2 from the performance curve.

The [Delta] P is any wind speed V _0, the power generation amount when the rotational speed r _0, the ratio A (%) of the rotation speed r ₀ (V ₀₎ and by defining, ΔP = (A / 100) × P 0 (V ₀ , r ₀ ).

  In the above example, the vertical axis is the power generation amount P, but instead, the rotational speed r may be used. In this case, the rotational speed performance curve can be expressed as a function of r (V). As for the normal range of the rotational speed, the normal ranges (I), (II), and (III) are defined for each of the class classifications F1 to F3 based on the wind speed, and the condition that defines the normal performance range is The fluctuation range is ± Δr / 2 from the performance curve.

The Δr is any wind speed V _0, the power generation amount when the rotational speed r _0, the ratio A (%) of the rotation speed r ₀ (V ₀₎ and by defining, ΔP = (A / 100) × P 0 (V ₀ , r ₀ ), Δr = (A / 100) × r ₀ (V ₀ ).
Furthermore, in the above example, the horizontal axis is the wind speed, but in FIG. 9, the horizontal axis may be the rotational speed r and the vertical axis may be the power generation amount P. In this case, the normal ranges (I), (II), and (III) are defined for each class classification corresponding to the rotational speed.

FIG. 10 is a normal definition from an approximately normal operating condition in which the variation range of the wind direction deviation is defined.
In FIG. 10, the horizontal axis represents the wind speed, and the vertical axis represents the wind direction deviation Δθ. Here, the “wind direction deviation” will be described. Usually, the “normal” operating condition of the wind turbine is based on the premise that the rotating surface of the wind turbine blade is always received in front of the wind direction.

  In other words, the difference between the actual wind direction and the direction of the wind turbine blade rotation surface is called “wind direction deviation”, and the wind direction deviation Δθ when they are in an ideal state directly in front is set to zero. The normal range of wind direction deviation Δθ, which is one of the “normal” operating conditions, is defined as normal ranges (I ′), (II ′), and (III ′) for each of the class classifications F1 to F3 based on wind speed. ing.

  That is, the direction of the windmill is supplemented by the wind while following the direction of the wind constantly changing under natural conditions so that it falls within the range of (I ′), (II ′), and (III ′). Power generation is performed to meet the conditions.

  In the above example, the horizontal axis is the wind speed V, but instead, the horizontal axis may be the rotational speed r. In this case, the normal ranges (I ′), (II ′), and (III ′) are defined for each class classification corresponding to the rotational speed.

  Logical expressions (AND, OR, NOT) of the normal ranges (I), (II), (III) and (I ′), (II ′), (III ′) stored in the normal data condition definition unit 26 )) Can be arbitrarily set by the user.

  In accordance with the definition of the normal range stored in the normal data condition definition unit 26, the normal data extraction unit 27 extracts only the data set determined as the normal range from the diagnosis data file and stores it in the second storage unit 28. By storing, only normal data is stored in the second storage unit 28. These normal data files stored in the second storage unit 28 are handled as “unit spaces” in the calculation processing in the index calculation unit 30.

  The diagnosis setting unit 29 extracts the whole or a part of the data set from the diagnosis data file in the first storage unit 25, creates a “diagnosis data file” (signal space) for diagnosis, and stores it in the second storage. The whole or a part of the data set is extracted from the normal data file of the unit 28, and a “reference data file” (unit space) for diagnosis is created.

  As described above, the class classification flag is added to each data set of these “diagnostic data file” and “reference data file”. Although details will be described later, basically, in the MT method, abnormality diagnosis is performed by comparing the “reference data file” and the “diagnostic data file” between the same class classifications. In addition, depending on the purpose of diagnosis, it is possible to perform comparative abnormality diagnosis between different classes, or make a plurality of classes as one group and reconstruct the class to perform comparative abnormality diagnosis.

  The index value calculation unit 30 determines the soundness of the bearing / speed increaser in the wind turbine 1 using a statistical diagnosis method based on the “diagnosis data file” and the “reference data file” created by the diagnosis setting unit 29. The indicated state index value is calculated. Specifically, the index value calculation unit 30 normalizes the “diagnosis data file” and the “reference data file” created by the diagnosis setting unit 29, and normalizes the “diagnosis data file” and the “reference data file”. ”Is obtained, the state in which the data distributions (groups) are deviated from each other is quantitatively obtained as the distance between the distributions, and the distance is treated as a soundness state index value.

  More specifically, the MT method is used as the statistical diagnosis method of the index value calculation unit 30, and the Mahalanobis distance (hereinafter referred to as “MD”) as a state index value of soundness that is one of the diagnostic output results obtained by the MT method. Value)). A specific calculation method will be described later.

  The abnormality determination unit 31 compares the state index value calculated by the index value calculation unit 30 with a preset threshold value, and evaluates the state of the windmill 1 according to the comparison result. For example, when the state index value exceeds a threshold value, it is determined to be abnormal, and an abnormality determination signal is output to the notification unit 32.

  When the abnormality determination signal is input, the notification unit 32 notifies the user of the abnormality by displaying the occurrence of the abnormality of the windmill 1 on a display or the like. In addition, instead of or in addition to the visual notification method, an abnormality may be notified by auditory notification, for example, a report sound. Thus, the notification method is not particularly limited.

Next, processing contents executed in each unit included in the monitoring apparatus 10 according to the present embodiment will be described in detail with reference to FIG. Various processes described later realized by the respective units shown in FIG. 3 are realized by the CPU 11 shown in FIG. 2 reading out the monitoring program stored in the auxiliary storage device 13 to the main storage device 12 and executing it. Is.
Moreover, in this embodiment, the case where the state of the bearing and speed-up gear (monitoring part) in the nacelle 3 of one component of the windmill 1 is assumed is assumed.

  First, the monitoring device 10 was measured by various sensors installed at predetermined locations together with measurement data measured by eight acceleration sensors attached to the bearing / speed increaser of the windmill 1 as a monitoring part. Measurement data, for example, wind speed, wind direction, generator speed, temperature, and the like are sequentially transmitted via a communication line in association with time information. These measurement data are sequentially accumulated in the measurement information storage unit 21 for each sensor.

  The data generation unit 22 unifies the sampling time of the measurement data sequentially stored in the measurement information storage unit 21 and calculates the diagnostic physical quantity from the measurement data of the eight sensors attached to the bearing / speed increaser that is the monitoring part. Then, a data file is generated for each natural frequency (step SA1 in FIG. 11).

  Subsequently, the class classification unit 24 integrates the data files to generate a diagnosis data file (step SA2 in FIG. 11), and for each data set associated with the same measurement time in the diagnosis data file A flag indicating the class classification is added (step SA3 in FIG. 11). The diagnosis data file to which the class classification flag is added is stored in the first storage unit 25 (step SA4 in FIG. 11).

  Next, the normal data extraction unit 27 extracts only normal data sets from the classified diagnostic data file stored in the first storage unit 25 to create a normal data file, which is stored in the second storage unit 28. Stored (step SA5 in FIG. 11).

  Next, the diagnosis setting unit 29 extracts all or part of the data set of the diagnostic data file in the first storage unit 25 to create a diagnostic data file for diagnosis, and the normality of the second storage unit 28 All or part of a data set of the data file is extracted and set as a normal data file for diagnosis (hereinafter referred to as “reference data file”) (step SA6 in FIG. 11).

When the “diagnostic data file” and the “reference data file” are set in this way, the index value calculation unit 30 calculates the state index value of the soundness (step SA7 in FIG. 11).
Hereinafter, the state index value calculation process will be described with reference to FIG.

[Data standardization]
First, the index value calculation unit 30 executes data normalization processing (step SB1 in FIG. 12).
For example, if the measurement time number of the reference data file set in the diagnosis setting unit 29 is i and the number of characteristic items is j, the reference data file forms a matrix of i rows and j columns. For example, if the number of data for one day is 1 minute interval, 1440 rows in 24 hours × 60 minutes = 1440 minutes. If there are 200 types of measured data item types, the data size is 200 columns.

The reason for standardizing the reference data is to treat characteristic values between different characteristic items (between measured physical quantities) fairly in statistical processing. Therefore, the characteristic value x _ij identified by each row and each column is normalized using the average value m _j and standard deviation σ _j calculated based on the following equations (1) and (2). The normalized value of the characteristic value x _ij is expressed as a sex standard value X _ij and is obtained by the following equation (3).
In the following explanation, as shown in FIG. 13, each data file of n rows and k columns is assumed.

Similarly, the index value calculation unit 30 normalizes the “diagnostic data file” by performing the same calculation as that of the reference data file. As the average value m _j and standard deviation σ _j used for normalization, the values of the “reference data file” calculated by the above formulas (1) and (2) are used. As a result, a characteristic standard value Y _ij obtained by standardizing each characteristic value y _ij of the “diagnosis data file” is calculated by the following equation (4).

  The index value calculation unit 30 reconstructs the respective data files by replacing the characteristic values of the “reference data file” and “diagnostic data file” with the normalized characteristic standard values.

[Calculation of correlation matrix]
Next, the index value calculation unit 30 calculates the correlation matrix R = (r _ij ) using the characteristic standard value X _ij of the reference data file (step SB2 in FIG. 12). The correlation matrix R is derived using the following equation (5). The correlation matrix is a k-th order matrix whose diagonal component is 1.

Here, a specific description for obtaining the correlation matrix will be given. When there are k (k columns) types of characteristic items j of the characteristic standard value _Xij of the reference data file, the number of correlation combinations is k × k. As an example, if the number of characteristic items j in the reference data file is k = 200 types (columns), there are 200 × 200 = 40000 correlation combinations, which simultaneously become the characteristics of a 200 × 200 regular matrix. The diagonal components of the regular matrix have the property of inevitably being 1 because they are the correlations between the same characteristic items. Further, the correlation coefficient other than the diagonal line is r _pq = r _qp , and its value is symmetrically equal across the diagonal line.

[Calculation of inverse matrix of correlation matrix]
Subsequently, the index value calculation unit 30 calculates the inverse matrix A = R ⁻¹ of the correlation matrix R of the reference data file using the following equation (6) (step SB3 in FIG. 12).

[Calculation of Mahalanobis distance]
Next, the Mahalanobis distance D ² _i (hereinafter referred to as “MD value”) using the correlation inverse matrix A of the reference data file obtained by the above equation (6) and each characteristic standard value Y _ij of the diagnosis data file after normalization. (Refer to step SB4 in FIG. 12). The MD value D ² _i is calculated using the following equation (7).

  Here, k is the number of characteristic items of the “diagnostic data file”, that is, the number of columns, and the MD value is calculated for each data set (for each row) of the “diagnostic data file”. For example, if the “diagnostic data file” is a data file for one day acquired at 1-minute intervals, the number of rows is 1440, and 1440 MD values are obtained. This means that the MD value, which is a soundness diagnostic index of the bearing / speed increaser, is sequentially calculated every measurement time.

Here, when calculating the MD value of a certain measurement time, the data line i (i is any one of 1 to n) corresponding to the certain measurement time of the standardized “diagnostic data file” is designated. , Y _i1 to Y _ik which are values of each column of the i row are substituted into the equation (7) for calculation. In the example of FIG. 6, when obtaining the MD value at the measurement time of 0:03, the characteristic standard values Y ₄₁ to Y _4k of the number of k columns in the i = 4th row in the data row are used. In this way, MD values are obtained for the number of measurement times of the “diagnostic data file”.

Here, it should be noted that in steps SB1 to SB4, m _j for normalization is determined depending on which class classification of the data set of the diagnosis data file is compared with the data set of the reference data file of which class classification. Although the correlation inverse matrix A at the time of σ _j and MD value calculation must be properly used, automatic calculation processing is possible by instructing in advance by programming.

The state index value calculated by the index value calculation unit 30 is output to the abnormality determination unit 31. The abnormality determination unit 31 compares each input MD value D ² _i with a preset threshold value (a value that can be arbitrarily set, for example, 3), and the MD value D ² _i is greater than the threshold value. Is also larger (step SA8 in FIG. 11). As a result, when the MD value D ² _i larger than the threshold is greater than or equal to a predetermined ratio, an abnormal signal is output assuming that the state of the windmill 1 is abnormal. Thereby, the notification unit 30 notifies the user of the abnormality of the windmill 1 (step SA9 in FIG. 11).

On the other hand, when the MD value D ² _i exceeding the threshold is equal to or less than the predetermined ratio in step SA8, it is determined that the state of the windmill 1 is normal, and the process returns to step SA1 in FIG. Repeat. As a result, as long as no abnormality is detected, the processing from step SA1 to step SA8 is repeated, and the state of the windmill 1 is monitored every predetermined period.

  Further, the state index value calculated by the index value calculation unit 30 is displayed on the display device as a diagnosis result as shown in FIG. FIG. 14 is a diagram illustrating an example of a diagnosis result. The horizontal axis shows the measurement time, and the vertical axis shows the MD value.

  As described above, according to the wind turbine monitoring apparatus and method and the program according to the present embodiment, the soundness of the wind turbine is compared with the “reference data file” that is an actually measured value classified into the class. Since it can be quantitatively determined by the MD value, it is possible to realize appropriate evaluation instead of qualitative evaluation based on experience and knowledge.

  In the above-described embodiment, normal data stored in the second storage unit 28 is created using data stored in the first storage unit 25, but normal data stored in the second storage unit 28 is used. The data is not limited to this example. For example, normal data calculated by predetermined simulation software or the like may be used.

  In the above embodiment, as shown in FIG. 11, the processing from step SA1 to step SA9 has been described as a series of processing. However, until the normal data file generation processing, that is, data is stored in the first storage unit 25. The processing to store and the processing to store data in the second storage unit 28 are handled as pre-processing necessary for monitoring the state of the windmill, and the processing after setting the reference data file and the diagnostic data file, specifically The processing from step SA6 to step SA9 in FIG. 11 may be handled as the main processing for evaluating the state of the windmill 1. The preprocessing and the main processing may have a time difference or may be realized by different computers.

[Second Embodiment]
Next, a second embodiment of the present invention will be described with reference to the drawings.
In the monitoring apparatus for the windmill 1 according to the first embodiment described above, the case of monitoring the speed increaser of the windmill has been described, but in this embodiment, the case of monitoring the load and strength applied to the entire windmill structure is described. To do.
In general, a windmill receives wind with windmill blades and generates mechanical energy by rotating the windmill blades, and converts the mechanical energy into electrical energy. At this time, the windmill structure is subjected to a load such as wind. Yes. In the initial design, the strength that satisfies the allowable load tolerance of each structural part is maintained, but the strength deterioration progresses due to wind and rain, corrosion, and aging.

  Therefore, in the present embodiment, the strain change of the windmill 1 is measured by an electric strain gauge or an optical fiber, and the state related to the soundness of the windmill 1 (cumulative load and its strength deterioration degree) is monitored based on this measurement data. Do. Each process according to the present embodiment is the same as the process performed in the first embodiment described above, and only the characteristic items classified as “characteristics” are changed. Therefore, description of each process is omitted.

  As described above, as the characteristic item according to the first embodiment described above, the MD value is obtained by using the measurement data measured by the strain sensor attached to the predetermined portion of the windmill 1 and comparing them with the normal data. calculate. Thereby, it becomes possible to evaluate the soundness of a windmill using MD value which is an objective statistical index value.

In the above-described first or second embodiment, the case where monitoring is performed for one monitoring part has been described as an example. However, the monitoring apparatus of the present invention has a plurality of monitoring parts set. It is also applicable when
In this case, in the diagnostic data file shown in FIG. 6, the number of characteristic items classified as “characteristics” increases according to the monitored part.
And about calculation of the said MD value, the state of a some monitoring site | part is integrated and it calculates as a soundness index value of a windmill with one type of MD value.

[Third Embodiment]
Next, a monitoring apparatus according to the third embodiment of the present invention will be described.
In the monitoring device according to the first or second embodiment described above, when an abnormality is detected by the abnormality determination unit 31, which characteristic item is involved in the abnormal state or is not involved. There is a need to identify quantitatively. The present embodiment has been proposed in view of the request.

  In the present embodiment, in the monitoring apparatus according to the first or second embodiment described above, when an abnormality is determined by the abnormality determination unit 31, as shown in FIG. The unit (factor analysis means) 50 is further provided. Hereinafter, regarding the monitoring device of the present embodiment, description of points that are common to the first embodiment will be omitted, and “factor / effect analysis” that is a different point will be mainly described.

  “Factor effect analysis” of the present embodiment is, for example, a quantitative analysis of which characteristic items have an influence on the length of MD values, assuming that there are 200 characteristic items constituting diagnostic data. As shown in FIG. 16, rank-up display is performed according to the magnitude of the SN ratio gain, which is the output value of the factor effect analysis.

  A method for calculating the SN ratio gain of each characteristic item, which is an index value of the factor effect, will be described with reference to FIG. For convenience of explanation, FIG. 17 shows an example in which there are five types of characteristic items. Each of the five types of characteristic items is assigned to a two-level orthogonal table of “use ○” and “not use ×”, and the five types. The MD value calculation process (see FIG. 12) is performed in accordance with 12 combination conditions (row number of orthogonal table) of “use ○” and “not use ×” of the characteristic items of 5 The MD value of each data set in the diagnostic data file consisting of individual characteristic items is calculated.

For example, when the measurement time interval determined to be abnormal in the diagnosis data file by the abnormality determination unit 31 is 2 minutes (two data lines), 12 characteristic items “ The MD value is calculated for each case of “use ○” and “not use ×”. As a result, 12 types × 2 MD values D ² (1) and D ² (2) are added as calculation results to the right end of the orthogonal table of FIG. From these MD values D ² (1) and D ² (2), the SN ratios η of 12 combinations are calculated using the following equation (8).

Here, n is the number of data (number of rows) of the factor effect target, and n = 2 in this example.
The results of ₁₂ types of SN ratios η _{1 to} η 12 calculated by Expression (8) are added to the right end of the orthogonal table as shown in FIG. This completes the preparation for calculating the S / N ratio gain for each characteristic item (characteristic item 1 to characteristic item 5) of the factor effect.
Specifically, considering the characteristic items of the wind turbine 1, for example, “characteristic item 1 = low speed meshing primary”, “characteristic item = medium speed overall value”, “characteristic item = wind speed turbulence”, It may be considered that “characteristic item = wind direction deviation”, “characteristic item = wind turbine blade center distortion value”, and the like.

  Hereinafter, the calculation of the S / N ratio and the factor analysis using the S / N ratio performed by the factor analysis unit 50 will be described in detail.

[Calculation of SN ratio]
As shown in the following equation, the SN ratio gains η _{c1 to} η _{c5 of the} characteristic items 1 to 5 obtained by the factor effect analysis in the abnormality diagnosis data file are the combinations (◯) of the characteristic items. This is the difference between the SN ratio and the SN ratio when the combination was not used (x).

  As a result, it can be determined that the characteristic value having a larger SN ratio gain is more likely to be involved in the abnormality. As the value of ηc (◯, x) to be substituted into the above equation, the average values of the SN ratio values in the auxiliary table list of FIG. 18 are used by using the calculated values in the orthogonal table of FIG.

[Factor analysis]
The factor analysis unit 50 can contribute to the cause of abnormality from a plurality of characteristic items in the diagnosis data file by quantifying the contribution rate of the factor effect to each characteristic item based on the gain of Expression (9). A characteristic item with high characteristics is selected, and the result of this factor effect is output to the notification unit 32. Thereby, the notification part 32 notifies the user of the analysis result of the factor analysis part 50.

  FIG. 16 is a diagram showing an example of a display screen showing the factor analysis results. It shows that a characteristic item having a large gain value can be a cause of occurrence of an abnormality detected this time.

  As described above, according to the monitoring apparatus according to the present embodiment, when a wind turbine state abnormality is detected, a characteristic item that is likely to cause the abnormality is analyzed, and the analysis result is analyzed by the user. Can be notified. Thereby, it is possible to promptly take an appropriate response to the occurrence of an abnormality.

  The factor analysis result obtained by the factor analysis unit 50 described above may be adopted for maintenance or after-sales service. As described above, by using the factor analysis result secondarily, it is possible to find a sign of abnormality, and thus it is possible to prevent occurrence of a serious abnormality such as replacement of a device in advance. As a result, it is possible to prevent a reduction in the operating efficiency of the windmill due to the occurrence of an abnormality, and it is possible to reduce maintenance costs.

[Fourth Embodiment]
Next, a wind turbine monitoring device according to a fourth embodiment of the present invention will be described.
In each of the embodiments described above, a normal data file is used as the reference data file. In the present embodiment, instead of this, the above-described MD value is calculated by using an abnormal data file as the reference data file. And when this MD value is smaller than a predetermined threshold value, it determines with abnormality having generate | occur | produced.

  As described above, by using the abnormal data file as the reference data file, for example, by dividing the various abnormal / failure states in the unit space and calculating the MD value, the factor as in the third embodiment described above is obtained. Without performing the effect analysis, it is possible to easily specify which characteristic item diagnosis data indicates what kind of abnormality.

[Application example]
Next, a monitoring system according to an embodiment of the present invention will be described.
The monitoring device according to each of the above-described embodiments performs state monitoring regarding each part of one windmill. The monitoring system according to the present embodiment monitors the state of some or all of the wind turbines 1 in a wind farm in which a plurality of wind turbines are installed.

The monitoring system of the present embodiment includes any one of the monitoring devices according to the first to fourth embodiments described above, and the state index value of each wind turbine obtained by each of these monitoring devices, and the plurality of operations. Based on the performance, the state of a plurality of wind turbines to be monitored is monitored.
Specifically, the monitoring system acquires the information of the monitoring result such as the state index value from each monitoring device via a communication network such as a wireless communication network, and outputs the total output power output from the wind turbine to be monitored. Get about the amount via the network.

  Then, the monitoring system calculates the MD values of these state index values by using the state index values acquired from the respective monitoring devices as parameters of the MT method. And the state of the windmill in a wind farm is determined by comparing this MD value with the preset threshold value.

  As described above, by monitoring the operating performance of the wind farm with the statistic of the state index value in units of wind turbines, a wind turbine that exhibits a characteristic value different from the others among a plurality of wind turbines constituting the wind farm. Can be determined. Thus, instead of specifying an abnormal windmill based on a predetermined threshold that is uniquely determined, by specifying a windmill showing characteristics different from others as a target for a plurality of windmills, It is possible to comprehensively determine wind turbine abnormalities from a wide range of viewpoints for various factors. Thereby, the monitoring precision of a windmill can be improved.

  Note that the monitoring system described above does not necessarily require a plurality of computers. For example, after calculating the state index values of a plurality of wind turbines in one computer system in order, these state index values are used. It is also possible to determine abnormality of the entire windmill. That is, by installing software for realizing the above functions and executing the software by the CPU, the above functions may be realized by a single device.

  As mentioned above, although embodiment of this invention was explained in full detail with reference to drawings, the specific structure is not restricted to this embodiment, The design change etc. of the range which does not deviate from the summary of this invention are included.

It is the figure which showed the whole schematic structure of the windmill. It is the figure which showed schematic structure of the monitoring apparatus which concerns on the 1st Embodiment of this invention. It is the functional block diagram which expanded and showed the function of the monitoring apparatus which concerns on the 1st Embodiment of this invention. It is a figure for demonstrating the process which calculates a diagnostic physical quantity from the measurement data acquired by each acceleration sensor at the time of giving the case of diagnosing a bearing and a gearbox as an example. It is the figure which showed an example of the new data file in which the diagnostic physical quantity was matched and stored with the channel and the natural frequency. It is the figure which showed an example of the diagnostic data file. It is the figure which showed an example of class classification. It is the figure which showed the other example of class classification. It is the figure which showed an example of the normal definition range. It is the figure which showed the other example of the normal definition range. It is the flowchart which showed the process sequence in the 1st Embodiment of this invention. It is the flowchart which showed the procedure of the calculation process of a state index value. It is explanatory drawing for demonstrating each data used by the calculation process of Mahalanobis distance. It is the figure which showed an example of the evaluation result of a state index value. It is the functional block diagram which expanded and showed the function of the monitoring apparatus which concerns on the 3rd Embodiment of this invention. It is the figure which showed an example of the factor analysis result. It is the figure which showed an example of the 2-level orthogonal table used in the 3rd Embodiment of this invention. FIG. 18 is a diagram in which MD values and SN ratio gains are added to the two-level orthogonal table shown in FIG. 17.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Windmill 3 Nacelle 4 Rotor head 5 Windmill blade 10 Windmill monitoring device 11 CPU
12 main storage device 13 auxiliary storage device 14 input device 15 output device 16 communication device 21 measurement information storage unit 22 data generation unit 23 class classification definition unit 24 class classification unit 25 first storage unit 26 normal data condition definition unit 27 normal data extraction Unit 28 second storage unit 29 diagnosis setting unit 30 index calculation unit 31 abnormality determination unit 32 notification unit 50 factor analysis unit

Claims (11)

A windmill monitoring device that monitors the state of the windmill using characteristic values created based on measurement data measured by a plurality of sensors provided on the windmill,
When a plurality of characteristic values associated with the measurement time are stored for each characteristic item and the characteristic values associated with the same measurement time are set as one data set, the data set includes predetermined characteristic items. First storage means stored with identification information indicating a class classification determined according to the characteristic value of
When a plurality of characteristic values associated with the measurement time are stored for each characteristic item and the characteristic values associated with the same measurement time are set as one data set, the data set includes predetermined characteristic items. Identification information indicating a class classification determined according to the characteristic value of the data set is assigned, and the characteristic value of the specific characteristic item constituting the data set belongs to a predetermined reference range defined in advance Second storage means,
A plurality of the data sets used for diagnosis extracted from the first storage means and set, and a plurality of the data sets used for diagnosis extracted from the second storage means and set,
An index value for calculating a state index value representing the state of the windmill using a statistical calculation method based on the data set of the diagnosis data file and the data set of the reference data file set by the diagnosis setting unit A calculation means;
Evaluation means for evaluating the state of the windmill based on the state index value calculated by the index value calculation means;
A wind turbine monitoring device comprising: notification means for notifying an evaluation result by the evaluation means.   The characteristic items are broadly classified into three categories: an environmental category related to the environment surrounding the windmill, a performance category related to the performance and power generation conditions of the windmill operation, and a characteristic category related to diagnosis of the operating state relating to various monitoring parts set in the windmill. The wind turbine monitoring device according to claim 1.   The wind turbine monitoring device according to claim 2, wherein the class classification is determined according to a characteristic value of a predetermined characteristic item classified into at least one of the environmental classification and the performance classification.   The plurality of data sets stored in the second storage means are preset with characteristic values related to the specific characteristic item among a plurality of data sets generated from the plurality of measurement data collected from the windmill. The wind turbine monitoring device according to any one of claims 1 to 3, wherein only the data set belonging to the reference range is extracted.   The wind turbine monitoring apparatus according to claim 1, wherein the reference range is set for each class classification.   The index value calculation means obtains the characteristic distribution of the reference data set by the diagnosis setting means, obtains the characteristic distribution of the diagnosis data, and quantitatively obtains the distance at which the characteristic distributions deviate from each other. The wind turbine monitoring device according to any one of claims 1 to 5, wherein the state index value is calculated.   The wind turbine monitoring device according to claim 6, wherein the state index value calculated by the index value calculating means is a Mahalanobis distance calculated using a Mahalanobis Taguchi method.   The wind turbine monitoring device according to any one of claims 1 to 7, further comprising factor analysis means for performing a factor analysis of an abnormality when the evaluation means evaluates that an abnormality has occurred. A wind turbine group monitoring system for monitoring the state of a part or the whole of a wind farm having a plurality of wind turbines,
A wind turbine monitoring device according to any one of claims 1 to 8, comprising: a state index value of each wind turbine obtained by the wind turbine monitoring device and a part or all of the operating performance of the wind farm. Wind turbine group monitoring system that monitors the state of part or the entire wind farm. A windmill monitoring method for monitoring a state of the windmill using characteristic values created based on measurement data measured by a plurality of sensors provided on the windmill,
A process of creating a diagnostic data file in which a plurality of characteristic values associated with measurement time are stored for each characteristic item;
In the diagnosis data file, when the characteristic values associated with the same measurement time are set as one data set, a class classification determined according to the characteristic value of a predetermined characteristic item is indicated in the data set. A process of providing identification information;
A process of creating a reference data file in which the characteristic values related to a specific characteristic item belong to a predetermined reference range defined in advance and the characteristic values of each characteristic item are associated with the measurement time;
In the reference data file, when the characteristic values associated with the same measurement time are set as one data set, an identification indicating a class classification determined according to the characteristic value of a predetermined characteristic item in the data set The process of giving information,
Extracting and setting a plurality of data sets used for diagnosis from the diagnosis data file, and extracting and setting a plurality of data sets used for diagnosis from the reference data file; and
Based on the data set of the diagnosis data file set and the data set of the reference data file, a process of calculating a state index value representing the state of the windmill using a statistical calculation method;
A process of evaluating the state of the windmill based on the state index value;
A wind turbine monitoring method comprising: notifying a result of the evaluation. A windmill monitoring program used to monitor the state of the windmill, using characteristic values created based on measurement data measured by a plurality of sensors provided on the windmill,
A process of creating a diagnostic data file in which a plurality of characteristic values associated with the measurement time are stored for each characteristic item;
In the diagnosis data file, when the characteristic values associated with the same measurement time are set as one data set, a class classification determined according to the characteristic value of a predetermined characteristic item is indicated in the data set. A process of providing identification information;
A process for creating a reference data file in which the characteristic value related to a specific characteristic item belongs to a predetermined reference range defined in advance and the characteristic value of each characteristic item is associated with the measurement time;
In the reference data file, when the characteristic values associated with the same measurement time are set as one data set, an identification indicating a class classification determined according to the characteristic value of a predetermined characteristic item in the data set A process of giving information,
A process of extracting and setting a plurality of the data sets used for diagnosis from the diagnosis data file, and extracting and setting a plurality of the data sets used for the diagnosis from the reference data file;
Based on the set data set of the diagnostic data file and the data set of the reference data file, a process of calculating a state index value representing the state of the windmill using a statistical calculation method;
A process for evaluating the state of the windmill based on the state index value;
A windmill monitoring program for causing a computer to execute processing for notifying the result of the evaluation.

Images