Abnormal early warning method and system for wind turbines based on artificial intelligence

Abstract

The present invention relates to the field of data processing technology, and in particular to an abnormal warning method and system for wind turbines based on artificial intelligence. The method comprises the steps of: constructing an isolated tree using wind turbine operation data; calculating the absolute value of the average layer difference between a data point and all time-series neighboring points in the isolated tree, calculating the time series mutation degree of the data point according to the absolute value of the difference and the standard deviation of the layers of all time series neighboring points, obtaining the spatial neighboring points of any data point in any dimension, calculating the density distribution difference of the data point according to the Euclidean distance between the data point and its spatial neighboring points, weighting the abnormal score of the multidimensional data point according to the time series mutation degree and density distribution difference of the data point in any dimension; and issuing an early warning according to the weighted abnormal score. The present invention has more accurate abnormal detection results for wind turbine operation data, and further more accurate early warnings for staff.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an abnormal early warning method and system for a wind turbine generator set based on artificial intelligence. Background Art

[0002] As the global demand for renewable energy continues to increase, wind energy, as a clean and sustainable form of energy, has occupied an important position in electricity production. As a key device for converting wind energy into electrical energy, the stability and reliability of wind turbine operation directly affect the power generation efficiency and economic benefits of wind farms. Wind turbines drive rotors through wind power, and the rotors are connected to generators to convert mechanical energy into electrical energy. In this process, parameters such as output electrical energy, voltage and current are key indicators for evaluating the operating status of wind turbines. Abnormal fluctuations in any of the indicators may indicate potential failures in the wind turbine, such as mechanical failures, failure of electronic components or changes in the external environment. Therefore, it is necessary to monitor multiple parameters during the operation of wind turbines and provide timely warnings.

[0003] The isolation forest algorithm is an algorithm for anomaly detection of multi-dimensional data. It uses multi-dimensional data to construct an isolation tree, and calculates the degree of abnormality of any data point based on the isolation tree, thereby completing anomaly detection. For example, the Chinese patent application document with publication number CN115496163A discloses a method for monitoring abnormal energy consumption of buildings based on K-Means++, BP neural network and isolation forest. The patent application document points out that an anomaly score is obtained by constructing an isolation tree using energy consumption data of the same category. When calculating the anomaly score, the traditional isolation forest algorithm can only obtain global abnormal data, but is not sensitive to local abnormal data. For the operating data of wind turbines, local anomalies may be caused by the influence of external factors, and the traditional isolation forest algorithm is not sensitive to it, resulting in the local abnormal data being detected as normal data, which leads to inaccurate detection results. Summary of the invention

[0004] The purpose of the present invention is to provide an artificial intelligence-based wind turbine abnormality warning method and system to solve the technical problem of inaccurate warning results for wind turbines.

[0005] In a first aspect, the present invention provides an abnormal warning method for wind turbines based on artificial intelligence, which adopts the following technical solutions:

[0006] Acquire multi-dimensional data points of wind turbine operation data, wherein the dimensions include: output electric energy, voltage, and current; and construct an isolation tree using an isolation forest algorithm based on the multi-dimensional data points;

[0007] Get any data point of any dimension time series neighboring points and the number of layers in the isolated tree, calculate the absolute value of the difference between the data point and the average number of layers of all time series neighboring points; calculate the time series mutation degree of the data point according to the absolute value of the difference and the standard deviation of the number of layers of all time series neighboring points, wherein the time series mutation degree is positively correlated with the absolute value and negatively correlated with the standard deviation;

[0008] Get any data point of any dimension spatially adjacent points, calculating the local density of the data point according to the Euclidean distance between the data point and its spatially adjacent points, wherein the local density is negatively correlated with the Euclidean distance; calculating the density distribution difference of the data point according to the difference between the average levels of the local density of the data point and its spatially adjacent points, wherein the density distribution difference is positively correlated with the difference;

[0009] According to the time series mutation degree and density distribution difference of the data points in any dimension, the anomaly scores of the multidimensional data points are weighted to obtain the weighted anomaly scores, which are positively correlated with the time series mutation degree and density distribution difference; and an early warning is issued according to the weighted anomaly scores.

[0010] For the abnormal warning system of wind turbines, the isolation forest algorithm can quickly and accurately detect abnormal changes in multi-dimensional data such as generator output power, voltage, current, etc., thereby timely warning of possible equipment failures or system operation abnormalities.

[0011] Preferably, the method of constructing an isolation tree using an isolation forest algorithm according to multidimensional data points comprises the steps of:

[0012] For any isolated tree, the data points collected in the same dimension are continuously segmented as segmentation samples of the isolation forest to obtain the isolated tree of the dimension, where multiple isolated trees are constructed in any dimension.

[0013] Preferably, the method for acquiring the time sequence neighboring points includes:

[0014] In any dimension, for any data point, get the closest time to the data point. data points as the temporal neighbors of the data point.

[0015] Preferably, the degree of the timing mutation satisfies the expression:

[0016] ;

[0017] In the formula, Indicates The dimension The degree of temporal mutation of each data point; Indicates The total number of isolated trees constructed by dimensions; represents the isolated tree ordinal number; Indicates the total number of temporal neighboring points of the preset data point; Indicates the ordinal number of the temporal neighboring points; Indicates The data point in The number of levels in an isolated tree; Indicates The data point The temporal neighbors are in The number of levels in an isolated tree; Indicates All the temporal neighbors of a data point are in The mean number of layers in an isolated tree; Represents the standard normalization function.

[0018] By calculating the degree of time series mutation, the abnormality of each data point in time series can be quantitatively evaluated, which helps to correct the abnormal score results of multi-dimensional data points in the future, making the abnormal score results more accurate, thereby improving the accuracy of anomaly detection results and early warning results.

[0019] Preferably, the method for acquiring the spatial neighboring points includes:

[0020] In any dimension, for any data point, get the closest Euclidean distance to the data point data points as the spatial neighbors of the data point.

[0021] Preferably, the local density satisfies the expression:

[0022] ;

[0023] In the formula, Indicates The dimension The local density of data points; Indicates the total number of spatial neighboring points of the preset data point; Indicates the ordinal number of spatial neighboring points; Indicates data point and its The Euclidean distance between spatially adjacent points.

[0024] By calculating the average level of the Euclidean distance between any data point and its spatial neighbors, the local density of the data point can be better evaluated, avoiding the problem of inaccurate local density calculation results of the data point due to the large difference between the Euclidean distance between a few data points and the data point and the average level.

[0025] Preferably, the density distribution difference satisfies the expression:

[0026] ;

[0027] In the formula, Indicates The dimension The density distribution difference of data points; Indicates The dimension The local density of data points; Indicates The dimension data points The local density mean of spatially adjacent points; Expressed as a natural constant The exponential function with base ; Represents the absolute value symbol.

[0028] The density distribution difference of a data point is quantified by calculating the difference between the local density of the data point and the local density mean of its spatial neighbors. The calculation result is more accurate by considering the difference between the local density mean of multiple data points.

[0029] Preferably, the weighted anomaly score satisfies the expression:

[0030] ;

[0031] In the formula, Indicates Weighted anomaly scores for multidimensional data points; Indicates Anomaly scores for multidimensional data points; Indicates the total number of dimensions; Indicates the dimension ordinal number; Indicates The total number of data points in each dimension; Indicates the ordinal number of the data point; Indicates The dimension The degree of temporal mutation of each data point; Indicates The dimension The density distribution difference of data points; Represents the standard normalization function.

[0032] By weighting the anomaly score by the degree of time series mutation and density distribution difference, a weighted anomaly score is obtained. This comprehensively evaluates the data characteristics of different dimensions, making the anomaly detection results more comprehensive and reliable. Through timely anomaly warnings, staff can take necessary measures more promptly to ensure the stable operation and long-term performance of the system.

[0033] Preferably, the step of providing an early warning based on the weighted anomaly score comprises the following steps:

[0034] Preset abnormal threshold , if the average weighted anomaly score of all multidimensional data points is greater than the preset anomaly threshold , the system will issue an early warning.

[0035] In a second aspect, the present invention provides an abnormal warning method and system for wind turbines based on artificial intelligence, which adopts the following technical solutions:

[0036] The wind turbine abnormality warning system based on artificial intelligence includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned wind turbine abnormality warning method based on artificial intelligence is implemented.

[0037] By adopting the above technical solution, the above-mentioned wind turbine abnormality warning method is generated into a computer program and stored in a memory to be loaded and executed by a processor, so that a terminal device is manufactured according to the memory and the processor for easy use.

[0038] The present invention has the following technical effects:

[0039] The present invention calculates the time series mutation degree and density distribution difference of any data point in any dimension, and weights the anomaly score of the multidimensional data point according to the time series mutation degree and density distribution difference of data points in all dimensions of any multidimensional data point to obtain a weighted anomaly score. Compared with the traditional isolation forest algorithm, the present invention fully considers the local anomaly of the wind turbine operation data by utilizing the time series mutation degree and density distribution difference, thereby making the anomaly detection result more accurate.

[0040] Furthermore, the present invention warns the staff according to the average level of the abnormal scores of all multi-dimensional data points, thereby avoiding false warnings to the staff due to individual extremely abnormal points. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts.

[0042] Figure 1 It is a flow chart of a method in an abnormal warning method for a wind turbine generator system based on artificial intelligence in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0044] It should be understood that when the terms "first", "second", etc. are used in the claims, descriptions, and drawings of the present invention, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the description and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their collections.

[0045] The embodiment of the present invention discloses an abnormal warning method for a wind turbine generator system based on artificial intelligence, referring to Figure 1 , comprising steps S1 to S5:

[0046] S1: Obtain wind turbine operation data and construct an isolation tree;

[0047] It should be noted that the operating data of wind turbines refers to various parameters during the operation of wind turbines, such as output power, current and voltage. By monitoring the operating data of wind turbines, the power generation efficiency and overall performance of wind turbines can be evaluated, and further, potential faults and abnormal conditions can be discovered in a timely manner.

[0048] In one embodiment, the wind turbine operating data includes but is not limited to: the output power, current and voltage of the wind turbine. The output power of the wind turbine is monitored in real time by installing a power metering device (such as an electric meter) on the wind turbine; the voltage and current are collected by arranging voltage sensors and current sensors.

[0049] It should be noted that the present invention does not limit the frequency of collecting wind turbine operating data, and the implementation personnel can adjust it according to the actual situation, for example .

[0050] In one embodiment, data points of all dimensions collected at the same time are combined into a multi-dimensional data point, and then all the multi-dimensional data points are used as segmentation samples of the isolation forest algorithm to construct an isolation tree.

[0051] It should be noted that when constructing an isolation tree, the dimensions of each layer in the same isolation tree are the same, that is, the same isolation tree is constructed using data of the same dimension. In the isolation forest algorithm, each isolation tree is constructed independently, and the segmentation threshold is selected randomly each time to ensure that the segmentation is random. When the same isolation tree is constructed using data of the same dimension, the distribution of the data can be fully utilized. This random selection of the segmentation threshold allows the isolation tree to effectively distinguish normal data points from abnormal data points, because abnormal points usually require fewer steps to be isolated, while normal data points require more steps.

[0052] At this point, the construction of the isolation tree is completed.

[0053] S2: Calculate the temporal mutation degree of a data point according to the position of the temporal neighboring points of data in any dimension in the isolation tree;

[0054] It should be noted that the operating data of wind turbines will change over time. By monitoring these data, the operating status of the system can be tracked and anomalies can be discovered in a timely manner. Assuming that the output power of the wind turbine suddenly increases or decreases significantly at a certain moment, this means that the system components may fail or operate abnormally. This mutation is only a local mutation, and the isolation forest is an algorithm for global anomaly detection and is less affected by local anomalies. Therefore, it is necessary to calculate the degree of local mutation of any data point to increase the local impact of the data point.

[0055] In one embodiment, in any dimension, for any data point, the closest time series to it is obtained. data points, as the data point Time series neighboring points, obtain the number of layers of the data point and all its time series neighboring points in the isolation tree. Among them, is the total number of preset time sequence neighboring points, which is not specifically limited in the present invention and can be adjusted by the implementer according to the specific situation. For example, .

[0056] The temporal mutation degree of any data point is calculated based on the number of layers of the data point and its temporal neighboring points in the isolation tree. The specific calculation formula is as follows:

[0057]

[0058] In the formula, Indicates The dimension The degree of temporal mutation of each data point; Indicates The total number of isolated trees constructed by dimensions; represents the isolated tree ordinal number; Indicates the total number of temporal neighboring points of the preset data point; Indicates the ordinal number of the temporal neighboring points; Indicates The data point in The number of levels in an isolated tree; Indicates The data point The temporal neighbors are in The number of levels in an isolated tree; Indicates All the temporal neighbors of a data point are in The mean number of layers in an isolated tree; Represents the standard normalization function.

[0059] in, Indicates the difference between the number of layers of the data point and the average number of layers of all its time series neighboring points. The larger the value, the greater the difference, that is, the greater the degree of time series mutation of the data point; Indicates the closeness of the number of layers of all the temporal neighboring points of the data point. The closer the value is, the closer the number of layers of all the temporal neighboring points of the data point is. At this time, if the difference between the number of layers of the data point and the average number of layers of all temporal neighboring points is, the greater the degree of temporal mutation of the data point is.

[0060] At this point, the calculation of the time series mutation degree of any data point is completed.

[0061] S3: Calculate the density distribution difference of any data point based on the Euclidean distance between the spatial neighboring points of the data point and the data point;

[0062] It should be noted that in the process of continuously segmenting all data points using the isolation forest algorithm, the data points that need to be segmented are only segmented according to the segmentation threshold size, and the density distribution between the data points is not taken into account. The anomaly scores of data points under different density distributions are different. The traditional isolation forest algorithm cannot identify such anomalies well. Therefore, it is necessary to calculate the density distribution difference of any data point based on the Euclidean distance between the spatial neighboring points of the data point and the data point.

[0063] In one embodiment, for any data point in any dimension, the Euclidean distance between the data point and all other data points is obtained, and the data point with the closest Euclidean distance is selected. data points as the spatial neighboring points of the data point. is the total number of preset spatial neighboring points, which is not specifically limited in the present invention and can be adjusted by the implementer according to the specific situation. For example, .

[0064] The local density of any data point is calculated based on the Euclidean distance between the data point and its spatial neighbors. The calculation formula is as follows:

[0065] ;

[0066] In the formula, Indicates The dimension The local density of data points; Indicates the total number of spatial neighboring points of the preset data point; Indicates the ordinal number of spatial neighboring points; Indicates data point and its The Euclidean distance between spatially adjacent points.

[0067] in, It represents the average level of the Euclidean distance between the data point and its spatial neighbors. If its value is smaller, it means that the data point is closer to its spatial neighbors, and the local density of the data point is greater.

[0068] The density distribution difference of any data point is calculated based on the local density between the data point and its spatial neighbors. The calculation formula is as follows:

[0069] ;

[0070] In the formula, Indicates The dimension The density distribution difference of data points; Indicates The dimension The local density of data points; Indicates The dimension data points The local density mean of spatially adjacent points; Expressed as a natural constant The exponential function with base ; Represents the absolute value symbol.

[0071] The density distribution difference of the data point is calculated by calculating the difference between the local density of the data point and the average level of the local density of its spatially adjacent points. The larger the value, the greater the density distribution difference of the data point. Conversely, the smaller the density distribution difference of the data point.

[0072] At this point, the density distribution difference calculation of any data point is completed.

[0073] S4: weight the anomaly scores of multidimensional data points according to the temporal mutation degree and density distribution difference of data points in all dimensions to obtain weighted anomaly scores of multidimensional data points;

[0074] It should be noted that the anomaly score obtained by using the traditional isolation forest algorithm is calculated only based on the segmentation result of the isolated tree, but the local changes of the isolated tree cannot be taken into account in the process of segmenting all data points using the isolated tree. Therefore, the present invention needs to weight the anomaly score of the multidimensional data point according to the temporal mutation degree and density distribution difference of the data points of all dimensions, so as to obtain the weighted anomaly score of the multidimensional data point.

[0075] In one embodiment, the abnormal scores of all multidimensional data points are obtained. The method of obtaining the abnormal scores of data points is a prior art of isolation forest, which will not be described in detail here.

[0076] The specific calculation formula for weighting the anomaly score of multidimensional data points is as follows:

[0077] ;

[0078] In the formula, Indicates Weighted anomaly scores for multidimensional data points; Indicates Anomaly scores for multidimensional data points; Indicates the total number of dimensions; Indicates the dimension ordinal number; Indicates The total number of data points in each dimension; Indicates the ordinal number of the data point; Indicates The dimension The degree of temporal mutation of each data point; Indicates The dimension The density distribution difference of data points; Represents the standard normalization function.

[0079] Among them, if the time series mutation degree of the data point is greater and the density distribution difference is greater, the anomaly score of the data point is higher, otherwise, the anomaly score of the data point is lower.

[0080] At this point, the weighted anomaly score calculation for any data point is completed.

[0081] S5: Early warning based on weighted anomaly scores of multidimensional data points;

[0082] In one embodiment, the preset experience threshold , if the mean of the weighted anomaly scores of all multidimensional data points is greater than or equal to the preset threshold , an early warning will be issued to the staff so that the wind turbine equipment can be repaired in time. It is a preset abnormal threshold, which is not specifically limited in the present invention and can be adjusted by the implementer according to the actual situation. For example, .

[0083] At this point, early warning based on weighted anomaly scores of multi-dimensional data points is completed.

[0084] Through the above steps, the abnormal detection of wind turbine operating data is completed, and timely warnings are issued based on the abnormal detection results.

[0085] An embodiment of the present invention also discloses an artificial intelligence-based wind turbine abnormality warning system, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an artificial intelligence-based wind turbine abnormality warning method according to the present invention is implemented.

[0086] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.

[0087] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory, a dynamic random access memory, a static random access memory, an enhanced dynamic random access memory, a high bandwidth memory, a hybrid storage cube, etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of the device or accessible or connectable to the device.

[0088] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

[0089] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. The abnormal warning method of wind turbines based on artificial intelligence is characterized by: Includes steps: Obtain multi-dimensional data points of wind turbine operation data, including output power, voltage, and current; and construct an isolation tree using the isolation forest algorithm based on the multi-dimensional data points; Get any data point of any dimension time series neighboring points and their layers in the isolated tree; calculate the time series mutation degree of the data point to satisfy the expression: ； In the formula, Indicates The dimension The degree of temporal mutation of each data point; Indicates The total number of isolated trees constructed by dimensions; represents the isolated tree ordinal number; Indicates the total number of temporal neighboring points of the preset data point; Indicates the ordinal number of the temporal neighboring points; Indicates The data point in The number of levels in an isolated tree; Indicates The data point The temporal neighbors are in The number of levels in an isolated tree; Indicates All the temporal neighbors of a data point are in The mean number of layers in an isolated tree; represents the standard normalization function; Get any data point of any dimension The spatial neighboring points are obtained by: in any dimension, for any data point, obtaining the Euclidean distance between the data point and all other data points, and taking the point with the closest Euclidean distance to it. data points as the spatial neighboring points of the data point; calculating the local density of the data point according to the Euclidean distance between the data point and its spatial neighboring points, wherein the local density is negatively correlated with the Euclidean distance; calculating the density distribution difference of the data point according to the difference between the local density of the data point and the average value of the local density of its spatial neighboring points, wherein the density distribution difference is positively correlated with the difference; According to the time series mutation degree and density distribution difference of the data points in any dimension, the anomaly scores of the multidimensional data points are weighted to obtain the weighted anomaly scores, which are positively correlated with the time series mutation degree and density distribution difference; and an early warning is issued according to the weighted anomaly scores.

2. The abnormal warning method for wind turbines based on artificial intelligence according to claim 1 is characterized in that: The method of constructing an isolated tree using an isolation forest algorithm according to multidimensional data points comprises the following steps: For any isolated tree, the data points collected in the same dimension are continuously segmented as segmentation samples of the isolation forest to obtain the isolated tree of the dimension, where multiple isolated trees are constructed in any dimension.

3. The abnormal warning method for wind turbines based on artificial intelligence according to claim 1 is characterized in that: The method for acquiring the time sequence neighboring points includes: In any dimension, for any data point, get the closest time to the data point. data points as the temporal neighbors of the data point.

4. The abnormal warning method for wind turbines based on artificial intelligence according to claim 1 is characterized in that: The local density satisfies the expression: ； In the formula, Indicates The dimension The local density of data points; Indicates the total number of spatial neighboring points of the preset data point; Indicates the ordinal number of spatial neighboring points; Indicates data point and its The Euclidean distance between spatially adjacent points.

5. The abnormal warning method for wind turbines based on artificial intelligence according to claim 1 is characterized in that: The density distribution difference satisfies the expression: ； In the formula, Indicates The dimension The density distribution difference of data points; Indicates The dimension The local density of data points; Indicates The dimension data points The local density mean of spatially adjacent points; Indicated by natural constant The exponential function with base ; Represents the absolute value symbol.

6. The abnormal warning method for wind turbines based on artificial intelligence according to claim 1 is characterized in that: The weighted anomaly score satisfies the expression: ； In the formula, Indicates Weighted anomaly scores for multidimensional data points; Indicates Anomaly scores for multidimensional data points; Indicates the total number of dimensions; Indicates the dimension ordinal number; Indicates The total number of data points in each dimension; Indicates the ordinal number of the data point; Indicates The dimension The degree of temporal mutation of each data point; Indicates The dimension The density distribution difference of data points; Represents the standard normalization function.

7. The abnormal warning method for wind turbines based on artificial intelligence according to claim 1 is characterized in that: The method of providing an early warning based on the weighted anomaly score comprises the following steps: Preset abnormal threshold , if the average weighted anomaly score of all multidimensional data points is greater than the preset anomaly threshold , the system will issue an early warning.

8. The abnormal warning system of wind turbines based on artificial intelligence is characterized by: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the abnormal warning method for wind turbines based on artificial intelligence according to any one of claims 1 to 7 is implemented.

Images