A health dynamic evaluation and early warning system for offshore wind turbine

Abstract

The present application relates to the technical field of wind turbine, specifically to a kind of offshore wind turbine health degree dynamic evaluation and early warning system, comprising: training module, for collecting the equipment operating parameter of wind turbine under different environmental intervals as training data, training based on training data to obtain benchmark threshold model;Calculation module, for obtaining real-time environmental factor data, input to benchmark threshold model, determine the real-time health degree benchmark threshold under current environment;Evaluation module, for inputting real-time equipment operating parameter and real-time environmental factor to pre-trained health degree evaluation model, obtain the real-time health degree of wind turbine under current environment;Early warning module, for comparing real-time health degree with real-time health degree benchmark threshold, obtain health degree evaluation result, can improve the accuracy of health degree evaluation and early warning of wind turbine.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wind turbine generators, in particular to a health degree dynamic evaluation and early warning system for offshore wind turbine generators. BACKGROUND

[0002] Offshore wind turbine generators are long-term operated in complex and harsh marine environments such as high salt fog, high humidity and strong typhoon, and the failure risk of their key components (such as main shaft, gear box, blade, etc.) is much higher than that of onshore units. Due to poor accessibility and extremely high maintenance cost, it is of great significance to carry out accurate health state monitoring and early failure warning for offshore wind turbine generators to ensure their safe and stable operation, reduce operation and maintenance cost and improve power generation efficiency.

[0003] The existing health degree evaluation and early warning of wind turbine generators generally use fixed health benchmarks (i.e. threshold values set based on specific environment or historical average data), which leads to the same health state score of the same unit operation state parameters in different environments. When the health degree score exceeds the preset threshold, an alarm will be triggered, but the actual health degree may reflect different health states due to environmental differences, which cannot adaptively adjust the health benchmark according to the environment of the wind turbine generator, thereby affecting the accuracy of health degree evaluation.

[0004] To solve the above problems, we propose a health degree dynamic evaluation and early warning system for offshore wind turbine generators. SUMMARY

[0005] The present application aims to provide a health degree dynamic evaluation and early warning system for offshore wind turbine generators to solve the problems raised in the background.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical solution: a health degree dynamic evaluation and early warning system for offshore wind turbine generators, comprising:

[0007] A training module for collecting equipment operation parameters of wind turbine generators in different environmental intervals as training data, and training a benchmark threshold model based on the training data;

[0008] A calculation module for obtaining real-time environmental factor data and inputting it into the benchmark threshold model to determine the real-time health degree benchmark threshold in the current environment;

[0009] An evaluation module for inputting real-time equipment operation parameters and real-time environmental factors into the pre-trained health degree evaluation model to obtain the real-time health degree of the wind turbine generator in the current environment;

[0010] An early warning module for comparing the real-time health degree with the real-time health degree benchmark threshold to obtain the health degree evaluation result, and generating and issuing a corresponding level of early warning for the wind turbine generator corresponding to the health degree evaluation result that does not meet the preset condition.

[0011] Preferably, the wind turbine generator sets collect equipment operating parameters in different environmental intervals as training data, and the step of training to obtain the baseline threshold model based on the training data comprises:

[0012] An information disc of the environment where the wind turbine generator set is located is constructed, and multiple different environmental intervals are dynamically divided based on at least one environmental factor through the information disc;

[0013] The historical equipment operating parameters of the wind turbine generator set are collected, and training data groups corresponding to each environmental interval are extracted from the historical data based on the multiple different environmental intervals divided;

[0014] The model is trained based on the multiple different environmental intervals and the corresponding training data groups, and a baseline threshold model is generated, which is used to represent the mapping relationship between the environmental interval and the health degree baseline threshold.

[0015] Preferably, the step of constructing the information disc of the environment where the wind turbine generator set is located, and dynamically dividing multiple different environmental intervals based on at least one environmental factor through the information disc comprises:

[0016] An information disc composed of multiple information circles is set, and each information circle corresponds to an environmental factor and constitutes an information page;

[0017] Multiple information points and corresponding information boxes are configured on the information page, and multiple information points corresponding to the same environmental factor are connected to obtain an information circle;

[0018] The management end is configured to the multiple information points, and the multiple information points are communicatively connected to the management end.

[0019] Preferably, the step of configuring multiple information points and corresponding information boxes on the information page comprises:

[0020] The corresponding information box is respectively configured on the information page with each information point as the center point, wherein the information box can cover one or more continuous information points;

[0021] The weight value of the environmental factor relative to the equipment operating parameter is obtained, and based on the weight value, the initial scaling size of the information box is automatically calculated and generated through a predefined scaling mapping relationship, wherein the scaling mapping relationship is that the higher the weight value, the smaller the calculated initial scaling size of the information box;

[0022] The change of the weight value is continuously monitored, and the scaling size of the information box is updated based on the change of the weight value.

[0023] Preferably, the step of obtaining real-time environmental factor data and inputting it into the baseline threshold model to determine the real-time health degree baseline threshold under the current environment comprises:

[0024] Real-time environmental factor data of the wind turbine currently in is collected, in each information page, the information point corresponding to the real-time environmental factor data is marked to obtain a marked information point, and the information frame corresponding to the marked information point on the information page is located with the marked information point as the center;

[0025] The marked information point is determined, the parameter range covered by the information frame located with the marked information point as the center on each information page is obtained, the parameter ranges covered by all the information frames are combined to obtain a plurality of different environmental intervals, and the real-time environmental factor data is obtained by combining the plurality of different real-time environmental intervals.

[0026] The real-time environmental factor data is input into the baseline threshold model to obtain a real-time health degree baseline threshold corresponding to the current environment.

[0027] Preferably, the step of obtaining the parameter range covered by the information frame located with the marked information point as the center on each information page comprises:

[0028] The center coordinates of the marked information point in the information page are determined;

[0029] The current scaling size of the information frame is obtained, wherein the scaling size represents the total number of information points covered by the information frame on the information page;

[0030] The parameter range covered by the information frame is obtained by extending the same number of information points in two directions respectively with the center coordinates as the center.

[0031] Preferably, the step of inputting the real-time device running parameters and the real-time environmental factors into the pre-trained health degree calculation model to obtain the real-time health degree of the device under the current environment comprises:

[0032] The features in the real-time device running parameters and the real-time environmental factors are extracted to obtain feature data;

[0033] The feature data is input into the trained health degree evaluation model, and the real-time health degree of the wind turbine under the current environment is output based on the health degree evaluation model.

[0034] Preferably, the step of comparing the real-time health degree with the real-time health degree baseline threshold to obtain a health degree evaluation result, and generating and issuing a corresponding level of warning for the wind turbine corresponding to the health degree evaluation result that does not meet the preset condition comprises:

[0035] The real-time health degree and the real-time health degree baseline threshold are obtained, and the deviation value between the real-time health degree and the real-time health degree baseline threshold is calculated as the health degree evaluation result;

[0036] The deviation value is compared with preset multiple, graded early warning thresholds to determine a current health degree deviation level; and a corresponding level of early warning is issued according to the health degree deviation level.

[0037] Compared with the prior art, the present application has the following advantages:

[0038] According to the correlation between the environmental factors and the operating parameters of the wind turbine, the real-time environmental interval is dynamically divided, and a benchmark threshold model capable of mapping the environmental interval to the health degree benchmark threshold is trained, so that the health degree benchmark threshold is calculated in real time according to the real-time environmental factors, so that the judgment standard of the health degree can dynamically change according to the real-time environment of the wind turbine, effectively reducing the false positive and false negative rates in complex environments, and improving the accuracy of the health degree evaluation and early warning of the wind turbine. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed for the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0040] Fig. 1 The system structure diagram of the present application.

[0041] Fig. 2 The structural diagram of the information disc of the present application. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0043] Embodiment one:

[0044] Please refer to Figs. 1-2 The present application provides a kind of offshore wind turbine health degree dynamic evaluation and early warning system technical scheme: a kind of offshore wind turbine health degree dynamic evaluation and early warning system, comprising the following steps:

[0045] Training module, for collecting the equipment operating parameters of wind turbine under different environmental intervals as training data, training data are used to obtain benchmark threshold model based on training;

[0046] The step of collecting equipment operation parameters of the wind turbine in different environment intervals as training data and training a benchmark threshold model based on the training data comprises: constructing an information disc of an environment where the wind turbine is located, dynamically dividing a plurality of different environment intervals through the information disc based on at least one environmental factor; collecting historical equipment operation parameters of the wind turbine; extracting training data groups corresponding to each environment interval from the historical data based on the plurality of different environment intervals; and training a model based on the plurality of different environment intervals and the corresponding training data groups to generate a benchmark threshold model, wherein the benchmark threshold model is used to represent a mapping relationship between the environment interval and the health degree benchmark threshold.

[0047] The step of constructing an information disc of an environment where the wind turbine is located, dynamically dividing a plurality of different environment intervals through the information disc based on at least one environmental factor comprises: setting an information disc composed of a plurality of information circles, each information circle corresponding to an environmental factor and constituting an information page; configuring a plurality of information points and corresponding information boxes on the information page, and connecting a plurality of information points corresponding to the same environmental factor to obtain an information circle; and configuring a management end for the plurality of information points, and connecting the plurality of information points to the management end in communication.

[0048] It should be noted that, on each information page, the parameter values of the environmental factors are sequentially stored on the corresponding information points in order;

[0049] Specifically, a plurality of information circles are arranged on the information disc, and each information circle corresponds to an information page, which is equivalent to a large database composed of a plurality of databases. In addition, a plurality of information points are arranged in the information page, and each information point corresponds to a plurality of data values of an environmental factor. Each data value is sequentially stored in the corresponding information point according to the size order. The information point is a discretized environmental parameter database. According to the information points and the information frame on the information disc, the environmental factor is dynamically divided into a plurality of different environmental intervals. The information frame is a dynamic and intelligent data sampling tool, and the size of the information frame is determined by the weight of the data. For example, the environmental factor is wind speed, and the data values corresponding to the wind speed are 3 m / s, 5 m / s, 7 m / s, 9 m / s, 11 m / s, 13 m / s, and a series of wind speed values. The information points are connected in order according to the numerical value to form the corresponding information circle of the "wind speed" environmental factor. A scalable information frame is preset for each information point. For example, the 9 m / s information point of the wind speed circle is configured with an information frame. According to the high correlation (high weight) of the "wind speed" to the fan vibration, a small-sized information frame is automatically set for the point. The frame may only cover a very narrow range of [8.5 m / s, 9.5 m / s]. When the system needs to analyze the health status of the fan under the wind speed of 9 m / s, it will accurately extract the data of the wind speed between 8.5 and 9.5 from the historical database to form a "training data group". The management terminal is connected and dynamically controlled through communication, and is a central processing and coordination unit located in the background of the system. The states (such as the size and position of the information frame) of all information points and their information frames are in real-time communication with the management terminal. For example, when a typhoon passes, according to the correlation of the wind speed to the influence of the equipment operating parameters, the information frame corresponding to the information point is scaled according to the correlation. For example, the information frame of the 9 m / s point of the wind speed circle is enlarged from [8.5, 9.5] to [8.0, 10.0], and a wider environmental interval is used to generate a health degree threshold. This makes the threshold more inclusive in extreme weather, effectively avoiding a large number of false alarms caused by environmental fluctuations, and ensuring the robustness of the system in special periods.

[0050] The steps of configuring a plurality of information points and corresponding information frames on the information page include: configuring a corresponding information frame for each information point on the information page as a center point, wherein the information frame can cover one or more continuous information points; obtaining a weight value of the environmental factor relative to the equipment operating parameter; and automatically calculating and generating an initial scaling size of the information frame based on the weight value through a predefined scaling mapping relationship, wherein the higher the weight value, the smaller the initial scaling size of the information frame calculated; and continuously monitoring the change of the weight value, and updating the scaling size of the information frame based on the change of the weight value.

[0051] It should be noted that the specific content of obtaining the weight value of the environmental factor relative to the device running parameter: the statistical correlation between each environmental factor and the device running parameter can be calculated, and the correlation is normalized as the weight value; receiving the weight value directly input by the user according to the field knowledge; calling from a configuration file pre-stored with different running parameters corresponding to different weight values, and assigning corresponding weight values to each environmental factor based on the correlation, wherein the environmental factor with higher correlation is assigned with larger weight value; the predefined scaling mapping relationship is a nonlinear function, which is configured to: in the interval with higher weight value, the information box size changes sensitively with the weight; in the interval with lower weight value, the information box size changes gently with the weight; continuously monitoring the change of the weight value, and when it is monitored that the change amount of the weight value exceeds a preset threshold, automatically triggering the weight-driven configuration step and the graphical rendering step to update the scaling size of the information box.

[0052] By the predefined scaling mapping relationship, the initial scaling size of the information box is automatically calculated and generated, wherein the scaling mapping relationship is: the higher the weight value, the smaller the calculated initial scaling size of the information box; the specific steps of continuously monitoring the change of the weight value and updating the scaling size of the information box based on the change of the weight value: establishing a scaling mapping relationship to define the functional relationship between the weight value w (usually normalized as 0 , wherein k and c are constants for adjusting the range and sensitivity of the scaling size. To ensure practicability, the minimum value and the maximum value of s need to be set, that is . When the system is initialized or the target is switched, the scaling mapping relationship is called. The current weight value is substituted into the function to calculate the initial scaling size . Then, the information box is rendered and generated on the graphical information page according to . After the system enters the running state, a background monitoring process is started, which periodically (such as every 24 hours) or based on events (such as receiving new batch data) recalculates or obtains the latest weight value . The absolute value of the difference between the new and old weights is calculated, and the absolute value is compared with a preset sensitivity threshold (such as ). If , it is determined that there is an effective change, and the update process is triggered. Once the change is triggered, the system performs the following operations: substituting the new weight value into the scaling mapping relationship to calculate the new target scaling size . To avoid interface jumping, an animation transition method is adopted to update the scaling size of the information box​ Within a frame (e.g. ), the current size of the information box is gradually transitioned from to . After the update is complete, the is updated to , the is updated to , and the new information box size is confirmed to be in effect.

[0053] Specifically, when the health degree of the wind turbine is warned, the real-time health degree is compared with the health degree threshold corresponding to the real-time health degree environment to ensure the accuracy of the health degree evaluation. In order to avoid misjudgment of the warning, the health degree threshold is calculated in real time according to the corresponding environment of the real-time health degree. Different real-time health degrees correspond to different health degree thresholds, and the health degree threshold can dynamically change, so that it can better match the health degree and ensure the accuracy of the evaluation health degree. The health degree threshold corresponding to the same operating parameter is different in different environments. Different health equipment operating parameters in different environment intervals are used as a training set to train a benchmark threshold model for outputting the health degree benchmark threshold of the equipment operating parameter corresponding to the environment interval. Then, the real-time environment factor data corresponding to the real-time environment factor data is determined through the information disc according to the real-time environment factor of the equipment operating parameter collected in real time, and the real-time environment factor data is transmitted to the management end. The management end inputs the real-time environment factor data into the benchmark threshold model to output the health degree benchmark threshold of the equipment operating parameter corresponding to the real-time environment factor. The benchmark threshold model is stored in the management end, and the management end is connected with multiple information points. The information point is equivalent to a database for storing data values of corresponding environment factors. The management end is used for marking multiple information points and collecting data on the information points and transmitting them to the benchmark threshold model. When the data value on the information point corresponds to the real-time environment factor, the information point is directly marked as a marked information point, and the environment factor value on the marked information point is collected through the management end. At the same time, the environment factor values of the information points covered by the information box corresponding to the marked information point are collected. The environment factor values of the information points covered by the information box corresponding to the same marked information point are taken as an environment interval of an environment factor. The parameter ranges covered by the information boxes corresponding to the marked information points of multiple different environment factors are combined to obtain the real-time environment factor corresponding to the real-time equipment operating parameter. The real-time environment factor is input into the benchmark threshold model to output the real-time health degree benchmark threshold corresponding to the real-time environment factor. When the health degree benchmark threshold is evaluated, the real-time environment factor data and the equipment operating parameter are input into the health degree evaluation model to obtain the health degree in the current environment. The health degree is compared with the benchmark threshold to determine the corresponding warning level according to the deviation between the health degree and the benchmark threshold. The corresponding level of warning is issued. The health degree benchmark threshold corresponding to the current environment can be calculated in real time according to the real-time environment factor data, so that it can better adapt to the evaluation of the health degree of the equipment operating parameter in the current environment, and the accuracy of the health degree evaluation of the wind turbine is improved.In the process of benchmark threshold model training, a corresponding information box is set for each information point, the information box is used to cover the parameter range of the environmental factor centered on the information point, and the parameter range of the environmental factor is determined according to the correlation degree of the environmental factor to the real-time device operation parameter; the stronger the correlation degree is, the smaller the parameter range is, and the lower the correlation degree is, the larger the parameter range is; because the environmental factor with strong correlation degree will cause fluctuation of the device operation parameter due to small change, thereby affecting the benchmark threshold for health degree evaluation of the wind turbine, and the change of the health degree may not be due to abnormal health state of the wind turbine, therefore, the correlation degree between the environmental factor and the wind turbine and the health degree benchmark threshold are associated, the value range of the environmental factor is dynamically determined according to the correlation degree between the two in real time, thereby the health degree benchmark threshold of the wind turbine under the same operation parameter is dynamically determined, the accuracy of the output health degree benchmark threshold is improved, the health degree benchmark threshold better adapts to the real-time health degree, the accuracy of the real-time health degree evaluation of the wind turbine is improved, and the misjudgment is reduced; a large information box is used for the low correlation degree, that is, the coverage range of the information box is increased, the generalization ability of the benchmark threshold model is enhanced, the normal fluctuation of the secondary factor should not trigger an alarm, the large information box processes the fluctuation as a same macro state, which is equivalent to a low-pass filter, the irrelevant interference is effectively smoothed, the model is more robust, the false alarm is reduced, thereby the benchmark threshold model can accurately capture the subtle degradation trend of the key component under the key working condition in the early stage, and will not be "alarmed by grass and trees" due to the normal fluctuation of the secondary factor, thereby the fault prediction accuracy and reliability of the benchmark threshold model are far superior to those of the traditional method.

[0054] The computing module is configured to acquire real-time environmental factor data, input the real-time environmental factor data into the benchmark threshold model, and determine the real-time health degree benchmark threshold under the current environment.

[0055] The step of acquiring real-time environmental factor data, inputting the real-time environmental factor data into the benchmark threshold model, and determining the real-time health degree benchmark threshold under the current environment includes the following steps: real-time collection of environmental factor data of the wind turbine currently located, marking of the real-time environmental factor data to obtain a marked information point, and positioning of a corresponding information box on the information page with the marked information point as the center; determination of the marked information point, acquisition of a parameter range covered by the information box positioned with the marked information point as the center on each information page; combination of the parameter ranges covered by all the information boxes to obtain a plurality of different environmental intervals; combination of the plurality of different real-time environmental intervals to obtain real-time environmental factor data; input of the real-time environmental factor data into the benchmark threshold model to obtain a real-time health degree benchmark threshold corresponding to the current environment.

[0056] The step of obtaining the parameter range covered by the information frame centered on the marked information point on each information page comprises: determining the center coordinates of the marked information point in the information page; obtaining the current zoom size of the information frame, wherein the zoom size represents the total number of information points covered by the information frame on the information page; and extending the same number of information points in two directions respectively from the center coordinates to obtain the parameter range covered by the information frame.

[0057] Specifically, the current scaling size of the information box is obtained, which defines the total number of information points N (N is an integer greater than or equal to 1) that the information box can cover on the information page; the (N-1) / 2 information points in the front or back or left or right direction are extended from the center of the information point that is lit, so as to calculate the accurate coverage range of the information box. When N is even, the extension rule is configured to extend one more information point in the increasing direction of the parameter value, or to extend one more information point in the decreasing direction of the parameter value. After the positioning of the information box is completed, the parameter value range corresponding to all the information points covered by the currently positioned information box is formally defined as the effective real-time environment interval of the environmental factor at the current time, and the effective real-time environment interval is input into the reference threshold model to determine the health degree reference threshold of the wind turbine corresponding to the current environment. Assuming that an information page of "wind speed" is matched, the information points represent wind speed values (such as 5, 6, 7, 8, 9, and 10 m / s), the real-time wind speed is 8.5 m / s, and the system is matched to the closest information point of "8 m / s". The information point is marked, the center pixel coordinates of the UI element of the information point representing "8 m / s" are obtained, and it is assumed that according to the weight calculation, the scaling size of the current information box is to cover 3 information points. Taking the "8 m / s" point that is lit as the center, the total number of information points N to be covered is 3, 1 point is extended to the left (in the decreasing direction of the wind speed), reaching "7 m / s", and 1 point is extended to the right (in the increasing direction of the wind speed), reaching "9 m / s". The calculated coverage range of the information box is the continuous interval [7 m / s, 9 m / s], and the information box is positioned to cover the three information points [7, 8, 9]. According to the range [7, 9], the engine calculates the final position and size of the information box on the screen, and the information box stably covers the three information points "7", "8", and "9". The lit "8" point is located at the visual center of the information box. After the positioning is completed, the system data flow module formally records that the real-time environment interval of the current "wind speed" factor is [7 m / s, 9 m / s]. This interval will be combined with the intervals of other environmental factors to form a multi-dimensional environment interval, which is input into the reference threshold model. Through the weight-driven scaling of the information box, the correlation between the environmental interval related to the health degree reference threshold and the device operating parameter can be associated, so as to dynamically adjust the selected interval range according to the influence of the environmental factor on the device operating parameter. The adjustment of the interval range can make the health degree reference threshold more suitable for the corresponding device operating parameter, and improve the adaptability of the health degree reference threshold to the device operating parameter.

[0058] The real-time environmental factor data is collected, and the information points marked in different information circles are associated to obtain the maximum data and the minimum data in the information frame corresponding to each marked information point. The maximum data, the minimum data and the data between the maximum data and the minimum data, that is, the data on all information points covered by the information frame, are used as the environmental input value of the health degree reference threshold corresponding to the information point, for determining the health degree reference threshold corresponding to the information point. A plurality of information pages are set on the information disc according to different environmental factors, and a plurality of information points are set on the information pages. Each information point is connected with the management end. The real-time environmental factors are respectively corresponding on the information pages according to categories, and the information points of the environmental factors on the information pages are determined according to the parameters of the environmental factors. The parameters of the environmental factors correspond to an information point, and the information point is marked as a marked information point. The plurality of marked information points and the corresponding information frame are combined as a real-time environmental factor interval, and the data of the plurality of marked information points is transmitted to the management end. The management end transmits it to the reference threshold model for calculating the health degree reference threshold corresponding to the real-time environmental factor interval. According to the center of the marked information point, the corresponding information is collected according to the information frame. The information frame can be adaptively scaled according to the size of the weight of the environmental factor. The weight of the environmental factor is determined according to the importance of the environment to the corresponding device operating parameter. The health degree reference threshold is dynamically floating with the environment, so that the judgment standard is more in line with the actual operation law of the wind turbine, reduces the false alarm and the missing report, and improves the accuracy of the early warning;

[0059] The evaluation module is used for inputting the real-time device operating parameter and the real-time environmental factor into the pre-trained health degree evaluation model to obtain the real-time health degree of the wind turbine under the current environment.

[0060] The step of inputting the real-time device operating parameter and the real-time environmental factor into the pre-trained health degree calculation model to obtain the real-time health degree of the device under the current environment includes: extracting the features in the real-time device operating parameter and the real-time environmental factor to obtain feature data; inputting the feature data into the trained health degree evaluation model, and outputting the real-time health degree of the wind turbine under the current environment based on the health degree evaluation model.

[0061] The training step of the health degree evaluation model comprises: selecting data of a time period in which the equipment is in a known health state from historical data of the wind turbine as a training data set, for each data sample in the training data set, assigning a health degree label value to it manually or based on rules according to the actual health state of the corresponding equipment; wherein the time period without any fault record and smooth operation is marked as a high health degree label (such as 1 or 100), the time period with an explicit fault is marked as a low health degree label (such as 0), the time period with known performance degradation but no fault is marked as a health degree label of an intermediate value, for each time point in the training data set, a multi-dimensional training feature vector is constructed; the training feature vector comprises two types of features: equipment operating parameter features: one or more sensor readings (such as vibration, temperature, pressure, current, power, etc.); environmental working condition parameter features: one or more environmental factor data (such as wind speed, ambient temperature, wave height, humidity, etc.) of the equipment, the training feature vector is used as input and the corresponding health degree label value is used as the target of supervised learning; a regression type machine learning algorithm (such as gradient boosting tree, random forest, neural network) is used to train the model; the training target is to enable the model to accurately fit the corresponding "health degree label value" from "equipment operating parameters" and "environmental working condition parameters", a historical verification data set not participating in the training is used to evaluate the plurality of candidate models trained; the model with the smallest error between the predicted health degree value and the true health degree label on the verification set is selected as the final health degree calculation model, the final selected health degree calculation model is packaged as a callable service and integrated into the wind turbine health state monitoring system for receiving real-time data and outputting real-time health degree scores;

[0062] The early warning module compares the real-time health degree with the real-time health degree benchmark threshold to obtain a health degree evaluation result, and generates and issues a corresponding level of early warning for the wind turbine corresponding to the health degree evaluation result that does not meet the preset condition;

[0063] The step of comparing the real-time health degree with the real-time health degree benchmark threshold to obtain a health degree evaluation result, and generating and issuing a corresponding level of early warning for the wind turbine corresponding to the health degree evaluation result that does not meet the preset condition comprises: obtaining the real-time health degree and the real-time health degree benchmark threshold, calculating a deviation value between the real-time health degree and the real-time health degree benchmark threshold as a health degree evaluation result; comparing the deviation value with a plurality of preset early warning threshold values classified by levels to determine the current health degree deviation level; issuing a corresponding level of early warning according to the health degree deviation level.

[0064] The environment information of the corresponding interval is used as a test parameter group for training the model to obtain the corresponding health degree benchmark threshold in the environment. The real-time environment interval is input into the model to determine the real-time health degree benchmark threshold. Through training of the working condition information, the real-time operation parameters and the environment are input into the health degree calculation model to obtain the health degree under the same environment. Then, the health degree is compared with the health degree benchmark threshold to obtain the health degree deviation value. The attention threshold, alarm threshold and danger threshold are set according to the health degree deviation value. When the health degree deviation value is lower than the attention threshold, a warning of the attention level is generated to prompt that the unit state appears slight degradation. When the health degree deviation value is lower than the alarm threshold, an alarm warning of the alarm level is generated to prompt that the unit has potential fault risk and needs to be arranged for inspection. When the health degree deviation value is lower than the danger threshold, a danger warning of the danger level is generated to prompt that the unit is about to or has occurred serious failure, and immediate shutdown for repair is suggested.

[0065] In the description of the present specification, the description referring to the terms "one embodiment", "an example", "a specific example" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0066] Although the embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made thereto without departing from the principles and spirit of the present application, the scope of which is defined by the following claims and their equivalents.

Claims

1. A system for offshore wind turbine health dynamic evaluation and early warning, characterized in that, The method comprises the following steps: The training module is used to collect equipment operation parameters of the wind turbine in different environmental intervals as training data, and to train a baseline threshold model based on the training data; The step of collecting equipment operation parameters of the wind turbine in different environmental intervals as training data, and training a baseline threshold model based on the training data comprises the following steps: An information disc of the environment where the wind turbine is located is constructed, and multiple different environmental intervals are dynamically divided based on at least one environmental factor; Historical equipment operation parameters of the wind turbine are collected, and training data groups corresponding to each environmental interval are extracted from the historical data based on the multiple different environmental intervals; A baseline threshold model is generated by training based on the multiple different environmental intervals and the corresponding training data groups, and the baseline threshold model is used to represent the mapping relationship between the environmental intervals and the health degree baseline threshold; The step of constructing an information disc of the environment where the wind turbine is located, and dynamically dividing multiple different environmental intervals based on at least one environmental factor comprises the following steps: An information disc composed of multiple information circles is set, and each information circle corresponds to an environmental factor and constitutes an information page; Multiple information points and corresponding information boxes are configured on the information page, and the multiple information points corresponding to the same environmental factor are connected to obtain an information circle; The multiple information points are configured with a management terminal, and the multiple information points are all communicatively connected with the management terminal; The step of configuring multiple information points and corresponding information boxes on the information page comprises the following steps: The corresponding information boxes are respectively configured on the information page with each information point as the center point, and the information box can cover one or more continuous information points; The weight value of the environmental factor relative to the equipment operation parameter is obtained, and based on the weight value, the initial scaling size of the information box is automatically calculated and generated through a predefined scaling mapping relationship, wherein the scaling mapping relationship is that the higher the weight value, the smaller the calculated initial scaling size of the information box; The change of the weight value is continuously monitored, and the scaling size of the information box is updated based on the change of the weight value; The calculation module is used to obtain real-time environmental factor data, input the real-time environmental factor data into the baseline threshold model, and determine the real-time health degree baseline threshold in the current environment; The evaluation module is used to input the real-time equipment operation parameter and the real-time environmental factor into the pre-trained health degree evaluation model, and obtain the real-time health degree of the wind turbine in the current environment; The early warning module is used to compare the real-time health degree with the real-time health degree baseline threshold, obtain the health degree evaluation result, and generate and issue a corresponding level of early warning for the wind turbine corresponding to the health degree evaluation result that does not meet the preset condition.

2. The offshore wind turbine health dynamic evaluation and early warning system according to claim 1, characterized in that: The step of obtaining real-time environmental factor data and inputting the real-time environmental factor data into the baseline threshold model to determine the real-time health degree baseline threshold in the current environment comprises the following steps: Real-time environmental factor data of the wind turbine is collected, and in each information page, the information points corresponding to the real-time environmental factor data are marked to obtain marked information points, and the corresponding information box on the information page is located with the marked information point as the center. The marked information point is determined, and a parameter range covered by an information frame positioned with the marked information point as a center on each information page is obtained; parameter ranges covered by all information frames are combined to obtain a plurality of different environmental intervals; and the plurality of different real-time environmental intervals are combined to obtain real-time environmental factor data. The real-time environmental factor data is input into a benchmark threshold model to obtain a real-time health degree benchmark threshold corresponding to the current environment.

3. The offshore wind turbine health dynamic evaluation and early warning system according to claim 2, characterized in that: The step of obtaining the parameter range covered by the information frame positioned with the marked information point as a center on each information page includes: determining a center coordinate of the marked information point in the information page; obtaining a current scaling size of the information frame, wherein the scaling size represents a total number of information points covered by the information frame on the information page; extending the same number of information points in two directions respectively from the center coordinate to obtain the parameter range covered by the information frame.

4. The offshore wind turbine health dynamic evaluation and early warning system according to claim 1, characterized in that: The step of inputting the real-time device running parameter and the real-time environmental factor into the pre-trained health degree calculation model to obtain the real-time health degree of the device under the current environment includes: extracting features in the real-time device running parameter and the real-time environmental factor to obtain feature data; inputting the feature data into the trained health degree evaluation model to output the real-time health degree of the wind turbine under the current environment based on the health degree evaluation model.

5. The offshore wind turbine health dynamic evaluation and early warning system according to claim 1, characterized in that: The step of comparing the real-time health degree with the real-time health degree benchmark threshold to obtain a health degree evaluation result, and generating and issuing a corresponding level of warning for the wind turbine corresponding to the health degree evaluation result that does not meet a preset condition includes: obtaining the real-time health degree and the real-time health degree benchmark threshold, and calculating a deviation value between the real-time health degree and the real-time health degree benchmark threshold as the health degree evaluation result; comparing the deviation value with a plurality of preset warning threshold values classified by levels to determine a health degree deviation level, and issuing a corresponding level of warning according to the health degree deviation level.

Images