Method and system for operational state risk assessment of a wind turbine

Abstract

The application relates to the technical field of wind power generation, and discloses a method and system for evaluating the running state risk of a wind turbine generator unit, the method comprising the following steps: constructing a risk evaluation model based on a neural network model, collecting historical first running data and corresponding state data of the wind turbine generator unit under a normal running state, and training the risk evaluation model; obtaining real-time running data of the wind turbine generator unit, also obtaining measured state data corresponding to the real-time running data, inputting the real-time running data into the risk evaluation model, and outputting predicted state data corresponding to the real-time running data by the risk evaluation model; obtaining a difference reference value corresponding to current environment data, calculating the difference between the predicted state data and the measured state data, comparing the difference with the difference reference value, and determining the running state of the wind turbine generator unit, so that the accuracy of risk evaluation is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wind power generation, in particular to a method and system for evaluating the running state risk of a wind turbine generator system. BACKGROUND

[0002] The state monitoring of conventional wind turbine generators mainly relies on fixed threshold values to determine abnormalities, which has significant limitations. In a complex and variable environment, such as sudden temperature changes and wind speed fluctuations, fixed threshold values are difficult to adapt to different working conditions, leading to false positives or false negatives. For example, at high temperatures and low wind speeds, the heat dissipation efficiency of the gearbox decreases, and normal temperature fluctuations are easily misjudged as faults. However, at low temperatures and high wind speeds, a temporary increase in vibration amplitude may be ignored, and potential risks may be missed. Existing methods fail to effectively integrate historical fault data and real-time environmental parameters, cannot dynamically adjust threshold values, and rely on single sensor data, making it difficult to capture multi-parameter degradation characteristics.

[0003] Similar prior art includes Chinese patent application CN119829957A, which discloses a device fault detection data analysis method and system for a digital factory, comprising S1: synchronously collecting comprehensive maintenance data of the device using high-precision clock synchronization technology; S2: constructing a graph neural network by taking the position data of the device as a graph node and the functional association and data interaction data between devices as the edges of the graph, and constructing a device association model based on the graph neural network, inputting the state data of the device into the device association model, and outputting state correlation factors; S3: inputting the comprehensive maintenance data and the state correlation factors into the state of the reinforcement learning agent, training the agent using the Q-learning algorithm, inputting real-time state data into the agent, outputting a state score, and setting an abnormal state threshold value. When the state score is lower than the abnormal state threshold value, the corresponding real-time data is marked as abnormal. Similar prior art also includes Chinese patent application CN119826908A, which discloses a method and device for detecting the abnormal operation state of a crane pipe. The method comprises: during the operation of the crane pipe, real-time acquisition of the operation parameters of the crane pipe; fusion of the operation parameters to obtain fusion data; input of the fusion data into a pre-trained abnormality detection model to obtain a reconstruction error according to the output of the abnormality detection model; and determination of whether the operation state of the crane pipe is abnormal according to the reconstruction error.

[0004] However, the above two technical solutions do not consider the dynamic influence of environmental parameters (such as temperature and wind speed) on the state of the device, and there is still a deficiency in the accuracy of the evaluation. SUMMARY

[0005] To solve the above technical problems, the present application provides a method and system for evaluating the running state risk of a wind turbine generator system to improve the accuracy of risk evaluation.

[0006] In a first aspect, the present application provides a method for risk assessment of an operating state of a wind turbine, sensors are installed on each component of the wind turbine, and the method comprises:

[0007] Step S1: constructing a risk assessment model based on a neural network model, collecting and training the risk assessment model based on historical first operating data and corresponding state data of the wind turbine in a normal operating state;

[0008] Step S2: obtaining real-time operating data of the wind turbine, and obtaining real-time measured state data corresponding to the real-time operating data, inputting the real-time operating data into the risk assessment model, and the risk assessment model outputting predicted state data corresponding to the real-time operating data;

[0009] Step S3: obtaining a difference reference value corresponding to current environmental data, calculating and comparing a difference between the predicted state data and the measured state data with the difference reference value, and determining an operating state of the wind turbine, the operating state including a normal operating state, a low-risk operating state, and a high-risk operating state.

[0010] In combination with the first aspect, in a first implementation manner of the first aspect of the present application, in step S3, the difference reference value corresponding to the current environmental data is obtained, comprising:

[0011] collecting historical second operating data and corresponding state data of the wind turbine in an abnormal operating state, grouping the state data corresponding to the historical second operating data according to environmental data in the historical second operating data to obtain a plurality of environmental groups, clustering the state data in each environmental group, and obtaining and calculating a difference reference value corresponding to the environmental group based on a clustering result;

[0012] based on the environmental groups and the difference reference values corresponding to the environmental groups, establishing a mapping table of environmental groups and difference reference values, matching the current environmental data with the environmental groups, and obtaining a corresponding difference reference value from the mapping table of environmental groups and difference reference values based on the matched environmental groups.

[0013] In combination with the first aspect, in a second implementation manner of the first aspect of the present application, the historical second operating data and corresponding state data of the wind turbine in an abnormal operating state are collected, comprising:

[0014] Based on the fault records in the wind turbine fault record table, operation data of the entire wind turbine is divided into operation data of multiple stages, from the operation data of each stage, operation data and corresponding state data in a preset time period before a second time point are obtained as target operation data and target state data, a first time point is determined from the preset time period based on the target state data, and the first time point is taken as a time demarcation point between the normal operation state and the abnormal operation state of the wind turbine.

[0015] Operation data from each time of failure recovery to the first time point is taken as first-stage operation data, and operation data from the first time point to the second time point is taken as second-stage operation data, multiple first-stage operation data constitute historical first operation data, and state data corresponding to the historical first operation data is also obtained, multiple second-stage operation data constitute historical second operation data, and state data corresponding to the historical second operation data is also obtained, wherein the second time point is the start time of any one fault in the wind turbine fault record table.

[0016] In combination with the first aspect, in a third implementation manner of the first aspect of the present application, calculating the difference reference value corresponding to the environment group comprises:

[0017] Any one parameter in the state data in any one of the environment groups is taken as a third parameter, the numerical value of the third parameter is clustered, multiple clusters and a standard deviation corresponding to each cluster are obtained, a standard deviation corresponding to each cluster is multiplied by a preset safety factor to obtain a difference reference value corresponding to each cluster, an average difference reference value of the multiple clusters is calculated and taken as a difference reference value of the third parameter corresponding to the environment group, this step is repeated to calculate difference reference values of other parameters corresponding to the environment group, and the difference reference values of each parameter in the state data in the environment group constitute the difference reference value corresponding to the environment group.

[0018] In combination with the first aspect, in a fourth implementation manner of the first aspect of the present application, determining the first time point from the preset time period based on the target state data comprises:

[0019] A third time point corresponding to a first parameter in the target state data and a fourth time point corresponding to a second parameter in the target state data are calculated, if the third time point is less than or equal to the fourth time point, the third time point is taken as the first time point, and if the third time point is greater than the fourth time point, the fourth time point is taken as the first time point.

[0020] In conjunction with the first aspect, in the fifth implementation of the first aspect of this application, calculating the third time point corresponding to the first parameter in the target state data and the fourth time point corresponding to the second parameter in the target state data includes:

[0021] Take any one parameter in the target state data as the first parameter, calculate the first average value of the first parameter within the preset time period, calculate the difference between the value of the first parameter at all time points within the preset time period and the first average value as the first difference, and obtain and select the time point corresponding to the first parameter with the largest difference from multiple first differences as the third time point.

[0022] Take any parameter other than the first parameter in the target state data as the second parameter, repeat the above steps, and calculate the fourth time point corresponding to the second parameter in the target state data.

[0023] In conjunction with the first aspect, in the sixth implementation of the first aspect of this application, the setting of the preset security factor includes:

[0024] Collect the number of misjudgments of low-risk or high-risk operating status over the past N months. If there are M consecutive misjudgments, the safety factor is increased; otherwise, it remains unchanged.

[0025] If a high-risk operating condition is missed once in the past N months, the safety factor will be lowered; otherwise, it will remain unchanged. Here, N represents a positive integer between 1 and 12, and M represents a positive integer greater than or equal to 2 and less than or equal to 5.

[0026] In conjunction with the first aspect, in the seventh implementation of the first aspect of this application, determining the operating state of the wind turbine generator set includes:

[0027] If the difference between the measured value and the predicted value of each parameter in the status data is less than or equal to the difference reference value corresponding to the parameter, then the wind turbine generator is determined to be in normal operation, and the current monitoring cycle of the wind turbine generator continues.

[0028] If the difference between the measured value and the predicted value of any parameter in the status data is greater than the reference difference value corresponding to the parameter, but less than or equal to twice the reference difference value corresponding to the parameter, then the wind turbine generator is determined to be in a low-risk operating state, and the maintenance process is initiated.

[0029] If the difference between the measured value and the predicted value of all parameters in the state data is greater than the difference reference value corresponding to the parameter or the difference value of any one parameter is greater than twice the difference reference value corresponding to the parameter, it is determined that the wind turbine is in a high-risk operating state, the wind turbine is controlled to stop immediately, and an emergency maintenance instruction is sent to the person in charge.

[0030] In combination with the first aspect, in an eighth implementation form of the first aspect of the application, the difference between the predicted state data and the measured state data is calculated, including:

[0031] Any one parameter in the state data is taken as a fourth parameter, and the difference value of the fourth parameter is calculated by the formula wherein λ represents the difference value of the fourth parameter, A' represents the predicted value of the fourth parameter, and A represents the measured value of the fourth parameter; this step is repeated to calculate the difference between the measured value and the predicted value of all parameters in the state data.

[0032] Secondly, the application provides a running state risk assessment system for a wind turbine, sensors are installed on each component of the wind turbine, and the system includes:

[0033] a model construction unit configured to construct a risk assessment model based on a neural network model, collect historical first running data and corresponding state data under a normal operating state of the wind turbine, and train the risk assessment model based on the historical first running data and the corresponding state data;

[0034] a data acquisition unit configured to acquire real-time running data of the wind turbine, acquire measured state data corresponding to the real-time running data, input the real-time running data into the risk assessment model, and output predicted state data corresponding to the real-time running data from the risk assessment model;

[0035] a risk assessment unit configured to acquire a difference reference value corresponding to current environment data, calculate a difference between the predicted state data and the measured state data, compare the difference with the difference reference value, and determine an operating state of the wind turbine, wherein the operating state includes a normal operating state, a low-risk operating state, and a high-risk operating state.

[0036] Compared with the prior art, the application has at least the following advantages:

[0037] The risk assessment model based on the neural network model can efficiently learn the correlation characteristics of multiple parameters in the normal operation state, for example, the nonlinear relationship among the gear box temperature, the main shaft speed and the environmental wind speed, and reduce the prediction error. By collecting the operation data in real time and comparing the operation data with the model prediction value, the abnormal fluctuation can be quickly identified, and the unnecessary shutdown target can be achieved. The calculation of the dynamic difference reference value is realized by environmental grouping and cluster analysis, and the difference reference value is generated in combination with the standard deviation and the safety factor. Compared with the fixed threshold, the method can reduce the false positive rate, and the adaptive adjustment mechanism of the safety factor further optimizes the sensitivity of the method, ensures the dynamic change of the operation data, and reduces the missed alarm and misjudgment.

[0038] The application also captures the equipment degradation characteristics by integrating multi-dimensional data such as temperature, vibration and power generation, and can quickly identify and filter the short-term vibration anomaly when the gust causes instantaneous load fluctuation, so as to avoid false triggering of shutdown. In extreme working conditions such as low temperature of-20 DEG C or high humidity environment on the sea, the model can automatically match the difference reference value of the historical similar scene by environmental grouping and migration learning framework. The accuracy of the low-temperature embrittlement risk warning of the bearing is improved. In summary, the application can effectively improve the accuracy of risk assessment. BRIEF DESCRIPTION OF DRAWINGS

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

[0040] Figure 1 An embodiment schematic diagram of the operation state risk assessment method for the wind turbine generator set in the embodiment of the application;

[0041] Figure 2 A cluster result schematic diagram of the gear box temperature in the embodiment of the application;

[0042] Figure 3 A historical first operation data and historical second operation data definition schematic diagram in the embodiment of the application;

[0043] Figure 4 An embodiment schematic diagram of the operation state risk assessment system for the wind turbine generator set in the embodiment of the application. DETAILED DESCRIPTION

[0044] The embodiments of the present application provide a method and system for risk assessment of operating state of a wind turbine generator set. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not have to be used to describe a particular order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the term "comprising" or "having" and any variation thereof is intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0045] Embodiment I:

[0046] For ease of understanding, the specific process of the embodiments of the present application is described below. Please refer to Figure 1 One embodiment of the method for risk assessment of operating state of a wind turbine generator set in the embodiments of the present application includes: installing sensors on each component of the wind turbine generator set, and implementing by performing the following steps:

[0047] Step S1: constructing a risk assessment model based on a neural network model, collecting and training the risk assessment model based on historical first operating data and corresponding state data of the wind turbine generator set under normal operating state. The operating data refers to the data collected by the sensor when the wind turbine generator set is in normal operating state, and at least includes gear box bearing pressure, main shaft rotating speed, environmental wind speed and environmental temperature, and the state data at least includes gear box temperature and main bearing vibration amplitude, and can also include power generation and the like.

[0048] Specifically, by collecting historical operating data and corresponding state data of the wind turbine generator set under normal operating state, the neural network model is trained to construct a risk assessment model. The risk assessment model can learn the data distribution characteristics and patterns under normal operating state, capture the nonlinear relationship between key parameters, and provide a basis for subsequent state prediction. Through a large amount of historical data training, the model can effectively identify the state data characteristics under normal operating state, realize accurate characterization of the state of the wind turbine generator set, and improve the prediction accuracy of the risk assessment model. The historical data used here refers to the historical first operating data, and how to obtain the historical first operating data will be described in detail below.

[0049] Step S2: Obtain real-time operation data of the wind turbine generator set, and obtain measured state data corresponding to the real-time operation data. The real-time operation data is input into the risk assessment model, and the risk assessment model outputs predicted state data corresponding to the real-time operation data.

[0050] Specifically, real-time operation data of the wind turbine generator set is obtained and input into the trained risk assessment model for prediction. At the same time, measured state data corresponding to the real-time operation data is also obtained for comparison with the predicted state data to evaluate the operation state of the wind turbine generator set. The risk assessment model outputs predicted state data based on the real-time operation data in combination with the trained features and patterns. By comparing the predicted data with the measured data, a difference value is calculated, and a dynamic difference reference value is used for state determination.

[0051] Step S3: Obtain a difference reference value corresponding to the current environment data, calculate and compare the difference between the predicted state data and the measured state data with the difference reference value to determine the operation state of the wind turbine generator set. The operation state includes normal operation state, low-risk operation state and high-risk operation state.

[0052] Specifically, the calculation of the dynamic difference reference value is the core of the present application. The difference between the predicted state data and the measured state data is calculated and compared with the difference reference value. Specifically, the difference value of each parameter is first calculated, and the operation state of the wind turbine generator set is determined according to the setting of the difference reference value. The normal operation state means that the difference values of all parameters are within the allowed range; the low-risk operation state means that the difference values of some parameters exceed the allowed range but do not reach the dangerous level; and the high-risk operation state means that the difference values significantly exceed the allowed range, which will cause serious operation risk. Through the difference analysis, the operation state of the wind turbine generator set is accurately determined, and an explicit risk level evaluation result is provided. At the same time, the accuracy and adaptive ability of risk assessment are also improved.

[0053] In the embodiments of the present application, the cooperation between the above steps can improve the accuracy of risk assessment.

[0054] In a specific embodiment, the obtaining of the difference reference value corresponding to the current environment data in step S3 specifically includes the following steps:

[0055] Historical second-stage operating data and corresponding status data under abnormal operating conditions of wind turbine generators are collected. The status data corresponding to the historical second-stage operating data are grouped according to the environmental data within the historical second-stage operating data, resulting in multiple environmental groups. The status data within each environmental group is clustered, and the difference reference value corresponding to the environmental group is obtained and calculated based on the clustering results. A mapping table between environmental groups and their corresponding difference reference values ​​is established. Current environmental data is matched with environmental groups, and the corresponding difference reference value is obtained from the mapping table based on the matched environmental group. Current environmental data includes current ambient temperature and current ambient wind speed.

[0056] Specifically, in traditional operational status assessment methods, reference values ​​are generally fixed, which can lead to numerous false alarms when external environmental data changes. Therefore, it is necessary to analyze operational and status data together with environmental data to avoid misjudgments under different operating conditions using a single reference value. First, historical secondary operational data is collected and grouped according to environmental data. The definition of historical secondary operational data will be described in detail below. For example, a wind farm collected operational data before 100 gearbox overheating failures, i.e., historical secondary operational data, including parameters such as: environmental data: temperature (°C), wind speed (m / s); status data: gearbox temperature (°C), main bearing vibration amplitude (mm / s), and power generation (kW). Based on the environmental data in the historical secondary operational data, it is divided into several groups: Group A: High temperature and low wind speed (temperature ≥ 20°C, wind speed ≤ 5 m / s); Group B: Normal temperature and medium wind speed (temperature 5°C~15°C, wind speed 6~10 m / s); Group C: Low temperature and high wind speed (temperature ≤ 0°C, wind speed ≥ 12 m / s). Depending on the actual operating conditions of the wind turbine generators, they will be divided into more groups.

[0057] Next, we clustered a parameter from the environmental state data within each group, taking group A (high temperature, low wind speed) as an example. The data sample includes operational data prior to 20 failures, with gearbox temperatures ranging from 70℃ to 95℃. Using the K-means clustering algorithm with K=3, we clustered the gearbox temperatures. The clustering results are shown below. Figure 2 As shown. Based on the clustering results, the difference reference value corresponding to each cluster is calculated. The specific calculation method for each cluster's difference reference value will be described below. The average of the difference reference values ​​corresponding to multiple clusters is used as the difference reference value for group A. Similarly, the difference reference values ​​for groups B and C are calculated, as described below. Through joint analysis of environmental and state data, misjudgments of fixed reference values ​​under different operating conditions can be effectively avoided.

[0058] In a specific embodiment, historical second operation data of the wind turbine in an abnormal operation state and corresponding state data are collected, specifically including the following steps:

[0059] Based on the fault records in the wind turbine fault record table, the operation data of the entire wind turbine is divided into operation data of multiple stages. From the operation data of each stage, the operation data and corresponding state data in a preset time period before the second time point are obtained as target operation data and target state data. Based on the target state data, the first time point is determined within the preset time period, and the first time point is taken as the time demarcation point between the normal operation state and the abnormal operation state of the wind turbine. The operation data from each time of failure recovery to the first time point is taken as the first stage operation data, and the operation data from the first time point to the second time point is taken as the second stage operation data. The multiple first stage operation data constitutes the historical first operation data, and the state data corresponding to the historical first operation data is also obtained. The multiple second stage operation data constitutes the historical second operation data, and the state data corresponding to the historical second operation data is also obtained. The second time point is the start time of any one fault in the wind turbine fault record table.

[0060] Specifically, referring to Figure 3 As shown in the figure, taking a certain gearbox overheating fault as an example, the fault start time second time point is May 10, 2024 14:00, and the preset analysis window is 72 hours before the fault. Because before the fault occurs, the values of various parameters in the state data will appear irregular fluctuations. By analyzing the fluctuation, the first time demarcation point can be determined as May 7, 14:00 - at this time, the gearbox temperature rises from normal 75℃ to 85℃, and then drops to 80℃. It may fluctuate several times, but the cooling efficiency decreases significantly under stable wind speed conditions. At this time, attention should be paid and analysis should be carried out. The first stage data of multiple fault events constitutes a historical first operation data set, that is, the operation data in the normal operation state. The second stage data of multiple fault events constitutes a historical second operation data set, that is, the operation data in the abnormal operation state. By clearly dividing the normal and abnormal data, it can be distinguished which is the degradation feature of the equipment and which is the real fault precursor during model training. Through time dynamic demarcation, both short-term fluctuations and irregular abnormalities can be avoided, and the transformation from "after-maintenance" to "predictive maintenance" can be realized.

[0061] In one specific embodiment, calculating the difference reference value corresponding to an environment group includes: obtaining any parameter from the state data of any environment group as a third parameter; clustering the value of the third parameter to obtain multiple clusters and the standard deviation corresponding to each cluster; multiplying the standard deviation corresponding to each cluster by a preset safety factor to obtain the difference reference value corresponding to each cluster; calculating the average difference reference value of multiple clusters and using it as the difference reference value of the third parameter corresponding to the environment group; repeating this step to calculate the difference reference values ​​of other parameters corresponding to the environment group; the difference reference value of each parameter in the state data of the environment group constitutes the difference reference value corresponding to the environment group.

[0062] Specifically, taking a high-temperature, low-wind-speed environment group (temperature ≥ 20℃, wind speed ≤ 5m / s) as an example, the calculation process of the difference reference value is explained: Gearbox temperature is selected as the third parameter. This group contains 100 historical data points, with a temperature range of 70℃ to 95℃. The K-means algorithm (K = 3) is used to cluster the temperature data; see [link to relevant documentation]. Figure 2 As shown, when calculating the cluster-level difference reference value, assuming a safety factor of 1.5, the difference reference value for cluster 1 is 3℃ × 1.5 = 4.5℃, for cluster 2 it is 4℃ × 1.5 = 6.0℃, and for cluster 3 it is 2℃ × 1.0 = 2.0℃. Taking the inter-cluster average—(4.5 + 6.0 + 2.0) / 3 ≈ 4.2℃—means the gearbox temperature difference reference value under this environmental group is ±4.2℃. In other words, if the difference between the predicted and measured gearbox temperatures is within ±4.2℃, the gearbox temperature of the wind turbine generator is considered normal. Following the same process, the difference reference value for the main bearing vibration amplitude under this environmental group is calculated. When the measured gearbox temperature is 5℃ higher than the predicted value (exceeding 4.2℃), the wind turbine generator is determined to be in a low-risk operating state, rather than directly classified as high-risk, reducing false shutdowns by 50%. In high-temperature environments, the temperature fluctuation tolerance is improved compared to the traditional fixed threshold of ±3℃, avoiding frequent false alarms caused by fluctuations in heat dissipation efficiency. The above method preserves the data distribution characteristics by using a weighted average of cluster-level difference reference values, and dynamically adjusts the sensitivity of the assessment through a safety factor, thereby achieving refined status monitoring in complex environments.

[0063] In one specific embodiment, determining a first time point from a preset time period based on target state data includes: calculating a third time point corresponding to a first parameter in the target state data and a fourth time point corresponding to a second parameter in the target state data; if the third time point is less than or equal to the fourth time point, then the third time point is taken as the first time point; if the third time point is greater than the fourth time point, then the fourth time point is taken as the first time point.

[0064] Specifically, the first parameter can be taken as the gearbox temperature, the second parameter can be taken as the main bearing vibration amplitude, and other parameters such as power generation can also be included in the state data, which is not limited by the present application. Through cross-validation of multiple parameter abnormal time sequences, the earliest possible abnormal signal is captured, and the lagging fluctuation depending on a single parameter is avoided. The definition of the third time point and the fourth time point will be described below.

[0065] In a specific embodiment, the third time point corresponding to the first parameter in the target state data and the fourth time point corresponding to the second parameter in the target state data are calculated, comprising:

[0066] Taking any one parameter in the target state data as the first parameter, the first average value of the first parameter in the preset time period is calculated, the difference between the value of the first parameter corresponding to all time points in the preset time period and the first average value is taken as the first difference value, and the time point corresponding to the first parameter with the largest difference value is selected from the plurality of first difference values as the third time point; taking any one parameter in the target state data except the first parameter as the second parameter, repeating the above steps, and calculating the fourth time point corresponding to the second parameter in the target state data.

[0067] Specifically, for example, 72 hours of data before the failure is analyzed, and the temperature range in the preset time period is 70℃-92℃. The vibration amplitude in the same time period is 1.5mm / s-5.0mm / s. The temperature average in 72 hours is calculated as 80℃, and then the absolute difference between the temperature and the average is calculated point by point. It is found that the temperature suddenly rises to 90℃ at 12:00 the next day, and the difference is 10℃, which is the largest difference of the gearbox temperature in the preset time period. Therefore, the abnormal time point of the temperature, i.e. the third time point, is marked at 12:00 the next day. Still the same 72 hours of data before the failure, the vibration average is calculated as 2.5mm / s, and the vibration suddenly increases to 4.5mm / s at 13:00 the third day, with a difference of 2.0mm / s, which is the largest difference of the vibration amplitude in the preset time period. Therefore, the abnormal time point of the vibration amplitude, i.e. the fourth time point, is marked at 13:00 the third day. Through the above method, the earliest abnormal time point can be used as a reference when defining normal and abnormal operating states, so as to avoid missing possible risks. At the same time, the earliest abnormal signal is located by the maximum deviation of the parameter and the average, avoiding the dependence on a single threshold or fixed rule, and providing a basis for calculating dynamic difference reference values.

[0068] In a specific embodiment, the setting of the preset safety factor comprises: collecting the number of misjudgments of low-risk operating states or high-risk operating states in the last N months, if misjudgment occurs for M consecutive times, the safety factor is increased, otherwise it remains unchanged; if a high-risk operating state is missed once in the last N months, the safety factor is decreased, otherwise it remains unchanged, wherein N represents a positive integer between 1 and 12, and M represents a positive integer greater than or equal to 2 and less than or equal to 5.

[0069] Specifically, the initial parameters are set, the safety factor is 1.5, the evaluation period N = 3 months, and M = 3 consecutive misjudgments. During the summer high-temperature period, the dust accumulation on the heat dissipation fins causes occasional temperature fluctuations, for example, 75°C (normal upper limit) is misjudged as high risk for 3 consecutive times from July to September, threshold ± 4.5°C, actual measurement 79.5°C). Then increase the safety factor from 1.5 to 1.6, and the difference reference value is expanded to ± 4.8°C = 3°C x 1.6. Through subsequent operation observation, the misjudgment rate is reduced from 15% to 5%, and unnecessary shutdown inspection is reduced. In winter, the lubricating oil will be sticky, which will mask early wear, for example, a bearing wear was not timely warned in December, the actual vibration difference value is 1.8 mm / s, the difference reference value is 1.5 mm / s, and the actual value is 1.3 mm / s. Then decrease the safety factor from 1.5 to 1.4, and the vibration difference reference value is tightened to ± 1.4 mm / s = 1.0 mm / s x 1.4. Through subsequent operation observation, two similar abnormalities are captured in the next month, and the maintenance response time is shortened by 40%. Continuous misjudgment indicates that the difference reference value is too strict, and the coefficient is adjusted upward to relax the limit (e.g. ± 4.5°C → ± 4.8°C), avoiding the "wolf coming" effect. Single missed report reflects that the threshold is too loose, and the coefficient is adjusted downward to enhance the sensitivity (e.g. 1.5 mm / s → ± 1.4 mm / s), blocking the implicit risk. By dynamically adjusting the safety factor, the environmental changes can be accurately matched, and misjudgment or missed report can be avoided.

[0070] In a specific embodiment, the step S3 of determining the operating state of the wind turbine generator set comprises:

[0071] If the difference between the measured value and the predicted value of each parameter in the state data is less than or equal to the difference reference value corresponding to the parameter, it is determined that the wind turbine generator set is in a normal operating state, and the current monitoring period of the wind turbine generator set is maintained;

[0072] If the difference between the measured value and the predicted value of any one parameter in the state data is greater than the difference reference value corresponding to the parameter, and less than or equal to twice the difference reference value corresponding to the parameter, it is determined that the wind turbine generator set is in a low-risk operating state, and a maintenance process is started;

[0073] If the difference between the measured value and the predicted value of all parameters in the state data is greater than the difference reference value corresponding to the parameter, or the difference value of any one parameter is greater than twice the difference reference value corresponding to the parameter, it is determined that the wind turbine generator set is in a high-risk operating state, and the wind turbine generator set is immediately shut down, and an emergency maintenance instruction is sent to the person in charge.

[0074] Specifically, first, the difference value between the measured value and the predicted value of each parameter in the state data, such as the gearbox temperature, the main bearing vibration amplitude and the power generation, etc., is calculated, and compared with the difference reference value corresponding to the parameter under the current environmental data. The difference reference value is determined according to the historical second running data of the wind turbine generator set, and is used to measure the deviation range allowed by the parameter. The difference reference value setting method has been described in detail above. Secondly, according to the comparison result of the difference value between the predicted value and the measured value and the difference reference value, the running state of the wind turbine generator set is determined. The normal running state is that if the difference value of each parameter is within the preset difference reference value range, that is, the difference value ≤ the difference reference value, it is determined that the wind turbine generator set is running normally, and the current monitoring period is maintained to continue monitoring the equipment state. The low-risk running state is that if the difference value of any one parameter exceeds the difference reference value but does not exceed twice the difference reference value, that is, the difference value > the difference reference value and ≤ 2 × the difference reference value, it is determined that the wind turbine generator set is in a low-risk state. At this time, the preventive maintenance process is started to prevent the potential failure from further deteriorating. In the high-risk running state, if the difference value of all parameters exceeds the difference reference value, or the difference value of any one parameter exceeds twice the difference reference value, that is, the difference value > 2 × the difference reference value, it is determined that the wind turbine generator set is in a high-risk state, and immediate emergency measures are taken, including controlling the wind turbine generator set to stop and sending an emergency maintenance instruction to the relevant person in charge, so as to check and repair the problem as soon as possible.

[0075] By analyzing the difference value of each parameter one by one and comparing it with the difference reference value, the running state of the wind turbine generator can be accurately determined, and false positives caused by a single parameter anomaly can be avoided. The running risk of the wind turbine generator set is also divided into different levels, and corresponding measures are taken, which can timely deal with potential problems and avoid unnecessary over-maintenance.

[0076] In a specific embodiment, the difference between the predicted state data and the measured state data is calculated, including: taking any one parameter in the state data as a fourth parameter, calculating the difference value of the fourth parameter by the formula wherein λ represents the difference value of the fourth parameter, A' represents the predicted value of the fourth parameter, and A represents the measured value of the fourth parameter; repeat this step to calculate the difference between the measured value and the predicted value of all parameters in the state data.

[0077] Specifically, by calculating the difference value of each parameter in the state data respectively, it can be identified which parameters have a large deviation, and then compared with the corresponding difference reference value to provide a basis for evaluating the running state of the wind turbine generator set.

[0078] Embodiment two:

[0079] The method for evaluating the running state risk of the wind turbine generator set in the embodiments of the present application is described above, and the system for evaluating the running state risk of the wind turbine generator set in the embodiments of the present application is described below, please refer to Figure 4 One embodiment of the system for evaluating the running state risk of the wind turbine generator set in the embodiments of the present application includes installing sensors on each component of the wind turbine generator set, and is implemented through the following modules:

[0080] The model construction unit is configured to construct a risk evaluation model based on a neural network model, collect historical first running data and corresponding state data under the normal running state of the wind turbine generator set, and train the risk evaluation model based on the historical first running data and the corresponding state data;

[0081] The data acquisition unit is configured to acquire real-time running data of the wind turbine generator set, and acquire measured state data corresponding to the real-time running data, input the real-time running data into the risk evaluation model, and output predicted state data corresponding to the real-time running data by the risk evaluation model;

[0082] The risk evaluation unit is configured to acquire a difference reference value corresponding to the current environmental data, calculate a difference between the predicted state data and the measured state data, compare the difference with the difference reference value, and determine the running state of the wind turbine generator set, wherein the running state includes a normal running state, a low-risk running state and a high-risk running state.

[0083] Through the cooperation of the above-mentioned components, the accuracy of risk evaluation can be improved.

[0084] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-mentioned system, system and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described herein.

[0085] The integrated unit, if realized in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the embodiments of the present application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0086] The above-described embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for operating state risk assessment of a wind turbine generator system, characterized by, The method comprises: Step S1: constructing a risk assessment model based on a neural network model, collecting and training the risk assessment model based on historical first operation data and corresponding state data of the wind turbine under normal operation state; Step S2: obtaining real-time operation data of the wind turbine, also obtaining measured state data corresponding to the real-time operation data, inputting the real-time operation data into the risk assessment model, and the risk assessment model outputting predicted state data corresponding to the real-time operation data; Step S3: obtaining a difference reference value corresponding to current environment data, calculating and comparing the difference between the predicted state data and the measured state data with the difference reference value to determine the operation state of the wind turbine, wherein the operation state comprises normal operation state, low-risk operation state and high-risk operation state; Obtaining a difference reference value corresponding to current environment data comprises: collecting historical second operation data and corresponding state data of the wind turbine under abnormal operation state, grouping the state data corresponding to the historical second operation data according to environment data in the historical second operation data to obtain multiple environment groups, clustering the state data in each environment group, and obtaining and calculating a difference reference value corresponding to the environment group based on the clustering result; based on the environment groups and the difference reference values corresponding to the environment groups, establishing a mapping table of environment groups and difference reference values, matching the current environment data with the environment groups, and obtaining the corresponding difference reference value from the mapping table of environment groups and difference reference values based on the matched environment groups; Collecting historical second operation data and corresponding state data of the wind turbine under abnormal operation state comprises: based on the fault records in the fault record table of the wind turbine, dividing the operation data of the entire wind turbine into operation data of multiple stages, obtaining operation data and corresponding state data in a preset time period before a second time point from operation data of each stage as target operation data and target state data, determining a first time point from the preset time period based on the target state data, and taking the first time point as a time demarcation point between normal operation state and abnormal operation state of the wind turbine; taking operation data from each time of failure recovery to the first time point as first stage operation data, and taking operation data from the first time point to the second time point as second stage operation data, multiple first stage operation data constituting historical first operation data, also obtaining state data corresponding to the historical first operation data, multiple second stage operation data constituting historical second operation data, also obtaining the historical second operation data and corresponding state data, wherein the second time point is the start time of any one fault in the fault record table of the wind turbine; The difference reference value corresponding to the environment group is calculated, including: taking any one parameter in the state data in any one of the environment groups as a third parameter, clustering the numerical value of the third parameter, obtaining a plurality of clusters and a standard deviation corresponding to each cluster, calculating the standard deviation corresponding to each cluster multiplied by a preset safety factor to obtain a difference reference value corresponding to each cluster, calculating the average difference reference value of the plurality of clusters as the difference reference value of the third parameter corresponding to the environment group, repeating the step to calculate the difference reference value of other parameters corresponding to the target state data, and the difference reference value of each parameter in the state data in the environment group forms the difference reference value corresponding to the environment group.

2. The method of claim 1, wherein, The first time point is determined based on the target state data within the preset time period, including: The third time point corresponding to the first parameter in the target state data and the fourth time point corresponding to the second parameter in the target state data are calculated, and if the third time point is less than or equal to the fourth time point, the third time point is taken as the first time point, and if the third time point is greater than the fourth time point, the fourth time point is taken as the first time point.

3. The method of claim 2, wherein, The third time point corresponding to the first parameter in the target state data and the fourth time point corresponding to the second parameter in the target state data are calculated, including: Any one parameter in the target state data is taken as a first parameter, a first average value of the numerical value of the first parameter within the preset time period is calculated, a difference value between the numerical value of the first parameter corresponding to all time points within the preset time period and the first average value is taken as a first difference value, and the time point corresponding to the first parameter with the largest difference value is selected from a plurality of first difference values as a third time point. Any one parameter in the target state data except the first parameter is taken as a second parameter, and the above steps are repeated to calculate the fourth time point corresponding to the second parameter in the target state data.

4. The method of claim 1, wherein, The preset safety factor is set, including: The number of misjudgments of low-risk running state or high-risk running state in the last N months is collected, if misjudgment occurs for M times in succession, the safety factor is increased, otherwise it remains unchanged; if a high-risk running state is missed once in the last N months, the safety factor is decreased, otherwise it remains unchanged.

5. The method of claim 1, wherein, The running state of the wind turbine generator is determined, including: If the difference value between the measured value and the predicted value of each parameter in the state data is less than or equal to the difference reference value corresponding to the parameter, it is determined that the wind turbine generator is in a normal running state, and the current monitoring period of the wind turbine generator is continued to be maintained; If the difference value between the measured value and the predicted value of any one parameter in the state data is greater than the difference reference value corresponding to the parameter and less than or equal to twice the difference reference value corresponding to the parameter, it is determined that the wind turbine generator is in a low-risk running state, and a maintenance process is started; If the difference between the measured value and the predicted value of all parameters in the state data is greater than the difference reference value corresponding to the parameter or the difference value of any one parameter is greater than twice the difference reference value corresponding to the parameter, it is determined that the wind turbine is in a high-risk operating state, the wind turbine is controlled to shut down immediately, and an emergency maintenance instruction is sent to the person in charge.

6. The method of claim 1, wherein, The difference between the predicted state data and the measured state data is calculated, including: Any one parameter in the state data is taken as a fourth parameter, and a difference value of the fourth parameter is calculated by a formula , wherein, represents the difference value of the fourth parameter, represents a predicted value of the fourth parameter, and A represents a measured value of the fourth parameter; the step is repeated to calculate the difference between the measured value and the predicted value of all parameters in the state data.

7. An operating state risk assessment system for a wind power plant for implementing the operating state risk assessment method for a wind power plant according to any one of claims 1 to 6, wherein sensors are installed on each component of the wind power plant, characterized in that, The system comprises: A model construction unit is configured to construct a risk assessment model based on a neural network model, collect historical first operating data and corresponding state data under a normal operating state of the wind turbine, and train the risk assessment model; A data acquisition unit is configured to acquire real-time operating data of the wind turbine, acquire measured state data corresponding to the real-time operating data, input the real-time operating data into the risk assessment model, and output predicted state data corresponding to the real-time operating data from the risk assessment model; A risk assessment unit is configured to acquire a difference reference value corresponding to current environmental data, calculate the difference between the predicted state data and the measured state data, compare the difference with the difference reference value, and determine the operating state of the wind turbine, wherein the operating state comprises a normal operating state, a low-risk operating state, and a high-risk operating state.

Images