Fan shaft fault early warning method based on dynamic game optimization

Abstract

The present invention discloses a fan shaft fault warning method based on dynamic game optimization, comprising the following steps: S1. Collecting a monitoring data set of the fan shaft operating status; S2. Generating a set of characteristic vectors of the fan shaft operating status; S3. Constructing a fan shaft fault risk assessment model based on the characteristic vector set and combining a causal reasoning method; S4. Constructing a dynamic game optimization model for fan shaft fault warning based on the fan shaft fault risk assessment model; S5. Updating the fan shaft fault risk assessment model; S6. Generating a real-time fan shaft fault warning signal based on the updated fan shaft fault risk assessment model; and S7. Transmitting the fault warning signal to an intelligent operation and maintenance platform. The present invention uses an asymmetric reinforcement mechanism and dynamic causal weight updating to accurately identify the key driving factors of shaft faults and quantify the degree of influence of each factor on the fault risk.

Description

Technical Field

[0001] The present invention relates to the technical field of fans, and in particular to a fan shaft fault early warning method based on dynamic game optimization. Background Art

[0002] With the rapid development of wind power generation technology, the operational safety and economy of wind turbines have received increasing attention. As an important component of wind turbines, the operating status of the wind turbine shaft system is directly related to the reliability and service life of the entire machine. However, the wind turbine shaft system is exposed to complex dynamic loads and a changing environment for a long time and is easily affected by multiple factors such as vibration fatigue, mechanical wear and environmental changes, resulting in frequent failures of key components such as bearings, couplings and gears.

[0003] Currently, mainstream methods for fan shaft fault early warning rely on static data analysis or single-variable monitoring technologies. Static data analysis methods typically assess the operating status of the fan shaft through vibration analysis, temperature monitoring, or simple empirical models. Some vibration monitoring methods can use acceleration sensors to collect bearing vibration signals and extract fault characteristics through spectrum analysis. At the same time, temperature monitoring technology is also widely used to determine the operating status of bearings or gears. However, traditional technologies have significant limitations in practical applications:

[0004] On the one hand, static data analysis methods cannot fully reflect the complex interactive behavior of the wind turbine shaft system under dynamic operating conditions. For example, changes in wind speed, ambient humidity and temperature difference will have a significant impact on the operating status of the shaft system, but traditional methods find it difficult to include these dynamic factors in the evaluation scope, resulting in insufficient accuracy and adaptability of the early warning results.

[0005] On the other hand, most existing fault warning systems lack a closed-loop feedback mechanism and are unable to effectively use warning results to guide wind farm operation and maintenance decisions. This isolated warning model leads to irrational allocation of operation and maintenance resources, low maintenance timeliness, and may even cause additional downtime losses due to incorrect priority judgments.

[0006] In summary, the existing technology has significant deficiencies in dynamic environment adaptability, warning accuracy, and operation and maintenance decision support. There is an urgent need for a new fault warning method that can combine dynamic game optimization and causal reasoning to improve the accuracy, intelligence, and operation and maintenance efficiency of wind turbine shaft fault prediction. Summary of the Invention

[0007] One purpose of the present invention is to propose a fan shaft system fault early warning method based on dynamic game optimization. The present invention accurately identifies the key driving factors of shaft system failures and quantifies the impact of each factor on the failure risk through an asymmetric reinforcement mechanism and dynamic causal weight update.

[0008] According to an embodiment of the present invention, a method for early warning of a fan shaft fault based on dynamic game optimization includes the following steps:

[0009] S1. Collect monitoring data sets of the fan shaft operating status;

[0010] S2. Preprocess the monitoring data set to generate a set of feature vectors of the wind turbine shaft system operating status;

[0011] S3. Based on the feature vector set and combined with causal reasoning method, a fan shaft fault risk assessment model is constructed;

[0012] S4. Construct a dynamic game optimization model for fan shaft failure warning based on the fan shaft failure risk assessment model;

[0013] S5. Calibrate the parameters of the dynamic game optimization model using historical operating data and real-time monitoring data, dynamically adjust the weights of each causal variable in the wind turbine shaft failure risk assessment model based on the game optimization results, and update the wind turbine shaft failure risk assessment model;

[0014] S6. Generate a real-time wind turbine shaft failure warning signal based on the updated wind turbine shaft failure risk assessment model;

[0015] S7. Transmit the fault warning signal to the intelligent operation and maintenance platform, and combine the historical operation and maintenance data and real-time monitoring data in the intelligent operation and maintenance platform to generate repair priority recommendations and resource allocation plans for the wind turbine shaft system.

[0016] Optionally, the S1 includes the following steps:

[0017] S11. Using a vibration sensor to collect a vibration signal dataset V(t) of the fan shaft, the vibration signal dataset is obtained by recording the real-time vibration amplitude, frequency, and change trend of the fan bearing under the operating state;

[0018] S12. Using a temperature sensor to collect an operating temperature signal data set T(t) of the fan shaft, the operating temperature signal data set is obtained by recording temperature changes at key parts of the fan shaft;

[0019] S13. Using environmental monitoring equipment to collect environmental load parameter data set L(t) of the fan shaft system, the environmental load parameters include wind speed, air density and humidity;

[0020] S14. Synchronize the vibration signal dataset V(t), the operating temperature signal dataset T(t), and the environmental load parameter dataset L(t) to construct a monitoring dataset D(t) for the operating status of the fan shaft system:

[0021] D(t)={V(t),T(t),L(t)}.

[0022] Optionally, the S2 includes the following steps:

[0023] S21. Perform data denoising on the monitoring data set D(t), using a wavelet transform method to decompose the vibration signal data set and the operating temperature signal data set, retaining the characteristic frequency band and filtering out high-frequency noise to generate a denoised signal data set;

[0024] S22. Remove abnormal data from the denoised signal dataset by setting a threshold range to detect abnormal values and remove data points that deviate from the statistical range to generate a cleaned signal dataset;

[0025] S23. Perform feature extraction on the cleaned signal dataset. Use Fast Fourier Transform to extract the characteristic frequency of the vibration signal from the vibration signal dataset, extract the temperature change rate characteristics of key parts from the operating temperature signal dataset, and extract the statistical characteristics of wind speed and humidity changes from the environmental load parameter dataset to generate the feature set F(t):

[0026] F(t)={f v (t),f t (t),f l (t)};

[0027] Among them, f v (t) represents the vibration signal characteristics, f t (t) represents the temperature signal characteristics, f l (t) represents the environmental load characteristics;

[0028] S24. Perform multi-dimensional parameter normalization on the feature set F(t), use a standardization method to make the feature data dimensionless, and generate a feature vector set X(t) of the fan shaft system operating state after normalization:

[0029] X(t)={x1(t),x2(t),…,x n (t)};

[0030] Among them, x n (t) represents the nth normalized eigenvector at time t, and n is the total number of eigenvectors.

[0031] Optionally, S3 includes the following steps:

[0032] S31. Define a multi-layer nested causal inference variable set C based on the characteristic vector set X(t) of the fan shaft system operating state. m (t):

[0033] C m (t) = {C1(t), C2(t), C3(t)};

[0034] Among them, C1(t) represents the first-layer state variables, including the bearing vibration signal and operating temperature signal of the fan shaft system, C2(t) represents the second-layer external environmental variables, including environmental load parameters, and C3(t represents the third-layer potential fault variables, including the signal response characteristics of each fault mode in the fan shaft system;

[0035] S32. Using multi-layer causal reasoning structure to construct dynamic nested causal graph G m (C m ,R m ), the nested structure of the dynamic nested causal graph dynamically describes the temporal interaction process between the wind turbine shaft system operating status, environmental conditions and potential failure modes:

[0036] G m (C m ,R m )=(C m ,R m );

[0037] Among them, G m Represents a multi-layer causal variable set, R m Represents a multi-layer causal relationship set, where each layer of relationship R ij Describe the causal interaction from layer i to layer j;

[0038] S33. Dynamically update the relationship weights in the dynamic nested causal graph and introduce the time-sensitive causal weight function w through the change of wind turbine operation status. ij (t):

[0039] w ij (t)=f(ΔX(t),ΔE(t),T c );

[0040] Among them, ΔX(t) represents the change of the fan state, ΔE(t) represents the change of the environmental conditions, T c is a specific time window, and f represents a nonlinear function fitted by historical data;

[0041] S34. Asymmetric strengthening of causal relationships in dynamic nested causal graphs is performed by introducing the impulse response function φ ij (t) Describe the asymmetric impact of a unidirectional variable change on the overall system:

[0042]

[0043] Among them, β ij is the causal path influence coefficient, P(C j ∣C i ) is the conditional probability, describing the dependent variable C j With the independent variable C iProbability distribution of changes;

[0044] S35. Define a risk assessment model for fan shaft failure based on a dynamic nested causal graph

[0045]

[0046] in, It represents a multi-level fault risk function and comprehensively evaluates the impact of the wind turbine shaft system operating status on potential faults through dynamic nested causal relationships.

[0047] Optionally, the S4 includes the following steps:

[0048] S41. Establish a dynamic game optimization model Γ(t) for wind turbine shaft fault warning. The dynamic game optimization model is composed of the wind turbine shaft operating state, environmental load condition, strategy space, and multi-level fault risk function:

[0049]

[0050] Among them, P s Indicates the operating status of the fan shaft system, P e Indicates environmental load conditions, S s (t) represents the strategy set of the wind turbine shaft system operation status at time t, S e (t) represents the strategy set of the environmental load condition party at time t;

[0051] S42. For the dynamic game optimization model Γ(t), define the profit functions for the wind turbine shaft system operating state and the environmental load condition, and construct the dynamic objectives of both parties:

[0052]

[0053] Among them, U s (t) represents the profit function of the fan shaft system operating state, c s (t) represents the operation adjustment cost paid by the fan shaft system operating state at time t to reduce the failure risk, η s is the magnification factor of the strategic benefit of the other party, Ω s (S s (t)) represents the operating status of the fan shaft system in strategy S s (t) The additional benefit to the shafting performance, U e (t) represents the benefit function of the environmental load condition, Indicates that in the time window [tT c ,t] is the integral of the influence of failure risk on the change of environmental load parameter E(τ), ζ(·) is the mapping function that weighs the risk sensitivity, represents the partial derivative of the risk function with respect to the change of environmental load;

[0054] S43. Based on the profit function of both parties, a dynamic game optimization problem is established. In a given strategy space S s (t)×S e (t), the wind turbine shaft system operating state and the environmental load condition are made to minimize the failure risk through game interaction. The dynamic optimization problem is characterized as follows:

[0055]

[0056] The following Nash equilibrium conditions are met at the same time:

[0057]

[0058] in, It represents the optimal strategy of the wind turbine shaft system operating state and the environmental load condition when the game equilibrium is reached at time t;

[0059] S44. By monitoring the failure risk function Real-time changes, adjust the strategy space S s (t),S e (t), and based on the benefit function U of both parties s (t),U e (t) is iterated to modify the strategy, so that the dynamic game optimization model can dynamically migrate to a new strategy equilibrium point when the external environment changes or the operating state of the wind turbine shaft system changes abnormally:

[0060]

[0061] Among them, s (·) and Ψ e (·) represent the mapping functions for dynamically updating the strategies of the wind turbine shaft system operating status and the environmental load conditions, and Represents the gradient information for the profit function;

[0062] S45. Iteratively solve the dynamic game optimization problem to obtain the optimal strategy combination of the wind turbine shaft system operating state and environmental load conditions. And the failure risk level corresponding to the optimal strategy combination As a result of the assessment to minimize the risk of shaft failure in wind turbines:

[0063]

[0064] in, represents the failure risk value under the optimal strategy combination.

[0065] Optionally, the S5 includes the following steps:

[0066] S51. Arrange the historical operation data of the dynamic game optimization model Γ(t) and the real-time monitoring data D(t) into a time series form;

[0067] S52. Based on historical operation data and real-time monitoring data, the profit function parameter γ in the dynamic game optimization model is s ,α e and the causal weight function w ij (t) Perform parameter calibration and predict risk value by minimizing historical data and actual risk value Error calibration parameters;

[0068] S53. Based on the dynamic game optimization results and the calibrated parameters, adjust the weights of the causal variables in the fan shaft failure risk assessment model. The adjustment rules are as follows:

[0069]

[0070] Among them, λ1 is the adjustment step coefficient, and Represents the effect of the profit function on the causal variable C j The partial derivative of

[0071] S54. Based on the calibrated parameters and dynamically adjusted causal variable weights, the fan shaft system failure risk assessment model is updated, and the calibration and update process of S51-S54 is repeated to iteratively optimize the dynamic game optimization model and the fan shaft system failure risk assessment model until the error between the predicted fault risk value and the actual observed risk value converges to the preset threshold ε.

[0072] Optionally, the S6 includes the following steps:

[0073] S61. Define real-time fault risk classification rules based on the updated wind turbine shaft fault risk assessment model, dividing fault risks into normal operation state, early warning state, and abnormal alarm state. The classification rules are fitted based on historical operation data and actual alarm conditions;

[0074] S62. Utilize the real-time monitoring data of the wind turbine shaft system to assess the current fault risk level based on the wind turbine shaft system fault risk assessment model, and identify and quantify potential fault characteristics in the monitoring data using dynamically updated risk assessment parameters;

[0075] S63. Generate a fault status signal of the fan shaft system based on the real-time assessed fault risk value and classification rules. The fault status signal includes a normal operating status signal, an early warning signal, and an abnormal alarm signal;

[0076] S64. Optimize the generated fault status signal based on the game strategy results of the dynamic game optimization model, and adjust the sensitivity of the warning signal trigger and the response priority of the abnormal alarm signal according to the operating conditions;

[0077] S65. Output the optimized fault warning signal to the wind turbine operation and maintenance system in real time. The output content includes the fault status, current risk level, and recommended response strategy.

[0078] The beneficial effects of the present invention are:

[0079] (1) The present invention introduces a dynamic game optimization model to establish a game relationship between the wind turbine shaft system operating state and the environmental load condition, thereby realizing dynamic optimization of multi-party interaction strategies. It can adjust model parameters in real time and dynamically respond to changes in the external environment of wind speed and humidity and fluctuations in the shaft system operating state, significantly improving the adaptability and accuracy of fault warning. By introducing the profit function and Nash equilibrium conditions, the model can still reach the optimal state under complex operating conditions.

[0080] (2) The present invention uses causal reasoning methods to construct a multi-layer nested causal graph to deeply explore the causal relationship between the shaft system operating status, environmental conditions and potential fault signals. Compared with the traditional linear model that can only capture the correlation between variables, through the asymmetric reinforcement mechanism and dynamic causal weight update, the key driving factors of the shaft system failure are accurately identified, and the influence of each factor on the failure risk is quantified.

[0081] (3) The present invention transforms the dynamic game optimization results and real-time risk assessment results into targeted operation and maintenance suggestions by constructing an intelligent closed-loop operation and maintenance system, including fault priority sorting, maintenance resource allocation and dynamic early warning strategy adjustment. Compared with the isolated early warning process in the existing method, the present invention realizes closed-loop management of the entire life cycle based on the early warning of the fan shaft system fault, which greatly improves the maintenance efficiency and the utilization rate of operation and maintenance resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0083] Figure 1 This is a flow chart of a fan shaft fault early warning method based on dynamic game optimization proposed by the present invention. DETAILED DESCRIPTION

[0084] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0085] refer to Figure 1 A fan shaft fault early warning method based on dynamic game optimization includes the following steps:

[0086] S1. Collect monitoring data sets of the fan shaft operating status;

[0087] S2. Preprocess the monitoring data set to generate a set of feature vectors of the wind turbine shaft system operating status;

[0088] S3. Based on the feature vector set and combined with causal reasoning method, a fan shaft fault risk assessment model is constructed;

[0089] S4. Construct a dynamic game optimization model for fan shaft failure warning based on the fan shaft failure risk assessment model;

[0090] S5. Calibrate the parameters of the dynamic game optimization model using historical operating data and real-time monitoring data. Dynamically adjust the weights of the causal variables in the wind turbine shaft failure risk assessment model based on the game optimization results, and update the wind turbine shaft failure risk assessment model.

[0091] S6. Generate a real-time wind turbine shaft failure warning signal based on the updated wind turbine shaft failure risk assessment model;

[0092] S7. Transmit the fault warning signal to the intelligent operation and maintenance platform, and combine the historical operation and maintenance data and real-time monitoring data in the intelligent operation and maintenance platform to generate repair priority recommendations and resource allocation plans for the wind turbine shaft system.

[0093] In this embodiment, S1 includes the following steps:

[0094] S11. Use a vibration sensor to collect a vibration signal dataset V(t) of the fan shaft system. The vibration signal dataset is obtained by recording the real-time vibration amplitude, frequency, and change trend of the fan bearing under the operating state;

[0095] S12. Using a temperature sensor to collect an operating temperature signal data set T(t) of the fan shaft system, the operating temperature signal data set is obtained by recording temperature changes at key locations of the fan shaft system;

[0096] S13. Use environmental monitoring equipment to collect the environmental load parameter data set L(t) of the fan shaft system, where the environmental load parameters include wind speed, air density, and humidity;

[0097] S14. Synchronize the vibration signal dataset V(t), the operating temperature signal dataset T(t), and the environmental load parameter dataset L(t) to construct a monitoring dataset D(t) for the operating status of the fan shaft system:

[0098] D(t)={V(t),T(t),L(t)}.

[0099] In this embodiment, S2 includes the following steps:

[0100] S21. Perform data denoising on the monitoring data set D(t), using a wavelet transform method to decompose the vibration signal data set and the operating temperature signal data set, retaining the characteristic frequency band and filtering out high-frequency noise to generate a denoised signal data set;

[0101] S22. Remove abnormal data from the denoised signal dataset by setting a threshold range to detect abnormal values and remove data points that deviate from the statistical range to generate a cleaned signal dataset;

[0102] S23. Perform feature extraction on the cleaned signal dataset. Use Fast Fourier Transform to extract the characteristic frequency of the vibration signal from the vibration signal dataset, extract the temperature change rate characteristics of key parts from the operating temperature signal dataset, and extract the statistical characteristics of wind speed and humidity changes from the environmental load parameter dataset to generate the feature set F(t):

[0103] F(t)={f v (t),f t (t),f l (t)};

[0104] Among them, f v (t) represents the vibration signal characteristics, f t (t) represents the temperature signal characteristics, f l (t) represents the environmental load characteristics;

[0105] S24. Perform multi-dimensional parameter normalization on the feature set F(t), use a standardization method to make the feature data dimensionless, and generate a feature vector set X(t) of the fan shaft system operating state after normalization:

[0106] X(t)={x1(t),x2(t),…,x n (t)};

[0107] Among them, x n (t) represents the nth normalized eigenvector at time t, and n is the total number of eigenvectors.

[0108] In this embodiment, S3 includes the following steps:

[0109] S31. Define a multi-layer nested causal inference variable set C based on the characteristic vector set X(t) of the fan shaft system operating state. m (t):

[0110] C m (t) = {C1(t), C2(t), C3(t)};

[0111] Among them, C1(t) represents the first-layer state variables, including the bearing vibration signal and operating temperature signal of the fan shaft system, C2(t) represents the second-layer external environmental variables, including environmental load parameters, and C3(t represents the third-layer potential fault variables, including the signal response characteristics of each fault mode in the fan shaft system;

[0112] S32. Using multi-layer causal reasoning structure to construct dynamic nested causal graph G m (C m ,R m ), the nested structure of the dynamic nested causal graph dynamically describes the temporal interaction process between the wind turbine shaft system operating status, environmental conditions and potential failure modes:

[0113] G m (C m ,R m )=(C m ,R m );

[0114] Among them, G m Represents a multi-layer causal variable set, R m Represents a multi-layer causal relationship set, where each layer of relationship R ij Describe the causal interaction from layer i to layer j;

[0115] S33. Dynamically update the relationship weights in the dynamic nested causal graph and introduce the time-sensitive causal weight function w through the change of wind turbine operation status. ij (t):

[0116] w ij (t)=f(ΔX(t),ΔE(t),T c );

[0117] Among them, ΔX(t) represents the change of the fan state, ΔE(t) represents the change of the environmental conditions, T c is a specific time window, and f represents a nonlinear function fitted by historical data;

[0118] S34. Asymmetric strengthening of causal relationships in dynamic nested causal graphs is performed by introducing the impulse response function φ ij (t) Describe the asymmetric impact of a unidirectional variable change on the overall system:

[0119]

[0120] Among them, β ij is the causal path influence coefficient, P(C j ∣C i ) is the conditional probability, describing the dependent variable C j With the independent variable Ci Probability distribution of changes;

[0121] S35. Define a risk assessment model for fan shaft failure based on a dynamic nested causal graph

[0122]

[0123] in, It represents a multi-level fault risk function and comprehensively evaluates the impact of the wind turbine shaft system operating status on potential faults through dynamic nested causal relationships.

[0124] In this embodiment, S4 includes the following steps:

[0125] S41. Establish a dynamic game optimization model Γ(t) for wind turbine shaft system fault warning. The dynamic game optimization model is composed of the wind turbine shaft system operating state, environmental load condition, strategy space, and multi-level fault risk function:

[0126]

[0127] Among them, P s Indicates the operating status of the fan shaft system, P e Indicates environmental load conditions, S s (t) represents the strategy set of the wind turbine shaft system operation status at time t, S e (t) represents the strategy set of the environmental load condition party at time t;

[0128] S42. For the dynamic game optimization model Γ(t), define the profit functions for the wind turbine shaft system operating state and the environmental load condition, and construct the dynamic objectives of both parties:

[0129]

[0130] Among them, U s (t) represents the profit function of the fan shaft system operating state, c s (t) represents the operation adjustment cost paid by the fan shaft system operating state at time t to reduce the failure risk, η s is the magnification factor of the strategic benefit of the other party, Ω s (S s (t)) represents the operating status of the fan shaft system in strategy S s (t) The additional benefit to the shafting performance, U e (t) represents the benefit function of the environmental load condition, Indicates that in the time window [tT c,t] is the integral of the influence of failure risk on the change of environmental load parameter E(τ), ζ(·) is the mapping function that weighs the risk sensitivity, represents the partial derivative of the risk function with respect to the change of environmental load;

[0131] S43. Based on the profit function of both parties, a dynamic game optimization problem is established. In a given strategy space S s (t)×S e (t), the wind turbine shaft system operating state and the environmental load condition are made to minimize the failure risk through game interaction. The dynamic optimization problem is characterized as follows:

[0132]

[0133] The following Nash equilibrium conditions are met at the same time:

[0134]

[0135] in, It represents the optimal strategy of the wind turbine shaft system operating state and the environmental load condition when the game equilibrium is reached at time t;

[0136] S44. By monitoring the failure risk function Real-time changes, adjust the strategy space S s (t),S e (t), and based on the benefit function U of both parties s (t),U e (t) is iterated to modify the strategy, so that the dynamic game optimization model can dynamically migrate to a new strategy equilibrium point when the external environment changes or the operating state of the wind turbine shaft system changes abnormally:

[0137]

[0138] Among them, s (·) and Ψ e (·) represent the mapping functions for dynamically updating the strategies of the wind turbine shaft system operating status and the environmental load conditions, and Represents the gradient information for the profit function;

[0139] S45. Iteratively solve the dynamic game optimization problem to obtain the optimal strategy combination of the wind turbine shaft system operating state and environmental load conditions. And the failure risk level corresponding to the optimal strategy combination As a result of the assessment to minimize the risk of shaft failure in wind turbines:

[0140]

[0141] in, represents the failure risk value under the optimal strategy combination.

[0142] In this embodiment, S5 includes the following steps:

[0143] S51. Arrange the historical operation data of the dynamic game optimization model Γ(t) and the real-time monitoring data D(t) into a time series form;

[0144] S52. Based on historical operation data and real-time monitoring data, the profit function parameter γ in the dynamic game optimization model is s ,α e and the causal weight function w ij (t) Perform parameter calibration and predict risk value by minimizing historical data and actual risk value Error calibration parameters;

[0145] S53. Based on the dynamic game optimization results and the calibrated parameters, adjust the weights of the causal variables in the fan shaft failure risk assessment model. The adjustment rules are as follows:

[0146]

[0147] Among them, λ1 is the adjustment step coefficient, and Represents the effect of the profit function on the causal variable C j The partial derivative of

[0148] S54. Based on the calibrated parameters and dynamically adjusted causal variable weights, the fan shaft system failure risk assessment model is updated, and the calibration and update process of S51-S54 is repeated to iteratively optimize the dynamic game optimization model and the fan shaft system failure risk assessment model until the error between the predicted fault risk value and the actual observed risk value converges to the preset threshold ε.

[0149] In this embodiment, S6 includes the following steps:

[0150] S61. Define real-time fault risk classification rules based on the updated wind turbine shaft fault risk assessment model, classifying fault risks into normal operating state, early warning state, and abnormal alarm state. The classification rules are fitted based on historical operating data and actual alarm conditions.

[0151] S62. Utilize the real-time monitoring data of the wind turbine shaft system to assess the current fault risk level based on the wind turbine shaft system fault risk assessment model, and identify and quantify potential fault characteristics in the monitoring data using dynamically updated risk assessment parameters;

[0152] S63. Generate a fault status signal of the fan shaft system based on the real-time assessed fault risk value and classification rules. The fault status signal includes a normal operating status signal, an early warning signal, and an abnormal alarm signal;

[0153] S64. Optimize the generated fault status signal based on the game strategy results of the dynamic game optimization model, and adjust the sensitivity of the warning signal trigger and the response priority of the abnormal alarm signal according to the operating conditions;

[0154] S65. Output the optimized fault warning signal to the wind turbine operation and maintenance system in real time. The output content includes the fault status, current risk level, and recommended response strategy.

[0155] Example 1:

[0156] In the actual operation of a coastal wind farm, the shaft system of a wind turbine numbered W-35 was monitored from July 2024 to September 2024 to verify the actual effect of the present invention. The W-35 wind turbine is located in the center of the wind farm and is affected by wind speed changes and humid sea breeze on a daily basis. Its operating state is complex and has a high risk of failure.

[0157] On August 15, 2024, the system collected real-time monitoring data of the W-35 wind turbine, in which the average value of the main shaft vibration signal increased from the previous 0.15m / s 2 Significantly increased to 0.26m / s 2 The vibration spectrum characteristics showed a significant enhancement of low-frequency components. At the same time, the gearbox temperature slowly increased from the normal 65°C to 78°C. In addition, environmental load monitoring showed that the wind speed in the area increased from 12m / s to 16m / s within 24 hours from August 14 to August 15, accompanied by an increase in humidity from 70% to 88%.

[0158] The dynamic game optimization model evaluated the above data in real time and, combined with the causal reasoning model, generated an assessment result of a fault risk value of 0.82 (higher than the warning threshold of 0.8). Further analysis by the model showed that changes in the low-frequency components of the main shaft vibration signal were identified as potential characteristics of main bearing abnormalities, and there was a significant causal relationship between the increase in gearbox temperature and the increase in humidity, which may lead to an increased risk of poor shaft lubrication.

[0159] At 10:00 AM on August 16, 2024, the system automatically generated an early warning signal, recording the following information:

[0160] Fan No.: W-35

[0161] Trigger time: August 16, 2024, 10:00 AM

[0162] Warning level: Early warning

[0163] Risk Assessment Value: 0.82

[0164] Abnormal characteristics: The low frequency of the spindle vibration signal is enhanced, the gearbox temperature is increased, and the ambient humidity is high.

[0165] The warning signal was simultaneously transmitted to the wind farm's operations and maintenance management platform, which automatically recommended a maintenance strategy: inspect the W-35 turbine's main bearings, replenish lubricant, and monitor gearbox temperature and humidity trends. Following the recommendations, the operations team performed a maintenance inspection on the W-35 turbine at 4:00 PM that same day. The inspection revealed initial cracking of the main bearing inner ring and a decrease in lubricant quality due to humidity, but overall operational status remained stable.

[0166] The maintenance personnel repaired the main bearing online and replaced the lubricating oil. The total maintenance time was about 4 hours. The fan downtime was only from 16:00 to 20:00 on the same day. The maintenance cost was RMB 75,000.

[0167] In contrast, another wind turbine W-42 that did not adopt the method of the present invention suffered a serious main bearing failure on September 10, 2024, under similar operating conditions. After the failure occurred, the traditional early warning method only issued an abnormal alarm signal when the main bearing temperature rose to 95°C. The maintenance results showed that the main bearing of the W-42 wind turbine was completely damaged and had to be replaced. The maintenance time was as long as 3 days and the cost was as high as RMB 420,000.

[0168] It can be seen from this embodiment that the method of the present invention successfully identified the potential failure risk of the W-35 wind turbine and achieved early warning, avoiding the high repair costs and long downtime losses caused by serious damage to the main bearing. Traditional methods failed to provide effective warnings in similar situations, resulting in serious consequences of the failure. The method of the present invention significantly improves the accuracy and timeliness of wind turbine shaft system fault warnings, and has high application value in actual scenarios.

[0169] The present invention introduces a dynamic game optimization model to establish a game relationship between the wind turbine shaft system operating state and the environmental load conditions, thereby realizing dynamic optimization of multi-party interaction strategies. It can adjust model parameters in real time, dynamically respond to changes in the external environment of wind speed and humidity, and fluctuations in the shaft system operating state, significantly improving the adaptability and accuracy of fault warning. By introducing the profit function and Nash equilibrium conditions, the model can still reach the optimal state under complex operating conditions.

[0170] This paper uses causal reasoning methods to construct a multi-layer nested causal graph to deeply explore the causal relationship between the shaft system operating status, environmental conditions and potential fault signals. Compared with the traditional linear model that can only capture the correlation between variables, through the asymmetric reinforcement mechanism and dynamic causal weight update, it can accurately identify the key driving factors of shaft system failures and quantify the impact of each factor on the failure risk.

[0171] The present invention transforms dynamic game optimization results and real-time risk assessment results into targeted operation and maintenance suggestions by constructing an intelligent closed-loop operation and maintenance system, including fault priority sorting, maintenance resource allocation and dynamic early warning strategy adjustment. Compared with the isolated early warning process in the existing method, it realizes closed-loop management of the entire life cycle based on the early warning of the fan shaft system fault, greatly improving maintenance efficiency and the utilization rate of operation and maintenance resources.

[0172] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A wind turbine shaft fault early warning method based on dynamic game optimization, characterized in that: The steps include: S1. Collect monitoring data sets of the fan shaft operating status; S2. Preprocess the monitoring data set to generate a set of feature vectors of the wind turbine shaft system operating status; S3. Based on the feature vector set and combined with causal reasoning method, a fan shaft fault risk assessment model is constructed; The S3 includes the following steps: S31. Define a multi-layer nested causal inference variable set C based on the characteristic vector set X(t) of the fan shaft system operating state. m (t): C m (t)={C1(t),C2(t),C3(t)}; Among them, C1(t) represents the first-level state variables, including the bearing vibration signal and operating temperature signal of the fan shaft system; C2(t) represents the second-level external environmental variables, including environmental load parameters; C3(t) represents the third-level potential fault variables, including the signal response characteristics of each fault mode in the fan shaft system; S32. Using multi-layer causal reasoning structure to construct dynamic nested causal graph G m (C m ,R m ), the nested structure of the dynamic nested causal graph dynamically describes the temporal interaction process between the wind turbine shaft system operating status, environmental conditions and potential failure modes: G m (C m ,R m )=(C m ,R m ); Among them, G m Represents a multi-layer causal variable set, R m Represents a multi-layer causal relationship set, where each layer of relationship R ij Describe the causal interaction from layer i to layer j; S33. Dynamically update the relationship weights in the dynamic nested causal graph and introduce the time-sensitive causal weight function w through the change of wind turbine operation status. ij (t): w ij (t)=f(ΔX(t),ΔE(t),T c ); Among them, ΔX(t) represents the change of the fan state, ΔE(t) represents the change of the environmental conditions, T c is a specific time window, and f represents a nonlinear function fitted by historical data; S34. Asymmetric strengthening of causal relationships in dynamic nested causal graphs is performed by introducing the impulse response function φ ij (t) Describe the asymmetric impact of a unidirectional variable change on the overall system: Among them, β ij is the causal path influence coefficient, P(C j ∣C i ) is the conditional probability, describing the dependent variable C j With the independent variable C i Probability distribution of changes; S35. Define a risk assessment model for fan shaft failure based on a dynamic nested causal graph in, Representing a multi-level fault risk function, the impact of the wind turbine shaft system operating status on potential faults is comprehensively evaluated through dynamic nested causal relationships; S4. Construct a dynamic game optimization model for fan shaft failure warning based on the fan shaft failure risk assessment model; S5. Calibrate the parameters of the dynamic game optimization model using historical operating data and real-time monitoring data, dynamically adjust the weights of each causal variable in the wind turbine shaft failure risk assessment model based on the game optimization results, and update the wind turbine shaft failure risk assessment model; S6. Generate a real-time wind turbine shaft failure warning signal based on the updated wind turbine shaft failure risk assessment model; S7. Transmit the fault warning signal to the intelligent operation and maintenance platform, and combine the historical operation and maintenance data and real-time monitoring data in the intelligent operation and maintenance platform to generate repair priority recommendations and resource allocation plans for the wind turbine shaft system.

2. The wind turbine shaft fault early warning method based on dynamic game optimization according to claim 1 is characterized in that: Said S1 comprises the following steps: S11. Using a vibration sensor to collect a vibration signal dataset V(t) of the fan shaft, the vibration signal dataset is obtained by recording the real-time vibration amplitude, frequency, and change trend of the fan bearing under the operating state; S12. Using a temperature sensor to collect an operating temperature signal data set T(t) of the fan shaft, the operating temperature signal data set is obtained by recording temperature changes at key parts of the fan shaft; S13. Using environmental monitoring equipment to collect environmental load parameter data set L(t) of the fan shaft system, the environmental load parameters include wind speed, air density and humidity; S14. Synchronize the vibration signal dataset V(t), the operating temperature signal dataset T(t), and the environmental load parameter dataset L(t) to construct a monitoring dataset D(t) for the operating status of the fan shaft system: D(t)={V(t),T(t),L(t)}.

3. The wind turbine shaft fault early warning method based on dynamic game optimization according to claim 1 is characterized in that: The S2 comprises the following steps: S21. Perform data denoising on the monitoring data set D(t), using a wavelet transform method to decompose the vibration signal data set and the operating temperature signal data set, retaining the characteristic frequency band and filtering out high-frequency noise to generate a denoised signal data set; S22. Remove abnormal data from the denoised signal dataset by setting a threshold range to detect abnormal values and remove data points that deviate from the statistical range to generate a cleaned signal dataset; S23. Perform feature extraction on the cleaned signal dataset. Use Fast Fourier Transform to extract the characteristic frequency of the vibration signal from the vibration signal dataset, extract the temperature change rate characteristics of key parts from the operating temperature signal dataset, and extract the statistical characteristics of wind speed and humidity changes from the environmental load parameter dataset to generate the feature set F(t): F(t)={f v (t),f t (t),f l (t)}; Among them, f v (t) represents the vibration signal characteristics, f t (t) represents the temperature signal characteristics, f l (t) represents the environmental load characteristics; S24. Perform multi-dimensional parameter normalization on the feature set F(t), use a standardization method to make the feature data dimensionless, and generate a feature vector set X(t) of the fan shaft system operating state after normalization: X(t)={x1(t),x2(t),…,x n (t)}; Among them, x n (t) represents the nth normalized eigenvector at time t, and n is the total number of eigenvectors.

4. The wind turbine shaft fault early warning method based on dynamic game optimization according to claim 1 is characterized in that: The S4 comprises the following steps: S41. Establish a dynamic game optimization model Γ(t) for wind turbine shaft fault warning. The dynamic game optimization model is composed of the wind turbine shaft operating state, environmental load condition, strategy space, and multi-level fault risk function: Among them, P s Indicates the operating status of the fan shaft system, P e Indicates environmental load conditions, S s (t) represents the strategy set of the wind turbine shaft system operation status at time t, S e (t) represents the strategy set of the environmental load condition party at time t; S42. For the dynamic game optimization model Γ(t), define the profit functions for the wind turbine shaft system operating state and the environmental load condition, and construct the dynamic objectives of both parties: Among them, U s (t) represents the profit function of the fan shaft system operating state, c s (t) represents the operation adjustment cost paid by the fan shaft system operating state at time t to reduce the failure risk, η s is the magnification factor of the strategic benefit of the other party, Ω s (S s (t)) represents the operating status of the fan shaft system in strategy S s (t) The additional benefit to the shafting performance, U e (t) represents the benefit function of the environmental load condition, Indicates that in the time window [tT c ,t] is the integral of the influence of failure risk on the change of environmental load parameter E(τ), ζ(·) is the mapping function that weighs the risk sensitivity, represents the partial derivative of the risk function with respect to the change of environmental load; S43. Based on the profit function of both parties, a dynamic game optimization problem is established. In a given strategy space S s (t)×S e (t), the wind turbine shaft system operating state and the environmental load condition are made to minimize the failure risk through game interaction. The dynamic optimization problem is characterized as follows: The following Nash equilibrium conditions are met at the same time: in, It represents the optimal strategy of the wind turbine shaft system operating state and the environmental load condition when the game equilibrium is reached at time t; S44. By monitoring the failure risk function Real-time changes, adjust the strategy space S s (t),S e (t), and based on the benefit function U of both parties s (t),U e (t) is iterated to make strategy corrections, so that the dynamic game optimization model can dynamically migrate to a new strategy equilibrium point when the external environment changes or the operating state of the wind turbine shaft system changes abnormally: Among them, s (·) and Ψ e (·) represent the mapping functions for dynamically updating the strategies of the wind turbine shaft system operating status and the environmental load conditions, )and Represents the gradient information for the profit function; S45. Iteratively solve the dynamic game optimization problem to obtain the optimal strategy combination of the wind turbine shaft system operating state and environmental load conditions. And the failure risk level corresponding to the optimal strategy combination As a result of the assessment to minimize the risk of shaft failure in wind turbines: in, represents the failure risk value under the optimal strategy combination.

5. The wind turbine shaft fault early warning method based on dynamic game optimization according to claim 1 is characterized in that: The S5 comprises the following steps: S51. Arrange the historical operation data of the dynamic game optimization model Γ(t) and the real-time monitoring data D(t) into a time series form; S52. Based on historical operation data and real-time monitoring data, the profit function parameter γ in the dynamic game optimization model is s ,α e and the causal weight function w ij (t) Perform parameter calibration and predict risk value by minimizing historical data and actual risk value Error calibration parameters; S53. Based on the dynamic game optimization results and the calibrated parameters, adjust the weights of the causal variables in the fan shaft failure risk assessment model. The adjustment rules are as follows: Among them, λ1 is the adjustment step coefficient, and Represents the effect of the profit function on the causal variable C j The partial derivative of S54. Based on the calibrated parameters and dynamically adjusted causal variable weights, the fan shaft system failure risk assessment model is updated, and the calibration and update process of S51-S54 is repeated to iteratively optimize the dynamic game optimization model and the fan shaft system failure risk assessment model until the error between the predicted fault risk value and the actual observed risk value converges to the preset threshold ε.

6. The wind turbine shaft fault early warning method based on dynamic game optimization according to claim 1 is characterized in that: The S6 comprises the following steps: S61. Define real-time fault risk classification rules based on the updated wind turbine shaft fault risk assessment model, dividing fault risks into normal operation state, early warning state, and abnormal alarm state. The classification rules are fitted based on historical operation data and actual alarm conditions; S62. Utilize the real-time monitoring data of the wind turbine shaft system to assess the current fault risk level based on the wind turbine shaft system fault risk assessment model, and identify and quantify potential fault characteristics in the monitoring data using dynamically updated risk assessment parameters; S63. Generate a fault status signal of the fan shaft system based on the real-time assessed fault risk value and classification rules. The fault status signal includes a normal operating status signal, an early warning signal, and an abnormal alarm signal; S64. Optimize the generated fault status signal based on the game strategy results of the dynamic game optimization model, and adjust the sensitivity of the warning signal trigger and the response priority of the abnormal alarm signal according to the operating conditions; S65. Output the optimized fault warning signal to the wind turbine operation and maintenance system in real time. The output content includes the fault status, current risk level, and recommended response strategy.

Images