Intelligent fan early warning method and system based on self-adaptive scheme and computer equipment

Abstract

The invention relates to the technical field of fan early warning, in particular to a fan intelligent early warning method and system based on a self-adaptive scheme and computer equipment. Multi-frequency-band heterogeneous data are collected in real time through multiple sensors, and four-dimensional tensor data are constructed. And carrying out single-factor evaluation on the environmental risk and the fan state risk. And fusing the two types of risks by using a double-layer fusion mechanism to obtain a fusion risk value. And setting different risk probability early warning thresholds according to fan operation requirements, comparing the fused risk value with the different risk probability early warning thresholds, and judging the state of the fan. And once the risk value exceeds the threshold value, identifying the high-risk data cluster, associating the fault mode library to determine a potential risk mode, and triggering a corresponding early warning signal so as to timely process and ensure the safe operation of the fan. The real-time and accurate early warning of the potential risk of the fan is realized, and the safety and reliability of the operation of the fan are improved.

Description

Technical Field

[0001] This invention relates to the field of wind turbine early warning technology, specifically to a wind turbine intelligent early warning method, system, and computer equipment based on an adaptive scheme. Background Technology

[0002] With the increasing global demand for clean energy, wind power, as an important renewable energy source, is playing an increasingly important role in the energy structure. As the core equipment of wind power generation, the stable and reliable operation of wind turbines is crucial for ensuring the stability and economic efficiency of power supply. However, wind turbines are typically installed in complex natural environments, such as mountainous areas and offshore locations, facing various harsh weather conditions and complex geographical environments, which presents numerous challenges to their operation.

[0003] Traditional wind turbine fault detection methods mainly include two strategies: reactive maintenance and scheduled maintenance. Reactive maintenance involves repairing the turbine only after a fault occurs. While this method is simple, it leads to prolonged downtime, resulting in significant power loss and high repair costs. Scheduled maintenance involves inspecting and maintaining the turbine at fixed intervals. Although this method can prevent faults to some extent, the lack of real-time monitoring of the turbine's actual operating status can lead to over-maintenance or under-maintenance.

[0004] While existing wind turbine early warning technologies have improved the timeliness of fault detection to some extent, they still have some shortcomings. For example, some early warning technologies only focus on the operating parameters of the wind turbine, while ignoring the impact of surrounding environmental factors on the wind turbine; when processing multi-source heterogeneous data, they fail to fully explore the correlation between data, resulting in inaccurate early warning results; and they lack simulation and analysis of the risk propagation process, making it impossible to predict potential failure modes in advance. Summary of the Invention

[0005] The purpose of this invention is to provide a wind turbine intelligent early warning method, system, and computer equipment based on an adaptive scheme, so as to solve the problems of insufficient comprehensive consideration of the environment and its own status and low warning accuracy in the existing wind turbine early warning technology, and to realize real-time and accurate early warning of potential risks of wind turbines, thereby improving the safety and reliability of wind turbine operation.

[0006] The specific technical solution of the present invention is as follows:

[0007] One of the technical solutions of this invention is to provide a wind turbine intelligent early warning method based on an adaptive scheme, comprising:

[0008] Real-time acquisition of multi-band heterogeneous data from wind turbines, spatiotemporal alignment of the multi-band heterogeneous data, and construction of four-dimensional tensor data;

[0009] Based on four-dimensional tensor data, a single-factor risk assessment was conducted on environmental risk and wind turbine self-state risk; the aerodynamic stability coefficient of the wind turbine area, the external risk coefficient of the wind turbine, the abnormal operating state coefficient of the wind turbine, and the failure risk coefficient of wind turbine components were calculated respectively.

[0010] Environmental risk assessment includes:

[0011] Computational fluid dynamics simulation analysis was used to analyze environmental data in four-dimensional tensor data to calculate the aerodynamic stability coefficient of the area where the wind turbine is located.

[0012] Based on the dynamic obstacle data and wind turbine protection capability coefficient in the four-dimensional tensor data, a distance attenuation factor and a velocity amplification factor are introduced to generate the external risk coefficient of the wind turbine.

[0013] The risk assessment of the wind turbine's own condition includes:

[0014] A multivariate temporal graph convolutional network is used to analyze the topological correlation of wind turbine operation data in four-dimensional tensor data and output the anomaly coefficient of operation status.

[0015] Feature extraction is performed on state data in four-dimensional tensor data, and historical fault knowledge graphs are associated to assess the fault risk coefficient of wind turbine components.

[0016] The weights of environmental risks and wind turbine status risks are dynamically integrated to generate a combined risk value;

[0017] By setting risk probability warning thresholds and comparing risk values ​​with risk probability warning thresholds, high-risk data clusters are identified and associated with a fault mode library to identify potential risk modes and trigger warning signals of different levels.

[0018] As a further aspect of the present invention: the multi-band heterogeneous data includes wind turbine operation data, wind turbine status data, environmental data, and dynamic obstacle data;

[0019] The four-dimensional tensor feature constructed using the multi-band heterogeneous data includes: the first dimension representing time, the second dimension representing data type, corresponding to wind turbine operation data, status data, environmental data, and dynamic obstacle data, respectively, the third dimension representing the specific parameters in each data type, and the fourth dimension representing the spatial dimension.

[0020] As a further aspect of the present invention: the aerodynamic stability coefficient of the area where the wind turbine is located is calculated by means of:

[0021] A three-dimensional flow field model of the area where the wind turbine is located is constructed, and the three-dimensional flow field model is numerically solved to obtain the velocity vector distribution of the flow field around the wind turbine;

[0022] Calculate the velocity gradient tensor S based on the velocity vector distribution. xyzIts formula is:

[0023]

[0024] Where x, y, z correspond to the three directions of spatial coordinates, u x u y u z It is a velocity vector Components in the x, y, z directions This represents the partial derivative of the velocity component in the corresponding coordinate direction;

[0025] Solve the characteristic equation: det(S) xyz -λI)=0, obtain the eigenvalues ​​λ1, λ2, λ3 of the velocity gradient tensor;

[0026] Through the formula: K aero =|λ1|+|λ2|+|λ3|, and the aerodynamic stability coefficient K is obtained by calculation. aero .

[0027] As a further aspect of the present invention: the method for calculating the external risk factor of the wind turbine includes:

[0028] Calculate the risk contribution r of the i-th obstacle to the wind turbine. i The calculation formula is:

[0029]

[0030] Among them, v i Let d be the velocity of the i-th obstacle. i Let s be the current distance between the i-th obstacle and the wind turbine. i p is the equivalent size of the i-th obstacle; defense This refers to the wind turbine's own protection capability coefficient. For distance attenuation factor, Where α is the speed amplification factor, β is the distance attenuation factor, and β is the speed amplification factor.

[0031] External risk factor K of wind turbine ex The calculation formula is:

[0032]

[0033] As a further aspect of the present invention: the method for calculating the abnormal operating state coefficient includes:

[0034] Wind turbine operation data is extracted from four-dimensional tensor data, and a multivariate time series graph convolutional network is trained using time series graph data;

[0035] The real-time collected wind turbine operation data is used to construct a time series graph, which is then input into a trained MTGCN model. The output is a score S representing the degree of difference between the current operating state and the normal state. p ;

[0036] Operational status anomaly coefficient K ab The calculation formula is:

[0037]

[0038] Where τ is the score S p The threshold for comparison.

[0039] As a further aspect of the present invention: the method for calculating the failure risk coefficient of wind turbine components includes:

[0040] State data is extracted from four-dimensional tensor data, time-domain features are extracted based on the state data, and the time-domain features are transformed to the frequency domain through Fourier transform.

[0041] The extracted current state data features are matched with the fault features in the historical fault knowledge graph to obtain the similarity sim between the current state data features and the historical fault features.

[0042] Failure risk coefficient K com The calculation formula is:

[0043] K com = f(F)·sim;

[0044] F represents the historical fault with the highest similarity to the current state data features, and its occurrence frequency is f(F).

[0045] As a further aspect of the present invention: the fusion risk value calculation step includes:

[0046] Define the first-layer fusion mechanism, that is, for the aerodynamic stability coefficient K aero External risk coefficient K of the wind turbine ex Operational status anomaly coefficient K ab And the risk factor K of the fan component failure com Calculate the fusion weights separately, using the following formula:

[0047]

[0048]

[0049] Where, ω aero ω ex ω ab and ω comThese are the weights for aerodynamic stability coefficient, external risk coefficient, abnormal operating condition coefficient, and fault risk coefficient of wind turbine components, respectively. γ1 is an adjustment parameter used to adjust the allocation of fusion weights, and e is an exponential function.

[0050] Define a second-layer fusion mechanism, which involves calculating the environmental risk fusion weight ω separately. env The weighted ω of wind turbine status risk integration state The calculation formula is:

[0051]

[0052] Where, ω env and ω state These are the environmental risk fusion weights and the wind turbine condition risk fusion weights, respectively. γ2 is an adjustment parameter used to adjust the allocation of fusion weights, and e is an exponential function.

[0053] The environmental risk and wind turbine condition risk are integrated, and the formula for calculating the integrated risk value is as follows:

[0054] R fused =ω env (ω aero K aero +ω ex K ex )+ω state (ω ab K ab +ω com K com );

[0055] Among them, R fused To integrate risk values.

[0056] As a further aspect of the present invention: the set risk probability warning threshold is: a low-risk warning threshold R low Medium-risk warning threshold R mid and high-risk warning threshold R high ;

[0057] The method for comparing the fused risk value with the risk probability early warning threshold is as follows:

[0058] If R fused ≤R low If so, it is determined that the fan is currently in normal operating condition;

[0059] If R low <R fused ≤R mid If so, the wind turbine is determined to be in a low-risk state;

[0060] If R mid <R fused ≤Rhigh If so, the wind turbine is determined to be in a medium-risk state;

[0061] If R fused >R high If so, the wind turbine is determined to be in a high-risk state.

[0062] The second technical solution of the present invention is to provide a wind turbine intelligent early warning system based on an adaptive scheme, comprising:

[0063] Data acquisition and processing module: used to acquire multi-band heterogeneous data of wind turbine in real time. The multi-band heterogeneous data includes wind turbine operation data, status data, environmental data and dynamic obstacle data. The module also performs time alignment on the multi-band heterogeneous data to construct four-dimensional tensor data.

[0064] Single-factor risk assessment module: connected to the data acquisition and processing module, used to perform single-factor risk assessment on environmental risk and wind turbine self-state risk based on four-dimensional tensor data, and calculate the aerodynamic stability coefficient of the wind turbine area, the external risk coefficient of the wind turbine, the abnormal operating state coefficient of the wind turbine and the failure risk coefficient of wind turbine components respectively.

[0065] Risk fusion module: connected to the single-factor risk assessment module, used to dynamically fuse the weights of environmental risk and wind turbine status risk to generate a fused risk value;

[0066] Risk warning module: Connected to the risk fusion module, it is used to set a risk probability warning threshold, compare the fused risk value with the risk probability warning threshold, identify high-risk data clusters and associate them with the fault mode library, identify potential risk modes, and trigger warning signals of different levels.

[0067] The third technical solution of the present invention provides a computer device, the computer device comprising: a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, wherein the at least one instruction, the at least one program, the code set or instruction set is loaded and executed by the processor to realize the wind turbine intelligent early warning method based on an adaptive scheme as described in the first technical solution.

[0068] The beneficial effects of the technical solutions provided in this application include at least the following:

[0069] Beneficial effects:

[0070] This invention comprehensively collects wind turbine operation data, status data, environmental data, and dynamic obstacle data to construct four-dimensional tensor data and conduct single-factor risk assessment. It can comprehensively consider various factors affecting wind turbine operation, and accurately assess multiple dimensions such as aerodynamic stability, external collision risk, probability of abnormal operation status, and component failure risk. It overcomes the limitations of traditional methods that only focus on a single factor or some factors, greatly improves the accuracy and comprehensiveness of risk assessment, and provides a reliable basis for the safe and stable operation of wind turbines.

[0071] By dynamically integrating environmental risk and wind turbine status risk weights to generate a fused risk value, the system can automatically adjust the importance of different risk factors based on the real-time operating status of the wind turbine and environmental changes. Under severe weather conditions, it automatically increases the environmental risk weight, accurately reflecting the current actual risk level. Compared to fixed-weight assessment methods, this approach is more flexible and adaptable, effectively improving the timeliness and reliability of risk assessment.

[0072] By setting a risk probability warning threshold and comparing it with the integrated risk value, high-risk data clusters are quickly identified. This is then correlated with a fault mode library to determine potential risk patterns and trigger warning signals. This process enables timely detection of potential risks, provides graded warnings based on risk severity, helps maintenance personnel to be aware of and quickly locate problems, buys time for targeted measures, effectively reduces the probability of failures, minimizes downtime and maintenance costs, ensures continuous and stable wind turbine operation, and improves the economic efficiency and reliability of the wind power system. Attached Figure Description

[0073] Figure 1 This is a schematic diagram of the overall process of the wind turbine intelligent early warning method based on an adaptive scheme;

[0074] Figure 2 A detailed flowchart of the steps in the wind turbine intelligent early warning method S100 based on an adaptive scheme;

[0075] Figure 3 A detailed flowchart of the steps in the S200 method for intelligent early warning of wind turbines based on an adaptive scheme;

[0076] Figure 4 A detailed flowchart of the steps in the S300 intelligent early warning method for wind turbines based on an adaptive scheme;

[0077] Figure 5 A detailed flowchart of the steps of the S400 intelligent early warning method for wind turbines based on an adaptive scheme. Detailed Implementation

[0078] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0079] Traditional wind turbine fault detection methods are mostly based on reactive maintenance or periodic maintenance strategies, which cannot predict potential faults in a timely and accurate manner. Existing early warning technologies suffer from insufficient comprehensive consideration of environmental factors and the turbine's own condition, simplistic data processing methods, and low early warning accuracy. For example, when assessing wind turbine risk, they often only focus on the turbine's operational data, ignoring the impact of dynamic changes in the surrounding environment; or when processing multi-source heterogeneous data, they fail to fully explore the correlations between the data, leading to inaccurate early warning results.

[0080] To resolve the above issues, please refer to [link / reference]. Figure 1 This illustrates an embodiment of the present invention providing a wind turbine intelligent early warning method based on an adaptive scheme, the method comprising:

[0081] S100: Real-time acquisition of multi-band heterogeneous data from wind turbines, spatiotemporal alignment of the multi-band heterogeneous data, and construction of four-dimensional tensor data.

[0082] S200: Based on four-dimensional tensor data, single-factor risk assessments are conducted on environmental risks and wind turbine self-state risks; the aerodynamic stability coefficient of the wind turbine area, the external risk coefficient of the wind turbine, the abnormal operating state coefficient of the wind turbine, and the failure risk coefficient of wind turbine components are calculated respectively.

[0083] Environmental risk assessment includes:

[0084] Computational fluid dynamics simulation analysis was used to analyze environmental data in four-dimensional tensor data to calculate the aerodynamic stability coefficient of the area where the wind turbine is located.

[0085] Based on the dynamic obstacle data and wind turbine protection capability coefficient in the four-dimensional tensor data, a distance attenuation factor and a velocity amplification factor are introduced to generate the external risk coefficient of the wind turbine.

[0086] The risk assessment of the wind turbine's own condition includes:

[0087] A multivariate temporal graph convolutional network is used to analyze the topological correlation of wind turbine operation data in four-dimensional tensor data and output the anomaly coefficient of operation status.

[0088] Feature extraction is performed on state data in four-dimensional tensor data, and historical fault knowledge graphs are associated to assess the fault risk coefficient of wind turbine components.

[0089] S300: Dynamically integrates the weights of environmental risks and wind turbine status risks to generate a combined risk value.

[0090] S400: Set risk probability warning thresholds, identify high-risk data clusters and associate them with the fault mode library by merging risk values ​​and comparing them with risk probability warning thresholds, identify potential risk modes, and trigger warning signals of different levels.

[0091] The adaptive intelligent early warning method for wind turbines can effectively improve the safety and stability of wind turbine operation, detect potential failure risks in advance, and reduce losses caused by failures. This method mainly includes four core steps: real-time data acquisition to construct a four-dimensional tensor, single-factor risk assessment, risk fusion and digital twin modeling, and risk identification and early warning. Each step will be elaborated in detail below.

[0092] The S100 collects heterogeneous data across multiple frequency bands, including wind turbine operation, status, environment, and dynamic obstacles, through various sensors. It then performs time and space alignment to construct four-dimensional tensor data containing time, data type, specific parameters, and spatial dimensions, providing a foundation for subsequent analysis.

[0093] Please refer to Figure 2 The diagram illustrates a flowchart of an exemplary wind turbine intelligent early warning method S100 based on an adaptive scheme, the contents of which include:

[0094] S110: Real-time acquisition of multi-band heterogeneous data from wind turbines, including wind turbine operation data, wind turbine status data, environmental data, and dynamic obstacle data.

[0095] Wind turbine operation data is acquired in real time using sensors installed at various key components of the wind turbine, such as speed sensors, power sensors, and oil temperature sensors. The operation data includes speed, power, voltage, and current, with a sampling frequency of 10Hz.

[0096] The wind turbine status data includes bearing temperature, gearbox oil temperature, blade stress, etc., with a sampling frequency of 5Hz.

[0097] Environmental data include wind speed, wind direction, temperature, and humidity, with a sampling frequency of 1 Hz.

[0098] Dynamic obstacle data is collected by radar or cameras to capture the position and speed of obstacles such as birds and drones, with a sampling frequency of 20Hz.

[0099] After collecting heterogeneous data across multiple frequency bands, the data is filled with missing values ​​and subjected to noise filtering.

[0100] S120: Multi-band heterogeneous data time and space alignment processing.

[0101] For time alignment, a precise timestamp is added to each set of acquired data, based on a high-precision clock. In one possible implementation, time alignment is performed using interpolation or resampling methods for data acquired at different frequencies.

[0102] For spatial alignment, the precise position and orientation of each sensor within the wind turbine system are determined, and a unified spatial coordinate system is established. For example, the center of the bottom of the wind turbine tower is used as the origin, the tower's axial direction is the z-axis, the direction perpendicular to the tower and pointing towards the wind direction is the x-axis, and the y-axis is determined using the right-hand rule. For the collected dynamic obstacle data, coordinate transformation is performed within the unified spatial coordinate system based on its distance and orientation information to align it with the spatial position of the wind turbine.

[0103] S130: Construct four-dimensional tensor data.

[0104] Based on the collected data types, a four-dimensional tensor is constructed. The first dimension represents time, reflecting the changes in data over time; the second dimension represents the data type, corresponding to wind turbine operation data, status data, environmental data, and dynamic obstacle data, respectively; the third dimension represents the specific parameters in each data type, such as the speed and power parameters in wind turbine operation data; the fourth dimension represents the spatial dimension, used to describe the location information of the data in the wind turbine's spatial structure, such as the installation position of vibration sensors on components.

[0105] The spatiotemporally aligned data is filled into the tensor according to the corresponding dimensions and order.

[0106] For example, at time t j The collected fan speed n j Its position in the tensor is T(t) j ,1,1,k), where represents a T-dimensional tensor, and the first index t j Corresponding to the time dimension, the second index 1 indicates the data type of the wind turbine operation, the third index 1 indicates the speed parameter, and the fourth index k indicates the position number of the speed sensor in the wind turbine spatial structure.

[0107] S200 performs single-factor assessments of both environmental risk and the wind turbine's own condition risk. The environmental risk assessment yields the aerodynamic stability coefficient and the external risk coefficient of the wind turbine; the wind turbine's own condition risk assessment yields the operational anomaly coefficient and the wind turbine component failure risk coefficient.

[0108] Please refer to Figure 3 The diagram illustrates a flowchart of an exemplary wind turbine intelligent early warning method S200 based on an adaptive scheme, the contents of which include:

[0109] S210: Conduct single-factor risk assessment of environmental risks based on four-dimensional tensor data.

[0110] Environmental risk assessment includes:

[0111] Computational fluid dynamics simulation analysis is used to analyze environmental data in four-dimensional tensor data to calculate the aerodynamic stability coefficient of the area where the wind turbine is located.

[0112] In one possible calculation method, the aerodynamic stability coefficient of the area where the wind turbine is located is calculated as follows:

[0113] Using computational fluid dynamics software, a three-dimensional flow field model of the wind turbine's location is constructed based on the turbine's actual geometry and dimensions, as well as surrounding topographic information. Boundary conditions are set in the model, including wind speed and direction at the inlet boundary and pressure conditions at the outlet boundary. The constructed three-dimensional flow field model of the wind turbine's location is numerically solved using either a k-ε or k-ω turbulence model. Through iterative calculations, the velocity vector distribution of the flow field around the wind turbine is obtained.

[0114] Calculate the velocity gradient tensor S based on the velocity vector distribution. xyz Its formula is:

[0115]

[0116] Where x, y, z correspond to the three directions of spatial coordinates, u x u y u z It is a velocity vector Components in the x, y, z directions This represents the partial derivative of the velocity component with respect to the corresponding coordinate direction. This formula can be used to describe the rate and direction of velocity change within the flow field.

[0117] After obtaining the velocity gradient tensor S xyz Solve for its eigenvalues ​​λ1, λ2, and λ3, that is, solve the characteristic equation det(S). xyz -λI)=0 is used to obtain the result, where det represents determinant operation and I is the identity matrix.

[0118] Aerodynamic stability coefficient K aero Calculated from eigenvalues:

[0119] K aero =|λ1|+|λ2|+|λ3|.

[0120] K aero The coefficient reflects the magnitude of the velocity gradient tensor eigenvalue. The larger the coefficient, the more drastic and unstable the flow field velocity changes, and the greater the aerodynamic impact on the fan.

[0121] Environmental risk assessment also includes:

[0122] Based on the dynamic obstacle data and wind turbine protection capability coefficient in the four-dimensional tensor data, a distance attenuation factor and a velocity amplification factor are introduced to generate the external risk coefficient of the wind turbine.

[0123] In one possible calculation method, the external risk factor of the wind turbine is calculated as follows:

[0124] Dynamic obstacle data includes the velocity v of the i-th obstacle. i The current distance d between the i-th obstacle and the wind turbine. i The equivalent size s of the i-th obstacle i The wind turbine's own protection capability coefficient p defense The value range is [0,1].

[0125] To incorporate these factors into the risk calculation, a distance attenuation factor is introduced. and speed amplification factor Where α is the distance attenuation coefficient and β is the speed amplification coefficient. This indicates that as the distance between the obstacle and the wind turbine increases, the risk associated with it decreases exponentially. This demonstrates the amplifying effect of speed on risk; as speed increases, risk grows exponentially.

[0126] The risk contribution r of the i-th obstacle to the wind turbine i The calculation formula is:

[0127]

[0128] External risk factor K of wind turbine ex The sum of the risk contributions from all obstacles is calculated using the following formula:

[0129]

[0130] S220: Perform a single-factor risk assessment of the wind turbine's own condition risk based on four-dimensional tensor data.

[0131] The risk assessment of the wind turbine's own condition includes:

[0132] A multivariate temporal graph convolutional network is used to analyze the topological correlation of wind turbine operation data in four-dimensional tensor data and output the abnormal coefficient of operation status.

[0133] In one possible calculation method, the abnormal operation state coefficient is calculated as follows:

[0134] Wind turbine operation data is extracted from a four-dimensional tensor dataset. Parameters at each time point are used as nodes in a graph, and the correlations between parameters are used as edges. A Multivariate Temporal Graph Convolutional Network (MTGCN) is trained using temporal graph data. The MTGCN model contains multiple graph convolutional layers and temporal convolutional layers. The graph convolutional layers extract spatial features between nodes, while the temporal convolutional layers capture dynamic features over time. During training, temporal graph data under normal operating conditions is used as input, and the model parameters are adjusted through backpropagation to enable the model to learn the feature representations under normal operating conditions.

[0135] The real-time collected wind turbine operation data is constructed into a time series graph and input into the trained MTGCN model. The model outputs a score S representing the degree of difference between the current operating state and the normal state. p Then the abnormality coefficient K of the operating state ab The calculation formula is:

[0136]

[0137] Where τ is the score S p The comparison threshold. When the abnormal coefficient K of the running state. ab The closer the value is to 1, the greater the likelihood of abnormal operation of the wind turbine.

[0138] The risk assessment of the wind turbine's own condition also includes:

[0139] Feature extraction is performed on state data in four-dimensional tensor data, and historical fault knowledge graphs are associated to assess the fault risk coefficient of wind turbine components.

[0140] In one possible calculation method, the failure risk coefficient of wind turbine components is calculated as follows:

[0141] State data is extracted from the four-dimensional tensor data. This data mainly includes vibration signals and temperature changes of various components of the wind turbine. Time-domain features, namely mean, variance, and peak value, are extracted from the state data.

[0142] The time-domain characteristics are transformed to the frequency domain using Fourier transform, and frequency domain characteristics, namely frequency peaks and band energy, are extracted. Frequency domain characteristics are helpful in analyzing the frequency components and energy distribution of a signal.

[0143] The historical fault knowledge graph contains information on various faults that have occurred in the wind turbine in the past, including fault type, state data characteristics at the time of the fault, fault cause, and corresponding maintenance measures. The extracted current state data features are matched with fault features in the historical fault knowledge graph. Cosine similarity is used to measure the similarity (sim) between the current features and historical fault features.

[0144] The failure risk coefficient of wind turbine components is calculated based on the similarity score (sim) and the frequency of historical failures. Let F be the historical failure with the highest similarity to the current state data, its frequency be f(F), and its similarity score be sim. Then the failure risk coefficient K is... com The calculation formula is:

[0145] K com = f(F)·sim.

[0146] The higher the failure risk coefficient, the greater the likelihood of a failure occurring in the current operating state of the wind turbine components.

[0147] The S300 constructs a two-layer fusion mechanism, first calculating the fusion weight of the internal factors of environmental risk and wind turbine status risk, then calculating the fusion weight between the two, and finally fusioning the two risks based on these weights to obtain the fused risk value.

[0148] Please refer to Figure 4 The diagram illustrates a flowchart of an exemplary wind turbine intelligent early warning method S300 based on an adaptive scheme, the contents of which include:

[0149] S310: Define the first-layer fusion mechanism.

[0150] To determine the relative importance of each factor within both environmental risk and wind turbine condition risk, a first-layer fusion mechanism is constructed. Environmental risk includes the aerodynamic stability coefficient K. aero External risk coefficient K of the wind turbine ex The risk of wind turbine status includes the abnormal operating status coefficient K. ab And the risk factor K of the fan component failure com .

[0151] For the aerodynamic stability coefficient K aero External risk coefficient K of the wind turbine ex Operational status anomaly coefficient K ab And the risk factor K of the fan component failure com Calculate the fusion weights separately, using the following formula:

[0152]

[0153] Where, ω aero ω ex ω ab and ω com These represent the weights for aerodynamic stability coefficient, external risk coefficient, abnormal operating condition coefficient, and component failure risk coefficient, respectively. γ1 is an adjustment parameter used to adjust the allocation of the fusion weights, and e is an exponential function. This formula ensures that each weight value is between 0 and 1.

[0154] S320: Defines the second-layer fusion mechanism.

[0155] To integrate environmental risk and wind turbine condition risk, a second-layer integration mechanism is constructed, and the integration weight between the two is calculated. The environmental risk integration weight ω is calculated separately. env The weighted ω of wind turbine status risk integration state The calculation formula is:

[0156]

[0157] Where, ω env and ω state These are the environmental risk fusion weights and the wind turbine condition risk fusion weights, respectively. γ2 is an adjustment parameter used to adjust the allocation of fusion weights, and e is an exponential function.

[0158] This formula determines the relative importance of environmental risk in the overall risk assessment by calculating the results after weighting the environmental risk and the wind turbine condition risk internally.

[0159] S330: Based on the dual-layer fusion weights calculated by the dual-layer fusion mechanism, environmental risks and wind turbine status risks are fused together.

[0160] Based on the calculated two-layer fusion weights, environmental risk and wind turbine condition risk are fused. The formula for calculating the fused risk value is as follows:

[0161] R fused =ω env (ω aero K aero +ω ex K ex )+ω state (ω ab K ab +ω com K com );

[0162] Among them, R fused To integrate risk values.

[0163] The S400 sets low, medium, and high risk probability warning thresholds based on wind turbine operating requirements and scenarios, and compares the real-time calculated fused risk value with these thresholds to determine the wind turbine status. When the risk value exceeds the threshold, it performs data clustering analysis to identify high-risk data clusters, matches them with a fault mode library to determine potential risk patterns, and triggers different levels of warning signals and generates reports based on the risk level and pattern.

[0164] Please refer to Figure 5 The diagram illustrates a flowchart of an exemplary wind turbine intelligent early warning method S400 based on an adaptive scheme, the contents of which include:

[0165] S410: Set the risk probability warning threshold.

[0166] By analyzing four-dimensional tensor data, and based on the wind turbine's safety operation requirements and actual application scenarios, corresponding risk probability warning thresholds are set for different risk types. Multiple threshold levels are typically set, namely, a low-risk warning threshold R. low Medium-risk warning threshold R mid and high-risk warning threshold R high .

[0167] S420: Comparison of integrated risk values ​​and early warning thresholds.

[0168] During the real-time operation of the wind turbine, the fused risk value R is calculated in real time according to the method of dynamically fusing environmental risk and wind turbine state risk weights using the two-layer attention mechanism in S300. fused .

[0169] The real-time fusion risk value is compared with the set risk probability warning threshold.

[0170] If R fused ≤R low If so, it is determined that the fan is currently in normal operating condition;

[0171] If R low <R fused ≤R mid If so, the wind turbine is determined to be in a low-risk state;

[0172] If R mid <R fused ≤R high If so, the wind turbine is determined to be in a medium-risk state;

[0173] If R fused >R high If so, the wind turbine is determined to be in a high-risk state.

[0174] S430: Identify high-risk data clusters, associate them with the failure mode library, identify potential risk modes, and trigger early warning signals.

[0175] When the fusion risk value exceeds the warning threshold and enters a risk state, cluster analysis is performed on the current and recent four-dimensional tensor data. Data clusters are identified based on the cluster density connectivity between data points.

[0176] For example, if multiple parameters such as the wind turbine's speed, power, and oil temperature fluctuate abnormally in a data cluster at the same time, and the fusion risk value corresponding to the data cluster is high, then this data cluster is identified as a high-risk data cluster.

[0177] A fault mode library is pre-established, which contains various possible fault modes and their corresponding characteristic descriptions, fault causes, fault effects, and other information.

[0178] For example, for blade failure modes, the vibration signal characteristics of blade crack failure are recorded, such as abnormal increase in amplitude at a specific frequency and stress distribution characteristics. The cause of failure may be long-term fatigue, material defects, etc., and the impact of failure includes reduced power generation, increased wind turbine vibration, or even blade breakage.

[0179] The identified high-risk data clusters are matched with failure modes in the failure mode library. By comparing the characteristics of the data clusters with the characteristics of the failure modes, the best-matching failure mode is found.

[0180] Based on the matching results with the failure mode library, potential risk modes are identified. If a highly matching failure mode is found, it can be determined that the wind turbine may currently face the potential risk corresponding to that failure mode.

[0181] For example, if a high-risk data cluster is highly similar to a blade crack failure mode, then it can be determined that the current wind turbine has a potential risk of blade cracks.

[0182] Based on the risk level and potential risk patterns, different levels of early warning signals are triggered. For low-risk situations, internal system messages remind maintenance personnel to pay attention to the wind turbine's operating status, such as displaying a yellow warning icon on the monitoring software interface. For medium-risk situations, in addition to message reminders, SMS notifications can be sent to relevant maintenance personnel, and the risk event is recorded in the monitoring system. For high-risk situations, audible and visual alarms are immediately activated to notify all relevant personnel, and a detailed fault report is automatically generated, including information such as the risk level, potential risk patterns, and characteristics of the data clusters involved, so that maintenance personnel can take timely measures to prevent the fault from occurring or escalating.

[0183] This application also provides a wind turbine intelligent early warning system based on an adaptive scheme, including:

[0184] Data acquisition and processing module: used to acquire multi-band heterogeneous data of wind turbine in real time. The multi-band heterogeneous data includes wind turbine operation data, status data, environmental data and dynamic obstacle data. The module also performs time alignment on the multi-band heterogeneous data to construct four-dimensional tensor data.

[0185] Single-factor risk assessment module: connected to the data acquisition and processing module, used to perform single-factor risk assessment on environmental risk and wind turbine self-state risk based on four-dimensional tensor data, and calculate the aerodynamic stability coefficient of the wind turbine area, the external risk coefficient of the wind turbine, the abnormal operating state coefficient of the wind turbine and the failure risk coefficient of wind turbine components respectively.

[0186] Risk fusion module: connected to the single-factor risk assessment module, used to dynamically fuse the weights of environmental risk and wind turbine status risk to generate a fused risk value;

[0187] Risk warning module: Connected to the risk fusion module, it is used to set a risk probability warning threshold, compare the fused risk value with the risk probability warning threshold, identify high-risk data clusters and associate them with the fault mode library, identify potential risk modes, and trigger warning signals of different levels.

[0188] This application also provides a computer device, the computer device comprising: a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, wherein the at least one instruction, the at least one program, the code set or instruction set is loaded and executed by the processor to realize a wind turbine intelligent early warning method based on an adaptive scheme.

[0189] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0190] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0191] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0192] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features of the invention herein.

[0193] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications or equivalent substitutions made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A wind turbine intelligent early warning method based on an adaptive scheme, characterized in that, include: Real-time acquisition of multi-band heterogeneous data from wind turbines, spatiotemporal alignment of the multi-band heterogeneous data, and construction of four-dimensional tensor data; Single-factor risk assessments of environmental risks and wind turbine-specific risks are conducted based on four-dimensional tensor data. The aerodynamic stability coefficient of the area where the wind turbine is located, the external risk coefficient of the wind turbine, the abnormal operating status coefficient of the wind turbine, and the failure risk coefficient of the wind turbine components are calculated respectively. Environmental risk assessment includes: Computational fluid dynamics simulation analysis was used to analyze environmental data in four-dimensional tensor data to calculate the aerodynamic stability coefficient of the area where the wind turbine is located. Based on the dynamic obstacle data and wind turbine protection capability coefficient in the four-dimensional tensor data, a distance attenuation factor and a velocity amplification factor are introduced to generate the external risk coefficient of the wind turbine. The risk assessment of the wind turbine's own condition includes: A multivariate temporal graph convolutional network is used to analyze the topological correlation of wind turbine operation data in four-dimensional tensor data and output the anomaly coefficient of operation status. Feature extraction is performed on state data in four-dimensional tensor data, and historical fault knowledge graphs are associated to assess the fault risk coefficient of wind turbine components. The weights of environmental risks and wind turbine status risks are dynamically integrated to generate a combined risk value; By setting risk probability warning thresholds and comparing risk values ​​with risk probability warning thresholds, high-risk data clusters are identified and associated with a fault mode library to identify potential risk modes and trigger warning signals of different levels.

2. The wind turbine intelligent early warning method based on an adaptive scheme according to claim 1, characterized in that, The multi-band heterogeneous data includes wind turbine operation data, wind turbine status data, environmental data, and dynamic obstacle data; The four-dimensional tensor feature constructed using the multi-band heterogeneous data includes: the first dimension representing time, the second dimension representing data type, corresponding to wind turbine operation data, status data, environmental data, and dynamic obstacle data, respectively, the third dimension representing the specific parameters in each data type, and the fourth dimension representing the spatial dimension.

3. The wind turbine intelligent early warning method based on an adaptive scheme according to claim 1, characterized in that, The calculation method for the aerodynamic stability coefficient of the area where the wind turbine is located includes: A three-dimensional flow field model of the area where the wind turbine is located is constructed, and the three-dimensional flow field model is numerically solved to obtain the velocity vector distribution of the flow field around the wind turbine; Calculate the velocity gradient tensor S based on the velocity vector distribution. xyz Its formula is: Where x, y, z correspond to the three directions of spatial coordinates, u x u y u z It is a velocity vector Components in the x, y, z directions This represents the partial derivative of the velocity component in the corresponding coordinate direction; Solve the characteristic equation: det(S) xyz -λI)=0, obtain the eigenvalues ​​λ1, λ2, λ3 of the velocity gradient tensor; Through the formula: K aero =|λ1|+|λ2|+|λ3|, calculate the aerodynamic stability coefficient K. aero .

4. The wind turbine intelligent early warning method based on an adaptive scheme according to claim 1, characterized in that, The calculation method for the external risk factor of the wind turbine includes: Calculate the risk contribution r of the i-th obstacle to the wind turbine. i The calculation formula is: Among them, v i Let d be the velocity of the i-th obstacle. i Let s be the current distance between the i-th obstacle and the wind turbine. i p is the equivalent size of the i-th obstacle; defense This refers to the wind turbine's own protection capability coefficient. For distance attenuation factor, Where α is the speed amplification factor, β is the distance attenuation factor, and β is the speed amplification factor. External risk coefficient K of wind turbine ex The calculation formula is:

5. The wind turbine intelligent early warning method based on an adaptive scheme according to claim 1, characterized in that, The calculation method for the abnormal operating status coefficient includes: Wind turbine operation data is extracted from four-dimensional tensor data, and a multivariate time series graph convolutional network is trained using time series graph data; The real-time collected wind turbine operation data is used to construct a time series graph, which is then input into a trained MTGCN model. The output is a score S representing the degree of difference between the current operating state and the normal state. p ; Operational status anomaly coefficient K ab The calculation formula is: Where τ is the score S p The threshold for comparison.

6. The wind turbine intelligent early warning method based on an adaptive scheme according to claim 1, characterized in that, The calculation methods for the failure risk factor of wind turbine components include: State data is extracted from four-dimensional tensor data, time-domain features are extracted based on the state data, and the time-domain features are transformed to the frequency domain through Fourier transform. The extracted current state data features are matched with the fault features in the historical fault knowledge graph to obtain the similarity sim between the current state data features and the historical fault features. Failure risk coefficient K com The calculation formula is: K com =f(F)·sim; F represents the historical fault with the highest similarity to the current state data features, and its occurrence frequency is f(F).

7. The wind turbine intelligent early warning method based on an adaptive scheme according to claim 1, characterized in that, The steps for calculating the fusion risk value include: Define the first-layer fusion mechanism, that is, for the aerodynamic stability coefficient K aero External risk coefficient K of the wind turbine ex Operational status anomaly coefficient K ab And the failure risk coefficient K of the fan component com Calculate the fusion weights separately, using the following formula: Where, ω aero ω ex ω ab and ω com These are the weights for aerodynamic stability coefficient, external risk coefficient, abnormal operating condition coefficient, and fault risk coefficient of wind turbine components, respectively. γ1 is an adjustment parameter used to adjust the allocation of fusion weights, and e is an exponential function. Define a second-layer fusion mechanism, which involves calculating the environmental risk fusion weight ω separately. env The weighted ω of wind turbine status risk integration state The calculation formula is: Where, ω env and ω state These are the environmental risk fusion weights and the wind turbine condition risk fusion weights, respectively. γ2 is an adjustment parameter used to adjust the allocation of fusion weights, and e is an exponential function. The environmental risk and wind turbine condition risk are integrated, and the formula for calculating the integrated risk value is as follows: R fused =ω env (oh aero K aero +oh ex K ex )+ω state (oh ab K ab +oh com K com ); Among them, R fused To integrate risk values.

8. The wind turbine intelligent early warning method based on an adaptive scheme according to claim 7, characterized in that, The set risk probability warning threshold is: low risk warning threshold R. low Medium-risk warning threshold R mid and high-risk warning threshold R high ; The method for comparing the fused risk value with the risk probability early warning threshold is as follows: If R fused ≤R low If so, it is determined that the fan is currently in normal operating condition; If R low <R fused ≤R mid If so, the wind turbine is determined to be in a low-risk state; If R mid <R fused ≤R high If so, the wind turbine is determined to be in a medium-risk state; If R fused >R high If so, the wind turbine is determined to be in a high-risk state.

9. A wind turbine intelligent early warning system based on an adaptive scheme, characterized in that, include: Data acquisition and processing module: used to acquire multi-band heterogeneous data of wind turbine in real time. The multi-band heterogeneous data includes wind turbine operation data, status data, environmental data and dynamic obstacle data. The module also performs time alignment on the multi-band heterogeneous data to construct four-dimensional tensor data. Single-factor risk assessment module: connected to the data acquisition and processing module, used to perform single-factor risk assessment on environmental risk and wind turbine self-state risk based on four-dimensional tensor data, and calculate the aerodynamic stability coefficient of the wind turbine area, the external risk coefficient of the wind turbine, the abnormal operating state coefficient of the wind turbine and the failure risk coefficient of wind turbine components respectively. Risk fusion module: connected to the single-factor risk assessment module, used to dynamically fuse the weights of environmental risk and wind turbine status risk to generate a fused risk value; Risk warning module: Connected to the risk fusion module, it is used to set a risk probability warning threshold, compare the fused risk value with the risk probability warning threshold, identify high-risk data clusters and associate them with the fault mode library, identify potential risk modes, and trigger warning signals of different levels.

10. A computer device, characterized in that, The computer device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the wind turbine intelligent early warning method based on an adaptive scheme as described in any one of claims 1 to 8.

Images