New energy wind power system fault detection method, system and device based on multi-parameter fusion and medium

Abstract

The invention relates to the technical field of new energy power generation monitoring, in particular to a new energy wind power system fault detection method, system and device based on multi-parameter fusion and a medium. The method comprises the following steps: acquiring various operating parameters such as wind speed, wind direction, generator speed, temperature, voltage and current of a wind power system in real time, and performing feature extraction on the operating parameters to obtain key features related to a fault; analyzing the key features based on a deep learning algorithm, and judging the fault type of the wind power system; and outputting a fault diagnosis result in a visual mode, and sending the fault diagnosis result to a control center for remote monitoring and fault early warning. The technical problems that fault detection of a traditional wind power system is poor in real-time performance, not high in accuracy, low in intelligent degree and the like are solved, and reliable guarantee is provided for safe and stable operation of the wind power system.

Description

Technical Field

[0001] The present invention relates to the field of renewable energy power generation monitoring technology, and in particular to a new energy wind power system fault detection method, system, equipment and medium based on multi-parameter fusion. Background Art

[0002] With the continuous development of new energy technologies, wind power has gained widespread application as a clean energy source. As a crucial component of renewable energy generation, the operational stability and safety of wind power systems directly impact the reliability of the entire power system. However, wind power systems are prone to various types of failures during operation, including generator failure, sensor failure, electrical system failure, and mechanical system failure, due to factors such as equipment aging, environmental impacts, and load fluctuations.

[0003] Traditional wind power system fault detection methods rely primarily on regular manual inspections and single-parameter monitoring, which present significant limitations. First, manual inspections cannot achieve real-time monitoring, often detecting faults only after they occur, missing the optimal opportunity for troubleshooting. Second, single-parameter monitoring lacks comprehensiveness, making it difficult to accurately reflect the overall operating status of the wind power system and prone to false alarms or missed detections. Furthermore, traditional detection methods are inefficient and lack high accuracy, failing to meet the efficient and accurate fault detection requirements of modern wind power systems.

[0004] Existing fault detection methods generally suffer from poor real-time performance, low accuracy, and a lack of intelligent analysis capabilities. Many existing solutions only provide simple alarm signals, lacking detailed fault diagnosis information and visual displays, making it difficult for operations and maintenance personnel to quickly understand and address fault conditions. Furthermore, traditional methods struggle to achieve remote monitoring and fault warnings, limiting the level of intelligent management of wind power systems. Summary of the Invention

[0005] In view of the problems existing in the prior art, the present invention is proposed.

[0006] Therefore, the present invention aims to address the real-time nature of wind power system fault detection, the accuracy of multi-parameter fusion analysis, the visualization of fault diagnosis results, and the implementation of remote monitoring and intelligent early warning. By establishing an intelligent fault detection method based on multi-parameter fusion, comprehensive, real-time, and accurate monitoring of wind power system operating conditions can be achieved, improving operational safety, stability, and intelligent management of wind power systems.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, an embodiment of the present invention provides a new energy wind power system fault detection method based on multi-parameter fusion, which includes collecting operating parameters of the wind power system in real time, extracting features of the operating parameters, and obtaining key features related to the fault;

[0009] Analyze the key features based on a deep learning algorithm to determine the fault type of the wind power system;

[0010] The fault diagnosis results are output in a visual manner and sent to the control center for remote monitoring and fault warning.

[0011] As a preferred solution of the new energy wind power system fault detection method based on multi-parameter fusion according to the present invention, the real-time collection of operating parameters of the wind power system includes:

[0012] Collect wind speed parameters and wind direction parameters; collect generator speed parameters; collect temperature parameters, voltage parameters and current parameters.

[0013] As a preferred solution of the new energy wind power system fault detection method based on multi-parameter fusion according to the present invention, the feature extraction of the operating parameters includes:

[0014] Extract the changing trend characteristics of wind speed and wind direction; extract the abnormal fluctuation characteristics of generator speed; extract the correlation characteristics of temperature, voltage and current.

[0015] As a preferred solution of the new energy wind power system fault detection method based on multi-parameter fusion according to the present invention, the analysis of key features based on the deep learning algorithm includes:

[0016] Establish a multi-parameter fusion fault diagnosis model; input key features into the fault diagnosis model; output the fault type recognition probability.

[0017] As a preferred solution of the new energy wind power system fault detection method based on multi-parameter fusion described in the present invention, the fault types include generator fault, sensor fault, electrical system fault and mechanical system fault.

[0018] As a preferred solution of the new energy wind power system fault detection method based on multi-parameter fusion according to the present invention, the outputting of the fault diagnosis results in a visual manner includes:

[0019] Generate fault type report; generate fault location information; generate fault severity assessment results.

[0020] As a preferred solution of the fault detection method of the new energy wind power system based on multi-parameter fusion of the present invention, wherein: the sending to the control center for remote monitoring and fault warning includes:

[0021] The fault diagnosis results are sent to the wind power system control center through the communication interface; the corresponding level of fault warning signal is triggered; and the fault occurrence time and processing status are recorded.

[0022] In a second aspect, an embodiment of the present invention provides a new energy wind power system fault detection system based on multi-parameter fusion, which includes a data acquisition module for collecting operating parameters of the wind power system in real time;

[0023] The data processing module is used to extract the characteristics of the operating parameters and obtain the key characteristics related to the fault;

[0024] Fault diagnosis module, which uses deep learning algorithms to analyze key features and determine the fault type of the wind power system;

[0025] The result output module is used to output the fault diagnosis results in a visual manner and send the fault diagnosis results to the control center for remote monitoring and fault warning.

[0026] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the new energy wind power system fault detection method based on multi-parameter fusion as described in the first aspect of the present invention are implemented.

[0027] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium on which a computer program is stored, wherein: when the computer program instructions are executed by a processor, the steps of the new energy wind power system fault detection method based on multi-parameter fusion as described in the first aspect of the present invention are implemented.

[0028] The beneficial effects of the present invention are as follows: by simultaneously collecting multiple operating parameters such as wind speed, wind direction, generator speed, temperature, voltage, current, etc. in real time, it breaks the limitations of traditional single parameter monitoring and achieves a panoramic perception of the operating status of the wind power system. This multi-dimensional parameter fusion can not only capture the comprehensive operating status of the system that cannot be reflected by a single parameter, but more importantly, it can identify the inherent correlation and mutual influence relationship between parameters. When a parameter is abnormal, the system can cross-verify through the changing trends of other parameters, significantly reducing the probability of false alarms and missed alarms, and achieving the beneficial effect of improving the accuracy and reliability of fault detection.

[0029] By extracting the changing trends of wind speed and direction, the abnormal fluctuations of generator speed, and the correlations between temperature, voltage, and current, a multi-level feature analysis system was established. This feature extraction method not only identifies overt parameter anomalies but also uncovers latent fault symptoms and early-stage fault signals. In particular, by analyzing the correlations between parameters, it is possible to detect complex and progressive faults that are difficult to identify with traditional methods. This shift from passive fault detection to active fault warning significantly improves the sensitivity and foresight of fault detection.

[0030] By establishing a multi-parameter fusion fault diagnosis model and analyzing it based on a deep learning algorithm, the system has acquired autonomous learning and pattern recognition capabilities. This algorithm automatically extracts fault characteristic patterns from a large amount of historical operating data and continuously optimizes diagnostic accuracy as the data accumulates. Compared to traditional rule-based or threshold-based diagnostic methods, deep learning algorithms can handle complex nonlinear relationships and multivariable coupling, and have a stronger ability to identify multiple faults in complex operating conditions, significantly improving the accuracy and adaptability of fault type identification.

[0031] By generating fault type reports, fault location information, and fault severity assessments, complex diagnostic data is transformed into intuitive and easy-to-understand visual information. This output not only provides maintenance personnel with a clear description of the fault status but, more importantly, provides a scientific basis for maintenance decision-making. By quantifying fault severity, maintenance personnel can rationally prioritize maintenance and allocate resources, avoiding blind and excessive maintenance, thereby optimizing maintenance strategies and reducing maintenance costs.

[0032] By establishing a multi-level fault warning signal system and remote communication mechanism, real-time transmission of fault information and a graded response are achieved. The system automatically triggers warning signals at the appropriate level based on the severity of the fault, ensuring that critical faults are detected and addressed promptly. Furthermore, through fault recording and status tracking, a complete closed-loop fault handling management mechanism has been established, significantly shortening fault response time and processing cycles, thereby improving the operational safety and management efficiency of the wind power system.

[0033] By organically integrating functional modules such as data acquisition, feature extraction, intelligent diagnosis, visual output, and remote warning, a complete intelligent fault detection ecosystem has been formed. The collaborative work between these modules not only achieves functional complementarity but, more importantly, through information sharing and feedback mechanisms, enables the entire system to self-optimize and continuously improve. This system-level integrated innovation overcomes the technical bottlenecks of traditional fault detection methods, achieving a technological leap from single-point detection to systematic diagnosis, from qualitative judgment to quantitative analysis, and from manual inspection to intelligent monitoring, achieving the beneficial effect of comprehensively improving the intelligent management level of wind power systems.

[0034] Through the above-mentioned technological innovations, the present invention not only solves the technical problems of poor real-time performance, low accuracy, and low intelligence level of fault detection in traditional wind power systems, but more importantly, provides key technical support for the large-scale development and intelligent operation and maintenance of the new energy wind power industry. It has important engineering application value and broad industrialization prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0036] Figure 1 This is a flow chart of a fault detection method for a new energy wind power system based on multi-parameter fusion;

[0037] Figure 2 A computer device diagram for a fault detection method for a new energy wind power system based on multi-parameter fusion;

[0038] Figure 3 This is a flow chart of Example 2 of a new energy wind power system fault detection method based on multi-parameter fusion. DETAILED DESCRIPTION

[0039] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0040] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0041] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it constitute a separate or selective embodiment that is mutually exclusive with other embodiments.

[0042] Example 1

[0043] Reference Figures 1 and 2 , which is the first embodiment of the present invention, provides a new energy wind power system fault detection method based on multi-parameter fusion, including:

[0044] S100: Real-time collection of wind power system operating parameters, feature extraction of the operating parameters, and acquisition of key features related to the fault;

[0045] S200: Analyzes key features based on deep learning algorithms to determine the fault type of the wind power system;

[0046] S300: Output the fault diagnosis results in a visual manner and send them to the control center for remote monitoring and fault warning.

[0047] It should be noted that wind power systems face complex technical challenges and fault detection difficulties during operation. Traditional wind power system fault detection mainly has the following technical problems: real-time problem of wind power system fault detection. Traditional wind power system fault detection methods may have delays and cannot obtain the operating data of the wind power system in real time. However, this technical solution can obtain various parameters of the wind power system in real time through real-time data collection, so as to timely discover potential faults; the accuracy and efficiency of fault diagnosis. Traditional fault detection methods may be limited by data processing capabilities and algorithm accuracy, resulting in inaccurate or inefficient fault diagnosis results. This technical solution uses multi-parameter fusion processing and feature extraction, combined with efficient deep learning fault diagnosis algorithms, to more accurately determine the type and location of the fault and improve diagnostic efficiency; fault diagnosis The problem of visualization and comprehensibility of fault diagnosis results. Traditional fault detection methods may only provide simple alarm signals and lack detailed fault diagnosis information, which is not conducive to user understanding and processing. This technical solution uses the result output function to display the fault diagnosis results in a visual way, which is convenient for users to intuitively understand the fault situation and take corresponding treatment measures; the problem of implementation of remote monitoring and fault early warning. Traditional wind power system fault detection methods may not be able to achieve remote monitoring and fault early warning, which limits the intelligent management of wind power systems. This technical solution can send fault diagnosis results to the control center or other related equipment through the communication function, realize remote monitoring and fault early warning, and improve the operation safety and management efficiency of wind power systems.

[0048] Therefore, in response to the above-mentioned operation monitoring and health prediction problems, through the steps S100-S300, the operation status monitoring of the wind power system under typical working conditions is realized, the relationship between the operation data and fault characteristics of the wind power system under the influence of multiple parameters is obtained, and the internal operation status is accurately estimated; the real-time monitoring of the wind power system operation parameters is realized, the system operation status is dynamically analyzed, and early warning of system abnormalities is realized; at the same time, based on the deep learning algorithm and the fault diagnosis model, the fault type of the wind power system is accurately identified and predicted. The present invention can comprehensively reflect the operation status of the wind power system through the fusion of multiple parameters, provide accurate fault detection and diagnosis, and provide a strong guarantee for the safe and stable operation of the wind power system.

[0049] It should be noted that the key English abbreviations involved in this technical solution include: Global Positioning System (GPS), used for positioning and time synchronization of wind turbines; Supervisory Control and Data Acquisition (SCADA), used for centralized monitoring and data acquisition of wind power systems; Programmable Logic Controller (PLC), used for automated control of wind power systems; Human-Machine Interface (HMI), used as the interactive interface between operators and wind power systems; Doubly Fed Induction Generator (DFIG), a commonly used type of generator in wind power systems; Maximum Power Point Tracking (MPPT), used to optimize the power generation efficiency of wind power systems; Condition Monitoring System (CMS), used to monitor the operating status of wind power equipment in real time; these English abbreviations represent the core technologies and key equipment in wind power systems, and are an important technical basis for realizing the technical solution of the present invention.

[0050] Example 2

[0051] Reference Figure 1-Figure 3 , which is the second embodiment of the present invention.

[0052] In the embodiment of the present application, step S100 collects the operating parameters of the wind power system in real time and performs feature extraction, including the following steps A1-A2:

[0053] A1: Real-time collection of wind power system operating parameters includes:

[0054] Collect wind speed parameters and wind direction parameters; collect generator speed parameters; collect temperature parameters, voltage parameters and current parameters.

[0055] Specifically, the collection of wind speed parameters is achieved through a wind speed sensor installed on the top of the wind turbine tower. The collection frequency is 10 times per second, the measurement range is 0 to 80 meters per second, and the accuracy is plus or minus 0.1 meters per second. The collection of wind direction parameters is achieved through a wind direction sensor. The collection frequency is synchronized with the wind speed, the measurement range is 0 to 360 degrees, and the accuracy is plus or minus 1 degree. These sensors use the ultrasonic measurement principle to avoid wear of mechanical components and improve the reliability and long-term stability of the measurement. The wind speed sensor uses a three-cup anemometer or an ultrasonic anemometer, which can accurately measure wind speed changes under various wind conditions. The wind direction sensor uses a wind vane or an ultrasonic anemometer, and outputs the wind direction angle signal through a magnetoresistive sensor or a photoelectric encoder.

[0056] In an optional embodiment, wind speed and direction can also be collected using a LiDAR wind measurement system. This system can measure wind speed and direction distribution at different altitudes, providing richer data support for precise control of wind turbines. The LiDAR system has a measurement range of up to 200 meters and a vertical resolution of 10 meters, enabling real-time monitoring of key parameters such as wind shear and turbulence intensity. This system measures wind speed by emitting a laser beam and receiving Doppler shift signals from aerosol particles in the atmosphere, offering high measurement accuracy and fast response times.

[0057] Generator speed parameters are collected using a speed sensor mounted on the generator shaft. These sensors use either magnetoelectric or photoelectric sensors, with a sampling frequency of 100 times per second and a measurement range of 0 to 2000 revolutions per minute (RPM) with an accuracy of ±0.1 RPM. The speed signal is a key indicator of the wind turbine system's operating status, reflecting the efficiency of the wind turbine rotor and the working condition of the generator. Magnetoelectric speed sensors generate pulse signals by sensing the passage of gears or magnets, while photoelectric speed sensors generate pulse signals by detecting the grating patterns on the encoder disk using a photoelectric switch.

[0058] In another alternative implementation, speed parameters can be acquired using the speed feedback signal within the inverter. This approach offers higher accuracy and stability while reducing the installation and maintenance costs of additional sensors. The encoder feedback signal within the inverter can provide higher-resolution speed information, typically with an accuracy of ±0.01 rpm.

[0059] Temperature parameter collection includes monitoring at multiple measurement points, including generator, transformer, bearing, and gearbox temperatures. Generator temperatures include stator, rotor, and coolant temperatures, and are measured using platinum resistance temperature sensors with a measurement range of -40°C to +200°C and an accuracy of + / - 0.2°C. Transformer temperature monitoring includes oil and winding temperatures, using fiber optic temperature sensors that enable multi-point distributed measurement. Bearing temperature monitoring utilizes embedded temperature sensors, mounted directly on the bearing outer ring, to accurately reflect the bearing's operating status. Gearbox temperature monitoring includes oil, bearing, and gear temperatures, and comprehensive monitoring is achieved through multi-point temperature sensors.

[0060] It should be noted that the temperature monitoring system employs a tiered monitoring strategy, categorizing temperature sensors into three levels based on their importance: critical monitoring points, important monitoring points, and general monitoring points. Critical monitoring points include generator stator winding temperature and main bearing temperature, with a data acquisition frequency of once per second; important monitoring points include gearbox oil temperature and transformer oil temperature, with a data acquisition frequency of once every 10 seconds; and general monitoring points include nacelle temperature and tower temperature, with a data acquisition frequency of once per minute. This tiered monitoring strategy ensures real-time monitoring of key areas while reducing the system's data processing burden.

[0061] Voltage parameters collected include generator output voltage, transformer high and low voltage side voltages, and grid connection point voltage. High-precision voltage transformers are used for signal acquisition, covering a measurement range of 380 V to 35 kilovolts with an accuracy of ±0.2%. Voltage signals are collected 50 times per second, capable of capturing rapid voltage fluctuations and harmonic components. Generator output voltage monitoring utilizes three-phase voltage transformers, capable of simultaneously monitoring the amplitude and phase relationship of the three-phase voltages. Transformer voltage monitoring includes both primary and secondary voltages, with the voltage transformer converting the high-voltage signal into a low-voltage signal for measurement.

[0062] Current parameters collected include generator stator current, excitation current, and transformer primary and secondary currents. High-precision current transformers or Rogowski coils are used for signal acquisition, with a measurement range of 1A to 5000A and an accuracy of ±0.2%. Current signals can reflect the load status of the wind power system and the operating conditions of electrical equipment. Stator current monitoring uses three-phase current transformers to monitor the balance and harmonic content of the three-phase currents. Excitation current monitoring uses DC current sensors to reflect the generator's excitation status and reactive power output.

[0063] In an optional implementation, voltage and current monitoring can also be combined with a power analyzer, capable of simultaneously measuring power quality parameters such as active power, reactive power, power factor, and harmonic content, providing comprehensive data support for evaluating the grid-connected performance of wind power systems. A power analyzer can calculate multiple power parameters such as instantaneous power, average power, and maximum power, and analyze the spectral characteristics of voltage and current.

[0064] A2: Feature extraction of operating parameters includes:

[0065] Extract the changing trend characteristics of wind speed and wind direction; extract the abnormal fluctuation characteristics of generator speed; extract the correlation characteristics of temperature, voltage and current.

[0066] Specifically, the wind speed and wind direction change trend feature extraction uses a sliding window statistical method to calculate the statistical features of the wind speed within 10 minutes, such as the average value, standard deviation, maximum value, and minimum value. At the same time, the wind speed change rate and the wind direction change amplitude are calculated to identify abnormal wind conditions such as gusts and wind shear. The wind speed change trend feature extracts the upward trend, downward trend, and stability characteristics of the wind speed by analyzing the wind speed data within a continuous time window. The calculation formula for the wind speed change rate is:

[0067]

[0068] Where: V is the wind speed change rate, V t is the wind speed at the current moment, V t-1 The wind speed at the previous moment, Δt is the time interval.

[0069] This parameter reflects rapid changes in wind speed and is an important indicator for determining whether a wind turbine requires emergency braking. Wind direction variation characteristics include parameters such as wind direction standard deviation, wind direction extremes, and wind direction variation frequency, which are used to assess wind direction stability and turbulence intensity.

[0070] In an optional implementation, wind speed and direction feature extraction can also utilize wavelet transforms, decomposing wind speed signals into frequency components and extracting high-frequency fluctuations and low-frequency trends, thereby providing more accurate wind condition predictions for intelligent wind turbine control. Wavelet transforms provide both time-domain and frequency-domain information, making them particularly suitable for analyzing non-stationary wind speed signals. By selecting appropriate wavelet basis functions, both sudden and periodic variations in wind speed signals can be effectively extracted.

[0071] Abnormal generator speed fluctuation characteristics are extracted by analyzing the spectral characteristics of the speed signal. Under normal circumstances, the generator speed should remain relatively stable. However, when a mechanical or electrical fault occurs, the speed will exhibit characteristic fluctuations. Fast Fourier transform is used to extract the frequency domain characteristics of the speed signal and identify abnormal frequency components. Speed ​​fluctuation characteristics include speed variance, peak-to-peak speed, and speed spectral energy distribution.

[0072] Specifically, abnormal speed fluctuation detection uses statistical process control methods to establish a speed control chart. When the speed exceeds the control limit, an abnormal alarm is triggered. The control limits are calculated based on statistical analysis of historical data, and the upper and lower control limits are set to the mean plus or minus three standard deviations, respectively. The evaluation indicators for speed fluctuation include:

[0073]

[0074] Where: K is the fluctuation coefficient, σ n is the speed standard deviation, μ n is the average speed.

[0075] When the fluctuation coefficient exceeds the set threshold, it is determined to be an abnormal speed fluctuation. At the same time, the instantaneous rate of change of the speed is monitored. When the rate of change exceeds the set limit, a sudden speed change alarm is triggered.

[0076] In another alternative implementation, abnormal speed detection can also employ machine learning methods to train a speed anomaly pattern recognition model capable of identifying characteristic speed fluctuation patterns of mechanical faults such as bearing and gearbox failures. By collecting large amounts of normal and abnormal speed data and training a support vector machine or neural network model, automatic identification of abnormal speed can be achieved.

[0077] Correlation analysis and principal component analysis are used to extract the correlation features of temperature, voltage, and current. By calculating the correlation coefficients between different parameters, the inherent connections between them can be identified. For example, the stator current and stator temperature of a generator typically have a positive correlation. An abnormal correlation may indicate an insulation fault or cooling system failure. The formula for calculating the correlation coefficient is:

[0078]

[0079] Where: R is the correlation coefficient, COV(X,Y) is the covariance value, σ x is the standard deviation of parameter X, σ y is the standard deviation of parameter Y.

[0080] The correlation coefficient ranges from negative 1 to positive 1, and the larger the absolute value, the stronger the correlation.

[0081] Specifically, correlation features are extracted to establish a multivariate regression model, with temperature as the dependent variable and voltage and current as independent variables, to establish a temperature prediction model. When the deviation between the actual temperature and the predicted temperature exceeds a threshold, it is determined to be an abnormal state. The regression model is established using the least squares method, and the model accuracy is evaluated through cross-validation. The general form of the multivariate regression model is:

[0082] T=a+b×U+c×l+ε

[0083] Where: T is temperature, a is a constant term, b and c are regression coefficients, U is voltage, I is current, and ε is the error term.

[0084] The model parameters were determined by training with historical data, and the goodness of fit of the model was evaluated by the coefficient of determination.

[0085] It should be noted that multi-parameter correlation analysis utilizes a dynamic time window approach, adaptively adjusting the analysis window length based on operating conditions. Longer time windows are used during stable operation, while shorter time windows are used during startups, shutdowns, and load changes, improving the accuracy and timeliness of feature extraction. The length of the time window is determined by the speed of parameter change: short time windows are used for rapidly changing parameters, and long time windows are used for slowly changing parameters.

[0086] In the embodiment of the present application, step S200 analyzes the key features based on the deep learning algorithm, including the following steps B1-B2:

[0087] B1: Analyzing key features based on deep learning algorithms includes: establishing a multi-parameter fusion fault diagnosis model; inputting key features into the fault diagnosis model; and outputting the probability of fault type identification.

[0088] Specifically, the multi-parameter fusion fault diagnosis model utilizes a deep neural network architecture, consisting of an input layer, multiple hidden layers, and an output layer. The input layer receives all feature parameters extracted from steps A1 and A2, including wind speed and direction, rotational speed, and temperature, voltage, and current correlations. The input dimension is a 128-dimensional feature vector. The feature vector is constructed using the normalization principle, scaling all eigenvalues ​​to a range of 0 to 1 to minimize the impact of features of different dimensions on model training.

[0089] The network structure design adopts a hybrid architecture combining fully connected layers and convolutional layers. The first layer is a fully connected layer with 256 neurons and a linear rectifier function as the activation function. The second layer is a one-dimensional convolutional layer with a convolution kernel size of 3, a stride of 1, and 128 output channels. The convolutional layer can extract local correlations between features and enhance the model's feature learning ability. The third layer is a pooling layer, which uses a maximum pooling operation with a pooling kernel size of 2 to reduce dimensionality and enhance feature generalization. Two subsequent fully connected layers are included, containing 64 and 32 neurons, respectively, to gradually reduce network complexity.

[0090] The model is trained using the backpropagation algorithm and the cross-entropy loss function, which effectively handles multi-classification problems. The optimizer uses the adaptive moment estimation algorithm, which combines the advantages of momentum and adaptive learning rates, resulting in fast convergence and excellent stability. The learning rate is initially set to 0.001, and a learning rate decay strategy is used. If the validation set loss function does not decrease for five consecutive epochs, the learning rate is decayed by 0.5 times the original value. The training batch size is set to 32, with a maximum training epoch of 1000. An early stopping strategy is used to prevent overfitting.

[0091] In an optional implementation, the fault diagnosis model can also utilize a long-short-term memory (LSTM) network architecture. This network can process time series data, using historical operating data to predict future fault trends and provide early warning of faults. LSTM networks, through a gating mechanism, selectively memorize and forget information, making them particularly well-suited for processing time series data with long-term dependencies. The network structure includes a forget gate, input gate, and output gate, effectively avoiding the vanishing gradient problem of traditional recurrent neural networks.

[0092] Key model input features undergo feature selection and dimensionality reduction. Recursive feature elimination is used to select the most important 64 features from the 128-dimensional feature set as model input, improving the model's generalization and computational efficiency. Cross-validation is used in the feature selection process to compare the model performance of different feature combinations and select the optimal feature subset. Dimensionality reduction can also utilize principal component analysis (PCA) to combine highly correlated features into principal components, reducing feature dimensionality while retaining key information.

[0093] The model outputs the probability of fault type identification, using a multi-class output method, with one output node corresponding to each fault type. The output layer uses a softmax activation function to ensure that the sum of all output probabilities is 1, facilitating probabilistic interpretation. When the output probability of a particular fault type exceeds 0.8, it is determined to be that type of fault. When the maximum probability is less than 0.6, the fault type cannot be determined, requiring further manual confirmation. This confidence threshold setting balances detection accuracy and false alarm rate.

[0094] In another alternative implementation, the fault diagnosis model can also employ an ensemble learning approach, combining multiple neural network models with different architectures and using a voting mechanism to improve the accuracy and reliability of fault identification. The ensemble model includes a convolutional neural network, a recurrent neural network, and a fully connected neural network. Each model is trained independently, and a weighted voting process is used to generate a diagnosis result. Weights are assigned based on the performance of each model on the validation set.

[0095] B2: Fault types include generator fault, sensor fault, electrical system fault and mechanical system fault.

[0096] Specifically, generator faults include stator winding faults, rotor faults, bearing faults, cooling system faults, etc. Stator winding faults are mainly manifested as insulation aging, inter-turn short circuits, ground faults, etc., and are characterized by unbalanced stator current, abnormally high temperature, increased vibration, etc. Early signs of stator winding insulation aging include decreased insulation resistance, increased dielectric loss factor, and increased partial discharge activity. Inter-turn short circuit faults will cause the short-circuit loop current to increase, generating additional heat and mechanical stress, further accelerating insulation aging. Ground faults usually occur at the winding ends or slots and are caused by insulation damage or contamination.

[0097] Rotor faults include broken rotor windings, worn slip rings, and rotor imbalance, characterized by abnormal excitation current, elevated rotor temperatures, and mechanical vibration. Broken rotor windings can lead to unstable excitation current and fluctuating generator output voltage. Slip ring wear increases contact resistance, generating sparks and heat, impairing the proper functioning of the excitation system. Rotor imbalance can cause periodic mechanical vibration, which can, in severe cases, damage the shafting.

[0098] Bearing failure is a common mechanical fault in generators and can include wear, poor lubrication, and improper installation. Early signs of bearing failure include increased bearing temperature, the presence of characteristic frequencies in the vibration spectrum, and increased metal particles in the lubricating oil. Vibration signal analysis can identify different types of bearing defects, including inner race faults, outer race faults, and rolling element faults.

[0099] Sensor failures include decreased measurement accuracy, signal drift, and sensor failure. Sensor failures are identified by comparing measurement results from redundant sensors and analyzing signal continuity and rationality. A sensor health assessment model is established to monitor sensor operating conditions in real time. When a sensor failure is detected, it automatically switches to a backup sensor or replaces it with an estimated value. Sensor failure diagnosis methods include signal verification, trend analysis, and statistical testing. Signal verification verifies that sensor output is within a reasonable range, trend analysis observes the long-term trend of the signal, and statistical testing analyzes the statistical characteristics of the signal.

[0100] Electrical system faults include transformer failures, switchgear failures, protection system failures, and cable faults. Transformer faults primarily manifest as insulation degradation, partial discharge, and oil quality deterioration. Diagnosis is performed through methods such as oil chromatography and partial discharge testing. The composition and content of dissolved gases in transformer oil can reveal the type and severity of internal transformer faults. Elevated hydrogen levels typically indicate localized overheating, elevated acetylene levels indicate arc faults, and elevated carbon monoxide and carbon dioxide levels indicate fiber insulation degradation.

[0101] Switchgear faults, including circuit breaker failure, contactor erosion, and fuse blown, are diagnosed through electrical parameter monitoring and mechanical property testing. A circuit breaker's mechanical characteristics include parameters such as operating time, travel curve, and operating force. Changes in these parameters can reflect the breaker's health. Contactor faults typically manifest as contact erosion, spring fatigue, and coil failure, which can be detected through methods such as contact resistance and insulation resistance measurements.

[0102] Mechanical system failures include gearbox failures, main shaft failures, yaw system failures, and pitch control system failures. Gearbox failures primarily manifest as tooth wear, bearing damage, and oil deterioration. Diagnosis is performed through methods such as vibration analysis, oil analysis, and temperature monitoring. Gearbox vibration signals contain a wealth of fault information. Spectral analysis can identify gear meshing frequencies and bearing characteristic frequencies, assessing the health of gears and bearings. Oil analysis can detect parameters such as metal particles, moisture, and acidity in the oil, assessing the lubrication status and internal wear of the gearbox.

[0103] Spindle failures include bearing damage, seal failure, and loose connections, which are diagnosed through vibration and temperature monitoring. Spindle bearings bear heavy radial and axial loads and are prone to fatigue damage. Seal failure can lead to grease leakage and the ingress of external contaminants, accelerating bearing wear. Loose connections can alter the dynamic characteristics of the shaft system and produce abnormal vibration.

[0104] It should be noted that the fault classification system adopts a hierarchical structure, classifying faults at the system, subsystem, and component levels. Each fault level has corresponding characteristic parameters and diagnostic methods, forming a complete fault diagnosis knowledge base. Fault severity is divided into four levels: minor, moderate, severe, and critical, determined by the impact of the fault on system safety and economic efficiency. Minor faults do not affect normal system operation, but development trends require attention; moderate faults may affect system performance and require maintenance; severe faults will cause a significant decrease in system performance and require prompt repair; critical faults threaten system safety and require immediate shutdown for resolution.

[0105] In the embodiment of the present application, the fault diagnosis result is output and an early warning is issued in step S300, which includes the following steps C1-C2:

[0106] C1: Output the fault diagnosis results in a visual manner, including:

[0107] Generate fault type report; generate fault location information; generate fault severity assessment results.

[0108] Specifically, the fault type report uses a standardized format and includes information such as the time of fault occurrence, fault type, fault probability, relevant parameter change trends, possible cause analysis, and recommended treatment measures. Report generation is templated, automatically populating the corresponding diagnostic information and treatment recommendations based on the different fault types. The time of fault occurrence is recorded using a high-precision timestamp, accurate to the millisecond level, to facilitate the precise location and analysis of the fault event. The fault probability is expressed as a percentage, reflecting the confidence level of the diagnostic results. Relevant parameter change trends are displayed in graphical form, including parameter change curves before and after the fault occurs, facilitating analysis of the fault's development process.

[0109] Fault type reports also include historical fault statistics, analyzing fault patterns and development trends. Similar fault cases are searched in the fault database, providing reference handling experience and expected repair times. Historical statistics include data such as the frequency of similar faults, average repair time, and repair cost, providing a reference for maintenance decisions. Reports support multiple output formats, including text, charts, and audible alarms, to meet the needs of diverse users.

[0110] Fault location information is provided through a 3D visualization model, with the specific location of the fault marked on the 3D model of the wind turbine. Positioning accuracy reaches the component level, accurately indicating the specific device or component where the fault occurred. The 3D model contains detailed structural information about the wind turbine, including the precise position and dimensions of major components such as the tower, nacelle, blades, generator, and gearbox. Location information includes detailed information such as the equipment number, installation location, maintenance access, and required tools, providing complete guidance for maintenance personnel. Equipment numbers use standardized coding rules and include information such as the equipment type, installation location, and manufacturer. Maintenance access information includes optimal approach paths, safety precautions, and required permissions.

[0111] In an optional implementation, fault location can be integrated with augmented reality technology. Maintenance personnel can scan the device's QR code with their mobile device to overlay fault information and maintenance instructions on the physical device, improving maintenance efficiency and accuracy. The augmented reality system can identify the device's position and posture in real time and accurately overlay virtual information. This virtual information includes fault location markers, maintenance instructions, safety warnings, and other content, presented in the form of graphics, text, and animations.

[0112] Fault severity assessments are calculated based on a fault impact analysis model. This model considers the impact of a fault on power generation loss, equipment damage risk, personnel safety risk, and repair costs, among other factors. Severity scores are assigned on a scale of 0 to 100, with 0 to 25 representing a minor fault, 25 to 50 representing a moderate fault, 50 to 75 representing a severe fault, and 75 to 100 representing a critical fault. Scoring is calculated using a weighted average method, with the weighting of each factor determined based on expert experience and historical data.

[0113] The power generation loss assessment is determined by calculating the power drop and downtime caused by the fault. Power drop is estimated based on the fault type and severity, while downtime is estimated based on repair complexity and spare parts availability. Equipment damage risk assessment considers the potential secondary damage and cascading failures caused by the fault. Personnel safety risk assessment considers the potential safety threats posed to maintenance personnel by the fault. Repair cost assessments include labor costs, spare parts costs, and downtime losses.

[0114] The assessment results also include fault development trend predictions. By analyzing historical fault data and current operating status, the system predicts the likely rate of fault development and its ultimate impact. This trend prediction utilizes time series analysis to establish a mathematical model of fault development. Model parameters are determined by fitting historical data, and prediction accuracy is assessed through cross-validation. In the event of a critical fault, the system automatically initiates an emergency shutdown procedure to ensure the safety of personnel and equipment.

[0115] It should be noted that the visualization output uses a multi-level display format, including overview, detailed, and trend views. The overview view displays the operating status and fault distribution of the entire wind farm, using geographic information system technology to mark the location and status of each unit on an electronic map. The detailed view displays detailed fault information for a single unit, including fault type, location, severity, and handling recommendations. The trend view shows historical and predicted fault development trends, using time series charts to illustrate the evolution of fault parameters.

[0116] C2: Send to the control center for remote monitoring and fault warning, including: sending fault diagnosis results to the wind power system control center through the communication interface; triggering the corresponding level of fault warning signal; recording the fault occurrence time and processing status.

[0117] Specifically, the communication interface utilizes a variety of communication methods, including Ethernet, fiber optics, and wireless communication. Industrial Ethernet is the primary communication method, using a standard transmission protocol to ensure compatibility with monitoring systems from different manufacturers. Communication data is encrypted for security. Ethernet communication utilizes a star topology, with each wind turbine connected to the control center via a network switch. Fiber optic communication is used for long-distance transmission, offering strong anti-interference capabilities and long transmission distances. Wireless communication serves as a backup communication method, providing emergency communication support in the event of a wired communication failure.

[0118] Fault diagnosis results are transmitted using a combination of real-time and batch transmission. Real-time transmission is used for critical and severe faults to ensure the control center receives fault information immediately. Real-time transmission utilizes a priority mechanism, with critical faults receiving the highest priority and preempting communication resources. For minor and moderate faults, batch transmission is used, with summary information sent every 15 minutes to reduce communication burden. Batch transmission data includes fault statistics, trend analysis results, maintenance recommendations, and more.

[0119] The communication system features disconnection reconnection and data caching. When communication is interrupted, the system automatically caches fault information in local storage and retransmits the cached data when communication is restored. The cache can store 72 hours of fault data, ensuring data protection. Cached data is stored in a circular manner, automatically overwriting the oldest data when storage space is insufficient. Data retransmission uses a confirmation mechanism to ensure reliable data transmission.

[0120] Fault warning signals are divided into four levels according to severity, each corresponding to a different warning method. Level 1 warning (critical fault) uses multiple methods, including audible and visual alarms, text message notifications, and phone calls, to ensure that relevant personnel are immediately notified. The audible and visual alarm uses a high-decibel buzzer and flashing red light, creating a clear warning effect in the control room. SMS notifications are sent to the on-duty personnel and technical supervisors, containing basic fault information and handling suggestions. Phone notifications use an automatic dialing system, dialing a preset list of contacts in sequence until a call is answered.

[0121] Level 2 warnings (serious faults) use audible and visual alarms and text message notifications. The intensity of the audible and visual alarms is lower than that of level 1 warnings, and text message notifications are sent to relevant technical personnel. Level 3 warnings (moderate faults) use interface prompts and email notifications. A prominent fault icon is displayed on the monitoring interface, and an email is sent to maintenance personnel. Level 4 warnings (minor faults) only use interface prompts, displaying the fault information in the fault list on the monitoring interface, and no active notification is given to personnel.

[0122] Early warning signals also include a fault confirmation and response mechanism. Upon receiving an early warning signal, control center operators must confirm and respond within a specified timeframe. This confirmation and response process includes fault confirmation, solution development, and personnel dispatch. The system records key time points, including the time the warning was sent, the time it was confirmed, the time the action began, and the time it was completed, for use in evaluating fault handling efficiency. Warnings not confirmed within the specified timeframe are automatically escalated, triggering a higher-level warning signal.

[0123] Fault records are stored in a database to create a complete fault history archive. Records include basic fault information, diagnostic process data, treatment measures, repair results, and experience summaries. Basic fault information includes fault number, occurrence time, fault type, severity, and impact range. Diagnostic process data includes key parameter change curves, feature extraction results, and model output results. Treatment measures include maintenance actions taken, spare parts replaced, and human resources invested. Repair results include repair time, repair effectiveness, and verification test results.

[0124] Fault records support multi-dimensional query and statistical analysis, providing data support for fault prevention and optimizing equipment maintenance strategies. Query dimensions include time, equipment, fault type, and severity. Statistical analysis includes fault frequency statistics, fault trend analysis, fault cause analysis, and repair efficiency analysis. Analysis results are displayed in charts, allowing managers to easily understand equipment health and maintenance effectiveness.

[0125] In an optional implementation, remote monitoring can also integrate a video surveillance system. When a fault occurs, it automatically retrieves video images of the relevant area, providing visual information for fault analysis and resolution. The video surveillance system uses high-definition cameras with night vision, enabling clear images in a variety of lighting conditions. Cameras are installed in key locations within the wind turbine, including the nacelle, tower, and transformer. The video system supports remote control, enabling adjustment of camera angle and focus for optimal viewing angles.

[0126] In another optional implementation, fault warnings can also integrate meteorological information. When severe weather is predicted, fault warning thresholds are adjusted in advance, equipment monitoring is strengthened, and damage caused by meteorological disasters is prevented. Meteorological information includes parameters such as wind speed, temperature, humidity, precipitation, and lightning. Severe weather includes extreme weather such as typhoons, heavy rain, thunderstorms, and ice and snow. Before severe weather approaches, the system automatically lowers the fault warning threshold and increases monitoring frequency to ensure timely detection of abnormal conditions.

[0127] It should be noted that the entire communication and early warning system utilizes a redundant design, with primary and backup communication links and control centers configured to ensure high system reliability. The primary communication link is wired, while the backup link is wireless. In the event of a primary link failure, automatic failover occurs to the backup link. The primary control center is responsible for daily monitoring, while the backup control center takes over in the event of a primary failure. In the event of a primary system failure, the backup system automatically takes over, ensuring continuity of fault monitoring and early warning functions.

[0128] The system is also equipped with an uninterruptible power supply (UPS), which allows continued operation for at least four hours during a utility power outage, ensuring timely transmission of critical fault information. The UPS utilizes a double-conversion online structure, providing voltage stabilization, filtering, and backup functions. The battery pack utilizes maintenance-free lead-acid batteries, offering long service life and high reliability. The power supply system also features remote monitoring, enabling real-time monitoring of parameters such as battery status, load conditions, and remaining time.

[0129] In summary, the present invention realizes comprehensive monitoring and intelligent diagnosis of the operating status of the wind power system through a multi-parameter fusion wind power system fault detection method. This method can collect and analyze multiple operating parameters in real time, accurately identify various fault types, and promptly generate fault reports and warning information, providing reliable protection for the safe and stable operation of the wind power system. Compared with traditional fault detection methods, the present invention has the advantages of high detection accuracy, fast response speed, accurate fault type identification, and high degree of visualization, and can significantly improve the operation and maintenance efficiency and reliability level of the wind power system. Through the application of deep learning algorithms, the system can automatically learn fault modes, continuously improve diagnostic accuracy, and adapt to equipment aging and changes in the operating environment. Multi-parameter fusion technology can comprehensively utilize various monitoring information, improve the comprehensiveness and reliability of fault detection, and reduce false alarms and missed alarms. Visual output and remote monitoring functions provide operation and maintenance personnel with convenient fault handling tools, improving fault response speed and processing efficiency.

[0130] Example 3

[0131] The above is a schematic diagram of a method for detecting faults in a new energy wind power system based on multi-parameter fusion. It should be noted that the technical solution of the system for detecting faults in a new energy wind power system based on multi-parameter fusion and the technical solution of the method for detecting faults in a new energy wind power system based on multi-parameter fusion described above are based on the same concept. For details not described in detail in the technical solution of the system for detecting faults in a new energy wind power system based on multi-parameter fusion in this embodiment, please refer to the description of the technical solution of the method for detecting faults in a new energy wind power system based on multi-parameter fusion described above.

[0132] This embodiment also provides a new energy wind power system fault detection system based on multi-parameter fusion, including:

[0133] Data acquisition module, used to collect operating parameters of wind power system in real time;

[0134] The data processing module is used to extract the characteristics of the operating parameters and obtain the key characteristics related to the fault;

[0135] Fault diagnosis module, which uses deep learning algorithms to analyze key features and determine the fault type of the wind power system;

[0136] The result output module is used to output the fault diagnosis results in a visual manner and send the fault diagnosis results to the control center for remote monitoring and fault warning.

[0137] This embodiment also provides an electronic device suitable for new energy wind power system fault detection based on multi-parameter fusion, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the new energy wind power system fault detection method based on multi-parameter fusion proposed in the above embodiment.

[0138] This embodiment further provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the method for detecting faults in a new energy wind power system based on multi-parameter fusion as proposed in the above embodiment is implemented.

[0139] The storage medium proposed in this embodiment and the new energy wind power system fault detection method based on multi-parameter fusion proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0140] Through the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented with the help of software and necessary general hardware, and of course can also be implemented by hardware. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0141] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A new energy wind power system fault detection method based on multi-parameter fusion, characterized by: The method includes collecting operating parameters of the wind power system in real time, performing feature extraction on the operating parameters, and obtaining key features related to the fault; Analyze the key features based on a deep learning algorithm to determine the fault type of the wind power system; The fault diagnosis results are output in a visual manner and sent to the control center for remote monitoring and fault warning.

2. The new energy wind power system fault detection method based on multi-parameter fusion according to claim 1, characterized in that: The real-time collection of operating parameters of the wind power system includes: Collect wind speed parameters and wind direction parameters; collect generator speed parameters; collect temperature parameters, voltage parameters and current parameters.

3. The new energy wind power system fault detection method based on multi-parameter fusion according to claim 2, characterized in that: The feature extraction of the operating parameters comprises: Extract the changing trend characteristics of wind speed and wind direction; extract the abnormal fluctuation characteristics of generator speed; extract the correlation characteristics of temperature, voltage and current.

4. The new energy wind power system fault detection method based on multi-parameter fusion according to claim 3 is characterized in that: The analysis of key features based on the deep learning algorithm includes: Establish a multi-parameter fusion fault diagnosis model; input key features into the fault diagnosis model; output the fault type recognition probability.

5. The new energy wind power system fault detection method based on multi-parameter fusion according to claim 4 is characterized in that: The fault types include generator fault, sensor fault, electrical system fault and mechanical system fault.

6. The new energy wind power system fault detection method based on multi-parameter fusion according to claim 5, characterized in that: Outputting the fault diagnosis result in a visual manner includes: Generate fault type report; generate fault location information; generate fault severity assessment results.

7. The new energy wind power system fault detection method based on multi-parameter fusion according to claim 6, characterized in that: The remote monitoring and fault warning sent to the control center includes: The fault diagnosis results are sent to the wind power system control center through the communication interface; the corresponding level of fault warning signal is triggered; and the fault occurrence time and processing status are recorded.

8. A new energy wind power system fault detection system based on multi-parameter fusion, based on the new energy wind power system fault detection method based on multi-parameter fusion according to any one of claims 1 to 7, characterized in that: It also includes a data acquisition module for collecting operating parameters of the wind power system in real time; The data processing module is used to extract the characteristics of the operating parameters and obtain the key characteristics related to the fault; Fault diagnosis module, which uses deep learning algorithms to analyze key features and determine the fault type of the wind power system; The result output module is used to output the fault diagnosis results in a visual manner and send the fault diagnosis results to the control center for remote monitoring and fault warning.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the new energy wind power system fault detection method based on multi-parameter fusion according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the new energy wind power system fault detection method based on multi-parameter fusion according to any one of claims 1 to 7 are implemented.

Images