Fault diagnosis method and device for generator cooling system and electronic equipment

Abstract

The invention provides a fault diagnosis method and device for a generator cooling system and electronic equipment. The fault diagnosis method comprises the steps of collecting real-time data of the cooling system; preprocessing the collected data by adopting a self-adaptive sparse noise suppression and feature enhancement method; performing feature extraction on the preprocessed data by adopting a time sequence evolution mode feature extraction method to obtain a data set; training a fault diagnosis model by using the feature extracted data set based on an adaptive dynamic clustering network; and analyzing the running state of the cooling system in real time by using the trained fault diagnosis model. The method provided by the invention is suitable for real-time monitoring and fault diagnosis of the generator set cooling system, the signal-to-noise ratio is improved, key characteristic signals of equipment operation are dynamically enhanced, dynamic change characteristics of equipment operation data can be analyzed, intelligent analysis and diagnosis are realized, and the reliability and efficiency of the cooling system are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of evaporative cooling, and relates to a fault diagnosis method for a cooling system, and in particular to a fault diagnosis method, device and electronic equipment for a wind turbine cooling system based on feature enhancement and machine learning. Background Art

[0002] In many industrial fields, cooling systems are an important component to ensure the normal operation of equipment and improve efficiency. Especially in high-power equipment (such as generators, transformers and motors), the performance of the cooling system directly affects the working stability and service life of the equipment. Traditional cooling methods mostly use air or liquid cooling. However, under high temperature and high load environments, these traditional methods often cannot provide sufficient cooling effects, which leads to equipment overheating, reduced efficiency and even failure.

[0003] As an efficient heat dissipation technology, evaporative cooling technology has been widely used in industrial cooling systems in recent years. Evaporative cooling achieves the effect of rapid cooling through the characteristic of liquid absorbing a large amount of heat when evaporating on the surface. It can effectively reduce the operating temperature of the equipment and avoid equipment aging, damage and unstable operation caused by high temperature. This technology is widely used in key equipment in the fields of electricity, communications, metallurgy and energy, especially in the cooling of high-power electrical equipment.

[0004] However, although evaporative cooling technology can significantly improve the cooling effect, the operation of the cooling system still faces many challenges. Since the cooling system usually involves real-time changes in multiple physical parameters such as temperature, pressure, and flow, the complex coupling relationship of these parameters may cause potential failures in the system. For example, problems such as insufficient coolant flow, excessive temperature, and poor condensation effect may all cause a decrease in cooling efficiency, thereby affecting the normal operation of the equipment. Traditional fault diagnosis methods rely on manual inspections and regular maintenance, which not only cannot achieve real-time monitoring, but also easily miss potential problems, resulting in the inability to detect faults in a timely manner, increasing the risk of equipment downtime and maintenance. Therefore, the present invention provides a method for quickly identifying and judging the status of a system. Summary of the invention

[0005] The purpose of the present invention is to provide a fault diagnosis method, device and electronic equipment for a generator cooling system, so as to realize rapid identification and judgment of cooling system faults, realize intelligent fault diagnosis and improve system stability and safety.

[0006] In order to achieve the purpose of the invention, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a fault diagnosis method for a generator cooling system, the fault diagnosis method comprising:

[0008] Collect real-time data of cooling system;

[0009] The collected data are preprocessed using adaptive sparse noise suppression and feature enhancement methods;

[0010] The feature extraction method of time series evolution pattern is used to extract features from the preprocessed data to obtain a data set;

[0011] Training fault diagnosis models using feature-extracted datasets based on adaptive dynamic clustering networks;

[0012] The trained fault diagnosis model is used to analyze the operating status of the cooling system in real time.

[0013] The fault diagnosis method provided by the present invention pre-processes the collected system data based on noise suppression and feature enhancement, can accurately remove the noise in the data, improve the signal-to-noise ratio, retain key fault signals, dynamically enhance the key feature signals of the equipment operation, ensure the high-quality input of subsequent feature extraction and model training, improve the accuracy of diagnosis, and adopt the time series evolution pattern feature extraction method for feature extraction, which can analyze the dynamic change characteristics of the equipment operation data, and extract the key time series features reflecting the normal and abnormal states of the equipment, effectively capture the pattern changes in the system operation, and is particularly suitable for processing the complex dynamic operation state of the cooling system. The method provided by the present invention is suitable for real-time monitoring and fault diagnosis of the cooling system of the generator set, especially in large-scale power plants, which can significantly improve the reliability and efficiency of the cooling system. Through the intelligent diagnosis system, the operation and maintenance personnel can manage the equipment more efficiently, reduce the downtime of the fault, and extend the service life of the equipment. In addition, the system can also help the power plant optimize the operating parameters, reduce energy consumption, and improve the overall economic benefits.

[0014] Preferably, the collected data is preprocessed using an adaptive sparse noise suppression and feature enhancement method, including:

[0015] Model the collected raw data as a time series signal;

[0016] Sparse representation of the signal;

[0017] Set the optimization problem and perform adaptive dictionary learning to optimize the dictionary and sparse coefficients;

[0018] The optimized sparse coefficients are weighted and the signal is reconstructed to obtain a preprocessed signal.

[0019] Preferably, the calculation formula used in the weighted processing is as follows:

[0020]

[0021] Where W is a diagonal weighting matrix, is the optimized sparse coefficient, is the weighted sparse coefficient;

[0022] Preferably, the calculation formula used for the signal reconstruction is as follows:

[0023]

[0024] in, For an optimized dictionary, is the preprocessed signal.

[0025] Preferably, the preprocessing further comprises: interpolating data at missing positions of the preprocessed signal due to data formatting.

[0026] Preferably, the interpolation method comprises:

[0027] The interpolation and reconstruction signals are obtained by using cubic spline interpolation and wavelet transform interpolation methods respectively;

[0028] The weight of each method is calculated according to the error estimate of the interpolated reconstructed signal obtained by each method;

[0029] The interpolated and reconstructed signals are fused according to the weight of each method.

[0030] Preferably, the wavelet transform interpolation method comprises:

[0031] Perform discrete wavelet transform on the preprocessed signal to obtain approximation coefficients and detail coefficients;

[0032] Constructing a soft threshold function for the detail coefficients to optimize;

[0033] According to the optimized detail coefficients and approximation coefficients, the inverse wavelet transform is used to reconstruct the signal and obtain the reconstructed signal of wavelet transform interpolation.

[0034] Preferably, the calculation formula of the weight is as follows:

[0035]

[0036] Among them, e wavelet is the error estimate of the cubic spline interpolation method, e cubic is the error estimate of the wavelet transform interpolation method, ω1 is the weight of the cubic spline interpolation method, and ω2 is the weight of the wavelet transform interpolation method;

[0037] Preferably, the calculation formula used for the signal fusion is as follows:

[0038] x final (n) = ω1x cubic (n)+ω2x wavelet (n)

[0039] Among them, x final (n) is the fused signal, x cubic is the reconstructed signal of cubic spline interpolation, x wavelet The reconstructed signal is interpolated by wavelet transform.

[0040] Preferably, the method of extracting features from preprocessed data using a time series evolution pattern feature extraction method comprises:

[0041] Divide the preprocessed signal into time windows;

[0042] Calculate the local dynamic features of the signal in each time window, and classify the time window into evolution modes according to the local dynamic features;

[0043] Quantify the features that characterize each evolution mode and combine the features into a feature vector;

[0044] Feature screening is performed based on the correlation between the feature vector and the operating status of the cooling system.

[0045] Preferably, the method of training the fault diagnosis model using a feature-extracted data set based on an adaptive dynamic clustering network comprises:

[0046] Establish initial cluster centers;

[0047] Adjust clustering results in real time based on input data;

[0048] Use labeled samples for supervised learning, assign labels to clusters, and mark the cluster status.

[0049] Preferably, the training of the fault diagnosis model further includes: updating and adaptively optimizing the fault diagnosis model, specifically including: collecting real-time data and fault modes of the cooling system, performing incremental training on the model, and dynamically adjusting model parameters.

[0050] Preferably, the fault diagnosis method further comprises: evaluating the severity of the fault, specifically comprising:

[0051] Based on the classification results output by the fault diagnosis model and the severity of known fault modes in historical data, a fault severity quantification model is constructed.

[0052] Set dynamic thresholds for fault severity and classify fault severity into levels;

[0053] The severity of the fault is calculated for the detected fault mode to quantify the impact of different fault modes on the system operation safety and performance. The fitting calculation is performed through a nonlinear function. The calculation formula used in the fitting calculation is as follows:

[0054] s=αM+βlog(1+T)+γP 2

[0055] Among them, s is the severity of the fault, M is the fault mode type, T is the fault duration, P is the impact of the current fault mode on system performance, α, β and γ are weight coefficients, which adjust the impact of M, T and P on the fault severity s respectively;

[0056] According to the dynamic threshold, an alarm signal of a corresponding level is triggered according to the calculated severity of the fault.

[0057] In a second aspect, the present invention provides a fault diagnosis device for a generator cooling system, the fault diagnosis device comprising:

[0058] Real-time data acquisition module, used to collect real-time physical parameters of the cooling system;

[0059] A data preprocessing module, used to denoise the data of the cooling system and enhance fault features based on an adaptive sparse noise suppression and feature enhancement method;

[0060] A feature extraction module, used to obtain a data set of time series evolution features of cooling system parameters by using a time series evolution pattern feature extraction method;

[0061] Model training module, used to build fault diagnosis model using adaptive dynamic clustering network and realize real-time update of the model;

[0062] The fault classification and diagnosis module is used to classify and diagnose system faults on the real-time data of the cooling system and output the classification and diagnosis results.

[0063] In a third aspect, the present invention provides an electronic device, the electronic device comprising:

[0064] at least one processor; and

[0065] a memory communicatively connected to the at least one processor; wherein,

[0066] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the fault diagnosis method of the generator cooling system described in the first aspect.

[0067] Compared with the prior art, the present invention has the following beneficial effects:

[0068] The method provided by the present invention is suitable for real-time monitoring and fault diagnosis of the cooling system of a generator set, improves the signal-to-noise ratio, dynamically enhances the key characteristic signals of equipment operation, can analyze the dynamic change characteristics of equipment operation data, realizes intelligent analysis and diagnosis, and significantly improves the reliability and efficiency of the cooling system. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 is a flow chart of the fault diagnosis method provided in Example 1;

[0070] Figure 2 is a flowchart of step S102 of the fault diagnosis method provided in Example 1;

[0071] Figure 3 is a flowchart of step S103 of the fault diagnosis method provided in Example 1;

[0072] Figure 4 is a flowchart of step S104 of the fault diagnosis method provided in Example 1;

[0073] Figure 5 is a schematic diagram of the structure of the fault diagnosis device provided in Example 2;

[0074] Figure 6 It is a schematic diagram of the structure of an electronic device for implementing a fault diagnosis method provided in Example 3. DETAILED DESCRIPTION

[0075] The technical solution of the present invention is further described below by specific implementation methods. It should be understood by those skilled in the art that the embodiments are only to help understand the present invention and should not be regarded as specific limitations of the present invention.

[0076] Example 1

[0077] This embodiment provides a Figure 1 The fault diagnosis method of the wind turbine evaporative cooling system based on feature enhancement and machine learning shown in the figure comprises the following steps:

[0078] S101. Collect real-time data of the cooling system.

[0079] In the cooling system, a data acquisition system is set up to monitor the physical parameters of the cooling process.

[0080] Exemplarily, the following devices are included: temperature sensor: measures the temperature of the coolant, the ambient temperature and the surface temperature of the equipment; vibration sensor: monitors the operating vibration of the cooling pump or other equipment; pressure sensor: measures the fluid pressure of the coolant to evaluate the flow and pressure changes; flow sensor: monitors the flow of the coolant to ensure the normal circulation of the coolant; dryness sensor: measures the dryness of the coolant during the evaporative cooling process to ensure the cooling effect; PIV (Particle Image Velocimetry) system: used to monitor the flow pattern of the coolant and evaluate the distribution state of the coolant.

[0081] When the generator set is running, multi-dimensional data is collected through sensors and systems to provide comprehensive operating information of the cooling system.

[0082] S102, pre-processing the collected cooling system data to achieve noise suppression and feature enhancement, such as Figure 2 shown.

[0083] Specifically, the collected data is filtered out of abnormal data points caused by sensor failure or external interference, and the adaptive sparse noise suppression and feature enhancement method (ASNRFE) is used to process the data to remove random noise in the data while retaining key fault features. During data collection, if some data is missing, the data is completed by combining cubic spline interpolation and wavelet transform to ensure data continuity and accuracy.

[0084] (2.1) Signal modeling: The collected raw data is modeled as a time series signal x:

[0085]

[0086] in, is a set of real numbers, and n is the dimension of the time series signal, that is, the number of data points in the signal.

[0087] The time series signal x consists of the real signal s and noise m:

[0088] x=s+m

[0089] Among them, s is the real operation signal of the cooling system, and m is the external noise.

[0090] The external noise may be, for example, mechanical vibration during operation of a wind turbine, environmental noise at sea, and electromagnetic interference.

[0091] (2.2) Sparse representation: Sparse representation of the signal s.

[0092] ① Initialize dictionary D:

[0093]

[0094] in, is a set of real numbers, and K is the number of atoms in the dictionary.

[0095] The dictionary D represents different modes of the cooling system operation signal.

[0096] The dictionary D is different modes of sparse signals, that is, different representations or characteristics used to describe the signals in the system, and is obtained by using SVD and / or wavelet transform methods.

[0097] ② Assume that the signal s has a sparse representation under the dictionary D, that is:

[0098] s=Dα

[0099] Among them, α is the sparse coefficient.

[0100]

[0101] (2.3) Noise suppression: Recover and optimize the sparse coefficients from the noisy time series signal x to achieve signal denoising.

[0102] ①Optimize problem setting.

[0103] Construct the following optimization problem:

[0104]

[0105] Among them, ||x-Dα||2 represents the L2 norm, ||α||1 represents the L1 norm, λ is a tuning parameter that balances reconstruction error and sparsity, > 0, and the constraint condition is ||d k ||2=1,d k is the k-th column of dictionary D.

[0106] ② Adaptive dictionary learning: Use alternating minimization algorithm for iterative solution.

[0107] First, the dictionary D is fixed to solve the sparse coefficient α, and then the dictionary D is updated by fixing α, and it is iterated until convergence to obtain the optimized sparse coefficient and dictionary.

[0108] ③The signal can be reconstructed through optimized sparse coefficients and dictionaries:

[0109]

[0110] in, is the denoised signal, For an optimized dictionary, is the optimized sparse coefficient.

[0111] (2.4) Feature enhancement: The sparse coefficients are weighted and used to reconstruct the signal to highlight the fault characteristics of the evaporative cooling system.

[0112] ① The sparse coefficient is weighted using the following calculation formula:

[0113]

[0114] in, is the weighted sparse coefficient, W is a diagonal weighting matrix whose elements are adaptively adjusted according to the significance of the coefficient.

[0115] ② The weighted sparse coefficients are used to reconstruct the signal using the following formula:

[0116]

[0117] in, is the reconstructed signal after preprocessing.

[0118] (2.5) Data interpolation: For the data in the reconstructed signal after preprocessing that have missing positions due to data formatting, a method combining cubic spline interpolation and wavelet transform interpolation is used to interpolate the data.

[0119] ①Use cubic spline interpolation method to reconstruct the signal.

[0120] In the signal interval [x i ,x i +1] Construct the cubic spline function as follows:

[0121] S i (x) = a i (xx i ) 3 +b i (xx i ) 3 +c i (xx i ) 3 +d i

[0122] Among them, a i , b i 、c i and d i are constants used to fit the function.

[0123] Get the reconstructed signal x of cubic spline interpolation cubic .

[0124] ② Use wavelet transform interpolation method to reconstruct the signal.

[0125] Perform discrete wavelet transform (DWT) on the signal to obtain the approximation coefficient and detail factor The expression is as follows:

[0126]

[0127] Among them, x(n) is the signal, Φ L.k (n) is the low-frequency scaling function in discrete wavelet transform, which is used to extract the approximate part of the signal.

[0128]

[0129] Among them, ψ L,k (n) is the wavelet basis function, which is used to extract the high-frequency details of the signal.

[0130] The following soft threshold function is constructed for the detail coefficient:

[0131]

[0132] Among them, c d (k) is the detail coefficient of the signal, which represents the information related to the high-frequency components in the signal. d (k)) is a symbolic function that returns c d The sign of (k) indicates whether the value is positive, negative, or zero.

[0133]

[0134] Where σ is the standard deviation of the noise estimate and N is the signal length.

[0135] The inverse wavelet transform is used to reconstruct the obtained parameters and obtain the reconstructed signal x of wavelet transform interpolation. wavelet The expression is as follows:

[0136]

[0137] ③ Perform weighted fusion on the reconstructed signals obtained by cubic spline interpolation method and wavelet transform interpolation.

[0138] First, the weight ω of each method is calculated according to the error estimate e of each method, and then the signal is fused according to the obtained weight.

[0139] The weight calculation formula is as follows:

[0140]

[0141] Among them, e wavelet is the error estimate of the cubic spline interpolation method, e cubicis the error estimate of the wavelet transform interpolation method, ω1 is the weight of the cubic spline interpolation method, and ω2 is the weight of the wavelet transform interpolation method.

[0142] The calculation formula used for signal fusion is as follows:

[0143] x final (n) = ω1x cubic (n)+ω2x wavelet (n)

[0144] Among them, x final (n) is the fusion signal.

[0145] The final preprocessed signal is obtained.

[0146] Since wind turbines are in an offshore environment all year round, harsh climatic conditions (such as strong winds, salt spray, waves, etc.) will affect the cooling system. The flow, pressure, temperature and other data of the coolant may be interfered by multiple noises. The ASNRFE method can effectively suppress these noises and improve the extraction accuracy of key fault features.

[0147] S103, extracting features from the preprocessed data, such as Figure 3 shown.

[0148] Extract key features that reflect the status of the cooling system from the preprocessed data, monitor the rate of change of the coolant temperature, and determine whether there is insufficient cooling or overheating problems; analyze the equipment vibration pattern to identify whether there is abnormal vibration that may cause equipment damage; monitor the coolant pressure fluctuations to determine whether there is a pipe blockage or pump failure; analyze the coolant flow distribution to identify whether there is coolant flow obstruction. The extracted features are used to map the normal and abnormal states of the system, forming the basis for subsequent diagnosis.

[0149] The time series evolution pattern feature extraction method is used to extract the time series evolution characteristics of the parameters during the operation of the cooling system from the preprocessed signals, so as to capture the dynamic changes of the equipment during operation.

[0150] (3.1) Signal segmentation: Divide the preprocessed signal into time windows, and divide it into multiple time windows W according to fixed length or adaptive length. i , each time window contains N consecutive data points, expressed as:

[0151] W i ={x i,1 ,x i,2 ,…,x i,N}

[0152] In offshore wind turbines, the temperature, pressure, and flow data of the cooling system will change over time. The purpose of segmentation is to divide the long time series into processable windows to capture the data changes within each window. For example, each window can represent a working cycle of the coolant circulation.

[0153] (3.2) Pattern recognition: Calculate the local dynamic characteristics of the signal in each time window, such as the rising rate, falling rate or local extreme point distribution.

[0154] Define the local rate of change CR i The expression is as follows:

[0155]

[0156] Define local volatility CV i The expression is as follows:

[0157]

[0158] Among them, Δt is the sampling interval, std is the standard deviation, and mean is the mean.

[0159] Based on local dynamic characteristics (local change rate CR i and local volatility CV i ), each time window W i Classified into different evolution modes M k , where k represents the mode category, including normal mode, slight abnormal mode and serious fault mode.

[0160] Specifically, when the coolant temperature continues to rise within a time window, CR i It will be displayed as a positive value, and the greater the rate of change, the faster the system temperature rises; if the coolant flow rate fluctuates frequently and violently within a certain period of time, CV i It will show a higher value, thus indicating the instability of the system's operating state.

[0161] In the cooling system of a large offshore wind turbine, this step can identify whether the rate of increase of the coolant temperature is normal, whether there are abnormal fluctuations in the coolant pressure, etc. For example, if the coolant temperature rises sharply in a short period of time, this may be a potential problem caused by a decrease in cooling effect.

[0162] (3.3) Mode quantization: Calculate the frequency of occurrence of each evolution mode in the entire signal f k , average duration T k and standard deviation σ TK , and construct the mode conversion matrix P.

[0163] Frequency of occurrence f k The expression is as follows:

[0164]

[0165] Average duration T k The expression is as follows:

[0166]

[0167] Standard deviation TK The expression is as follows:

[0168]

[0169] The mode transition matrix P describes the transition from one mode M k Switch to another mode M l The probability of is expressed as follows:

[0170]

[0171] In offshore wind turbines, this step can quantify the stability of different cooling system operating modes; for example, the system may enter a high-load operating mode at certain times of the day and a normal operating mode at other times. Through quantification, the duration and changes of these modes can be accurately tracked; the mode transition matrix can reveal the dynamic evolution of the cooling system; for example, if the probability of transitioning from a normal operating mode to a mild abnormal mode increases significantly, it may indicate potential problems in the cooling system, such as blocked coolant flow or reduced heat dissipation efficiency.

[0172] (3.4) Feature aggregation: The quantitative features of the evolution pattern are combined into a feature vector F, and then the correlation coefficient between each feature and the operating state Y of the cooling system is calculated, and the features that are strongly correlated with the state are screened out to form a data set.

[0173] ① The frequency of occurrence of the evolutionary pattern f k , average duration T k , standard deviation σ Tk The key elements of the mode conversion matrix P are combined into the feature vector F, which is specifically expressed as:

[0174] F=[f1,f2,…,f K ,T1,T2,…,T K ,P 1→2 ,P 2→3 ,…]

[0175] The feature vector F is standardized to ensure that the scales of different features are consistent. The calculation formula for the standardization is as follows:

[0176]

[0177] Among them, μ j and σ j are the mean and standard deviation of the j-th feature, respectively.

[0178] ②For each feature F j ’ Calculate the Pearson correlation coefficient between it and the cooling system operating state Y. The calculation formula of the Pearson correlation coefficient is as follows:

[0179]

[0180] Where Y represents the operating status of the cooling system, including normal or faulty, cov(F j ’ ,Y) is the covariance F between the system state Y j ’ , σ is the standard deviation.

[0181] The threshold of the correlation coefficient is set, |ρ|>0.3, to filter out the features that are strongly correlated with the state as the input data set for subsequent fault diagnosis.

[0182] In offshore wind turbines, if certain features (such as the rate of change of the coolant flow pattern) are highly correlated with the fault mode, these features can be used as key inputs for subsequent fault diagnosis models. This time series evolution pattern feature extraction method is particularly suitable for the cooling system of offshore wind turbines. Due to the complex working environment of offshore wind turbines, the workload and environmental conditions of the cooling system (such as wind speed, temperature, humidity, seawater cooling conditions, etc.) will change frequently. Therefore, the time series evolution characteristics of the cooling system operating status can provide valuable reference information for fault diagnosis.

[0183] Through this method, the evolution law of the cooling system under different working modes can be identified, potential failure modes can be predicted, and the health status of the cooling system can be monitored in real time. When the cooling system frequently switches from normal mode to abnormal mode, the system can issue an early warning to help operation and maintenance personnel take timely intervention measures to prevent generator shutdown due to cooling system failure. This method greatly improves the reliability and operating efficiency of wind turbines.

[0184] S104. Build and train a machine learning model for the cooling system, such as Figure 4 shown.

[0185] Based on the extracted features, the model is trained with normal operating conditions and known failure modes in historical data. The model has the ability to automatically identify and classify the operating conditions of the cooling system, including:

[0186] Normal operation status: All parameters of the cooling system are within the normal range and the equipment runs smoothly;

[0187] Mild abnormality: such as a slight decrease in coolant flow, large pressure fluctuations, etc., which have not yet affected the overall performance, but require attention;

[0188] Serious failures: such as interruption of coolant flow, pump failure, etc., may cause the equipment to overheat or shut down.

[0189] Adaptive dynamic clustering network (ADCN) is used to build and train the fault diagnosis model to achieve real-time fault identification and classification of the cooling system. ADCN can dynamically adjust the clustering structure according to the changes in input data, identify new fault modes and perform fault classification.

[0190] (4.1) Establish initial cluster centers.

[0191] Randomly select some samples from the initial sample set as cluster centers C, expressed as:

[0192] C={C1,…,C k}

[0193] Here, k is the number of initial clusters.

[0194] The initial sample set is based on normal operating conditions and known failure modes from historical data.

[0195] (4.2) Adaptive clustering process: The clustering results are adjusted in real time according to the changes in input data.

[0196] Choose an appropriate distance metric such as Euclidean distance or cosine similarity to calculate the similarity between samples and cluster centers.

[0197] When the new input feature F i When it arrives, calculate its distance from all cluster centers C j The similarity between i ,C j ).

[0198] ① Select the cluster C with the highest similarity j As a candidate cluster, if the similarity is less than the set threshold θ, then F i Classify into the candidate cluster C j , and update the cluster center, expressed as:

[0199]

[0200] Among them, n j is cluster C j The current number of samples.

[0201] ②If the similarity between the input feature and all clusters is lower than the threshold θ, a new cluster is created to form a new cluster center.

[0202] (4.3) Machine learning: Use labeled data to train models.

[0203] Use the labels of some labeled samples for supervised learning and assign the label L j Assign cluster C j , and mark the status of the cluster (normal or faulty).

[0204] Adaptive Dynamic Clustering Network (ADCN) automatically classifies new samples without labels based on clustering results:

[0205] If the new sample is classified into a known cluster, the category label of the cluster is assigned; if the new sample forms a new cluster, it is marked as unknown and the category is updated after manual confirmation.

[0206] (4.4) Model adaptive update.

[0207] As the cooling system operates, the model will regularly collect new data, update existing failure modes, and perform incremental training. When external operating parameters or operating conditions change (such as coolant replacement, ambient temperature change), the model will adaptively adjust to ensure the accuracy of fault diagnosis. As the equipment ages, new failure modes may appear in the cooling system. The model is continuously optimized through incremental training to maintain its ability to identify new faults.

[0208] The adaptive dynamic clustering network (ADCN) continuously adjusts the clustering structure according to new data, adds training samples of new faults, and updates existing cluster centers through incremental learning to improve the fault identification ability of the model.

[0209] Regularly calculate the distance D(F) between each cluster center i ,C j ).

[0210] If the distance between two cluster centers is less than the merging threshold θ m , then merge the two clusters to prevent redundant clustering caused by excessive splitting.

[0211] If a cluster has no new samples entering for a long time and the number of samples is lower than the set threshold, the cluster will be pruned to ensure the efficiency of the model.

[0212] According to the aging of cooling system equipment or changes in external operating parameters, the clustering structure of the model is dynamically adjusted to ensure the diagnostic accuracy of the model under different working conditions; the ADCN model will regularly clean up redundant or outdated clusters to keep the model lightweight and efficient.

[0213] S105: Real-time fault diagnosis and classification.

[0214] When the generator set is running, the data collected by the cooling system will be analyzed in real time. The model will classify and diagnose the new data. If the model detects an abnormal state, the system will send an alarm signal in real time to prompt the operation and maintenance personnel to intervene in time; for minor abnormalities, the system recommends preventive maintenance to avoid further deterioration of the problem; for serious faults, the system will immediately alarm and trigger automatic safety measures, such as reducing the power generation load or performing an emergency shutdown to prevent equipment damage.

[0215] The trained fault diagnosis model is used to classify the real-time data of the evaporative cooling system and judge the fault or abnormal status in real time.

[0216] Each input sample is assigned to a cluster with a corresponding cluster label, such as normal operation, minor anomaly, or severe failure.

[0217] When the cooling system enters an abnormal or faulty state, the system sends out an alarm signal to help the operation and maintenance personnel take appropriate maintenance measures in a timely manner. For new fault modes detected, the system will mark them as new faults to remind the operation and maintenance personnel to take further action.

[0218] S106. Assessment of fault severity.

[0219] (6.1) Based on the output classification results and combined with the severity of known fault modes in historical data, a fault severity quantification model is constructed.

[0220] (6.2) Set dynamic thresholds for fault severity, which are divided into three levels: mild, moderate, and severe.

[0221] (6.3) For the fault mode detected in real time, the severity s of the current fault is calculated based on the similarity between its characteristics and the corresponding mode in the historical data, which is used to quantify the impact of different fault modes on the system operation safety and performance. The fitting calculation is performed through a nonlinear function. The calculation formula used in the fitting calculation is as follows:

[0222] s=αM+βlog(1+T)+γP 2

[0223] Where s is the severity of the fault, M is the fault mode type, T is the fault duration, P is the impact of the current fault mode on system performance, and α, β, and γ are weight coefficients, which respectively adjust the impact of M, T, and P on the fault severity s.

[0224] (6.4) Based on the fault severity s calculated in real time, if s exceeds the preset threshold, an alarm signal of the corresponding level is triggered, and the operation and maintenance personnel are guided to adopt different fault response strategies according to the fault level, thereby achieving fast and effective fault handling.

[0225] Example 2

[0226] This embodiment provides a fault diagnosis device for a generator evaporative cooling system. Figure 5 As shown, the fault diagnosis device comprises:

[0227] Real-time data acquisition module 110, data preprocessing module 120, feature extraction module 130, model training module 140 and fault classification and diagnosis module 150. Among them:

[0228] A real-time data acquisition module 110 is used to collect real-time physical parameters of the evaporative cooling system;

[0229] A data preprocessing module 120, for denoising and enhancing fault features of the data of the evaporative cooling system based on an adaptive sparse noise suppression and feature enhancement method;

[0230] A feature extraction module 130, for acquiring a data set of time series evolution features of evaporative cooling system parameters by using a time series evolution pattern feature extraction method;

[0231] A model training module 140 is used to construct a fault diagnosis model using an adaptive dynamic clustering network and to achieve real-time updating of the model;

[0232] The fault classification and diagnosis module 150 is used to classify and diagnose system faults according to the real-time data of the evaporative cooling system, and output classification and diagnosis results.

[0233] In the real-time data acquisition module 110, physical parameters during the operation of the evaporative cooling system are collected, and the physical parameters include temperature, vibration, pressure, flow rate, dryness, flow rate, coolant distribution state and flow pattern.

[0234] In the data preprocessing module 120, the collected real-time data is preprocessed. This includes:

[0235] The collected raw data is modeled as a time series signal.

[0236] Provide a sparse representation of the signal.

[0237] The dictionary and sparse coefficients are optimized through adaptive dictionary learning to reconstruct the signal.

[0238] The reconstructed signal is subjected to feature enhancement to obtain a preprocessed signal.

[0239] The data at the missing positions are interpolated using cubic spline interpolation and wavelet transform interpolation methods.

[0240] In the feature extraction module 130, feature extraction is performed on the preprocessed signal, including:

[0241] Divide the preprocessed signal into time windows.

[0242] The local dynamic features of the signal are calculated in each time window, and the time windows are classified into evolution modes according to the local dynamic features.

[0243] Characterize the characteristics of each evolution mode and construct a feature vector based on the characteristics.

[0244] Feature screening is performed based on the correlation between the feature vector and the operating status of the cooling system to obtain a data set.

[0245] In the model training module 140, the fault diagnosis model is trained, including:

[0246] Establish initial cluster centers.

[0247] The clustering results are adjusted in real time according to the input data.

[0248] Use labeled samples to train machine learning models and assign labels to clusters.

[0249] Update the model based on new data.

[0250] Example 3

[0251] This embodiment provides an electronic device for implementing a fault diagnosis method for a generator evaporative cooling system, such as Figure 6 As shown, the electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.

[0252] like Figure 6As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0253] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a second storage area, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0254] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a fault diagnosis method for a generator evaporative cooling system.

[0255] In some embodiments, the fault diagnosis method of the generator evaporative cooling system may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the fault diagnosis method of the generator evaporative cooling system described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to execute the fault diagnosis method of the generator evaporative cooling system in any other appropriate manner (e.g., by means of firmware).

[0256] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0257] Computer programs for implementing the methods of the present application may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable target determination device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0258] In the context of the present application, a computer readable storage medium may be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, device or equipment. A computer readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer readable storage medium may be a machine readable signal medium. A more specific example of a machine readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0259] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0260] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0261] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0262] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the information expected by the technical solution of this application can be achieved, and this document is not limited here.

[0263] The applicant declares that the above is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Those skilled in the art should understand that any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention shall fall within the protection scope and disclosure scope of the present invention.

Claims

1. A fault diagnosis method for a generator cooling system, characterized in that: The fault diagnosis method comprises: Collect real-time data of cooling system; The collected data are preprocessed using adaptive sparse noise suppression and feature enhancement methods; The feature extraction method of time series evolution pattern is used to extract features from the preprocessed data to obtain a data set; Training fault diagnosis models using feature-extracted datasets based on adaptive dynamic clustering networks; The trained fault diagnosis model is used to analyze the operating status of the cooling system in real time.

2. The fault diagnosis method according to claim 1, characterized in that: The collected data is preprocessed using an adaptive sparse noise suppression and feature enhancement method, including: Model the collected raw data as a time series signal; Sparse representation of the signal; Set the optimization problem and perform adaptive dictionary learning to optimize the dictionary and sparse coefficients; Performing weighted processing on the optimized sparse coefficients and performing signal reconstruction to obtain a preprocessed signal; Preferably, the calculation formula used in the weighted processing is as follows: Where W is a diagonal weighting matrix, is the optimized sparse coefficient, is the weighted sparse coefficient; Preferably, the calculation formula used for the signal reconstruction is as follows: in, For an optimized dictionary, is the preprocessed signal.

3. The fault diagnosis method according to claim 1 or 2, characterized in that: The preprocessing further includes: interpolating data at missing positions of the preprocessed signal.

4. The fault diagnosis method according to claim 3, characterized in that: The interpolation method includes: The interpolation and reconstruction signals are obtained by using cubic spline interpolation and wavelet transform interpolation methods respectively; The weight of each method is calculated according to the error estimate of the interpolated reconstructed signal obtained by each method; The interpolated and reconstructed signals are fused according to the weight of each method; Preferably, the wavelet transform interpolation method comprises: Perform discrete wavelet transform on the preprocessed signal to obtain approximation coefficients and detail coefficients; Constructing a soft threshold function for the detail coefficients to optimize; According to the optimized detail coefficient and approximation coefficient, the inverse wavelet transform is used to reconstruct the signal to obtain the reconstructed signal of wavelet transform interpolation; Preferably, the calculation formula of the weight is as follows: Among them, e wavelet is the error estimate of the cubic spline interpolation method, e cubic is the error estimate of the wavelet transform interpolation method, ω1 is the weight of the cubic spline interpolation method, and ω2 is the weight of the wavelet transform interpolation method; Preferably, the calculation formula used for the signal fusion is as follows: x final (n)=ω1x cubic (n)+ω2x wavelet (n) Among them, x final (n) is the fused signal, x cubic is the reconstructed signal of cubic spline interpolation, x wavelet The reconstructed signal is interpolated by wavelet transform.

5. The fault diagnosis method according to any one of claims 1 to 4, characterized in that: The method of extracting features from the preprocessed data using the time series evolution pattern feature extraction method includes: Divide the preprocessed signal into time windows; Calculate the local dynamic features of the signal in each time window, and classify the time window into evolution modes according to the local dynamic features; Quantify the features that characterize each evolution mode and combine the features into a feature vector; Feature screening is performed based on the correlation between the feature vector and the operating status of the cooling system.

6. The fault diagnosis method according to any one of claims 1 to 5, characterized in that: The method of training a fault diagnosis model using a feature-extracted data set based on an adaptive dynamic clustering network includes: Establish initial cluster centers; Adjust clustering results in real time based on input data; Use labeled samples for supervised learning, assign labels to clusters, and mark the cluster status.

7. The fault diagnosis method according to any one of claims 1 to 6, characterized in that: The training of the fault diagnosis model also includes: updating and adaptively optimizing the fault diagnosis model, specifically including: collecting real-time data and fault modes of the cooling system, performing incremental training on the model, and dynamically adjusting model parameters.

8. The fault diagnosis method according to any one of claims 1 to 7, characterized in that: The fault diagnosis method further includes: evaluating the severity of the fault, specifically including: Based on the classification results output by the fault diagnosis model and the severity of known fault modes in historical data, a fault severity quantification model is constructed. Set dynamic thresholds for fault severity and classify fault severity into levels; The severity of the fault is calculated for the detected fault mode to quantify the impact of different fault modes on the system operation safety and performance. The fitting calculation is performed through a nonlinear function. The calculation formula used in the fitting calculation is as follows: s=αM+βlog(1+T)+γP 2 Among them, s is the severity of the fault, M is the fault mode type, T is the fault duration, P is the impact of the current fault mode on system performance, α, β and γ are weight coefficients, which adjust the impact of M, T and P on the fault severity s respectively; According to the dynamic threshold, an alarm signal of a corresponding level is triggered according to the calculated severity of the fault.

9. A fault diagnosis device for a generator cooling system, characterized in that: The fault diagnosis device comprises: Real-time data acquisition module, used to collect real-time physical parameters of the cooling system; A data preprocessing module, used to denoise the data of the cooling system and enhance fault features based on an adaptive sparse noise suppression and feature enhancement method; A feature extraction module, used to obtain a data set of time series evolution features of cooling system parameters by using a time series evolution pattern feature extraction method; Model training module, used to build fault diagnosis model using adaptive dynamic clustering network and realize real-time update of the model; The fault classification and diagnosis module is used to classify and diagnose system faults on the real-time data of the cooling system and output the classification and diagnosis results.

10. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the fault diagnosis method for the generator cooling system according to any one of claims 1 to 8.

Images