Hydroelectric generating set fault early warning method and system based on SCN and VMD-approximate entropy

Abstract

The invention relates to a hydroelectric generating set fault early warning method based on SCN and VMD-approximate entropy, and the method comprises the steps: building a time domain health model of a hydroelectric generating set based on a random configuration network, and calculating a relative error between a time domain feature actual value and a health value of the hydroelectric generating set, and taking the relative error as a time domain fault symptom index of the hydroelectric generating set; combining a variational mode decomposition algorithm with an approximate entropy algorithm to construct a frequency domain fault symptom index of the unit; and fusing to obtain a time-frequency domain fault symptom index as a unit vibration degradation index, and realizing unit vibration abnormity real-time early warning according to the unit vibration degradation index. According to the invention, effective combination of model driving and data driving is realized; and through real-time monitoring and early warning, potential faults of the unit can be found and processed in time, and the operation reliability of the unit can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault warning of hydro-generator units, and particularly relates to a method and system for fault warning of hydro-generator units based on SCN and VMD-approximate entropy. Background Technique

[0002] As a key device for hydroelectric energy conversion, the safety and stability of hydro-generator units have always been the focus of attention in the power industry. Moreover, hydro-generator units are constantly developing towards large-scale and complex directions, and their operating states are directly related to the hydroelectric energy conversion efficiency, the safety of hydropower stations, and the stability of the power grid. It is of great significance to master the real-time operating state of hydro-generator units, timely detect fault signs, and make rapid and accurate judgments on equipment faults.

[0003] A hydro-generator unit is a complex non-linear dynamic system with highly coupled hydraulic, mechanical, and electrical factors, and its operating state has a crucial impact on the safety of the power station and the stability of the power grid. Conducting research on fault warning of equipment is to detect fault signs in the early stage of equipment deterioration and take corresponding measures to avoid irreversible faults, which has a guiding role in accurately grasping the trend of the unit's operating state. In traditional hydro-generator unit monitoring systems, most use the method of single eigenvalue fixed threshold alarm to achieve the unit fault alarm function. The triggering of the threshold alarm means that the equipment has deteriorated to a certain extent, and it is impossible to achieve true early warning. At the same time, the signals of large mechanical equipment have non-stationary and non-linear characteristics, and a single eigenvalue is difficult to represent the true state of the unit. Therefore, researching and developing effective non-linear and non-stationary signal processing methods and deeply excavating the unit state characteristic information contained in the vibration signal have important guiding significance for accurately analyzing the operating conditions of the unit and reasonably formulating operation and maintenance strategies. Summary of the Invention

[0004] The purpose of the present invention is to address the above problems and provide a method for fault warning of hydro-generator units based on SCN and VMD-approximate entropy. By combining the Stochastic configuration network (SCN) algorithm, the Variational mode decomposition (VMD) algorithm, and the approximate entropy algorithm, they are respectively used to extract the time-domain and frequency-domain characteristic information of the unit vibration signal, and realize the fault warning of the unit based on the time-frequency domain fault symptom index. First, based on the stochastic configuration network, the time-domain fault symptom index of the unit is constructed. Then, based on VMD-approximate entropy and clustering analysis, the frequency-domain fault symptom index of the unit is constructed, and a fault symptom index that integrates the time-frequency domain information of the signal is constructed. Furthermore, based on the Bootstrap sampling method and the Gaussian threshold method, the upper and lower limits of the fault symptom index are calculated, and by judging whether the index exceeds the limit, real-time warning of abnormal unit vibration is realized.

[0005] To achieve the above object, the technical solution provided by the present invention is as follows: A fault warning method for a hydropower unit based on SCN and VMD-approximate entropy, comprising the following steps: Step 1: Establish a time-domain health model for the hydropower unit, using the historical operating condition data under the healthy state of the unit as the input of the model, and using the time-domain characteristics of the unit vibration signal as the output of the model; Step 2: Input the operating condition data in the real-time monitoring sample of the unit into the time-domain health model to obtain the time-domain characteristic health value of the unit vibration signal at the current moment, calculate the relative error between the actual value of the time-domain characteristic and the time-domain characteristic health value, and use it as the time-domain fault symptom index of the unit. Set the time-domain index threshold, and perform preliminary anomaly warning and fault diagnosis according to the real-time time-domain fault symptom index value of the unit vibration signal and the time-domain index threshold; Step 3: Decompose the unit vibration signal under the same operating condition in the healthy state of the unit into modal components in different frequency bands, calculate the approximate entropy value of the intrinsic modal component, and construct an intrinsic modal component-approximate entropy feature vector set under the same operating condition; Use the K-means clustering algorithm to obtain the central vector of the feature vector set as the standard frequency-domain characteristic of the unit under the corresponding operating condition; Step 4: Extract the intrinsic modal component-approximate entropy feature vector of the unit vibration signal in the real-time monitoring sample of the unit, and calculate its Euclidean distance from the central vector of the feature vector set obtained in Step 3 as the frequency-domain fault symptom index of the unit; Step 5: Perform symbol correction on the frequency-domain fault symptom index obtained in Step 4 to keep it identical to the time-domain fault symptom index of the unit, and fuse the corrected frequency-domain fault symptom index with the time-domain fault symptom index to construct a signal time-frequency domain fault symptom index as the unit vibration deterioration index; Step 6: Combine the unit vibration deterioration index, sample the unit vibration signal sample in the normal state, calculate the standard deviation of each sampled sub-sample, obtain the overall standard deviation estimator based on the normal prior distribution of the unit vibration deterioration index, and use the Gaussian threshold method to determine the upper and lower limit values of the unit vibration deterioration index. By judging whether the unit vibration deterioration index exceeds the limit, real-time warning of unit vibration anomalies is realized.

[0006] Preferably, in the above Step 1, the standard deviation and peak-to-peak value of the unit vibration signal are selected as the time-domain characteristics of the vibration signal, and a random configuration network SCN is used to construct a time-domain health model for the hydropower unit, specifically including: Construct a random configuration network, which includes an input layer, a hidden layer, and an output layer, and initialize the weights and biases: ; (1) Among them, , respectively represent the The input weights and biases of a hidden node; Denotes the scale factor, Is the maximum number of hidden nodes in the hidden layer; m Denotes the dimension of the input vector of the hidden node; Select Sigmoid( ) as the activation function and calculate l The output of a hidden node, and the output value of the randomly configured network SCN is: ; (2) ; (3) In the formula , Respectively represent the input weights and biases of the th hidden node; Is the output weight of the l th hidden node; Denotes the kernel function; Denotes the l th hidden node's output; Denotes the l th hidden node's activation function used; Denotes the l th hidden node's output weight; Denotes the initial output is 0; X represents the input of the hidden node; L Denotes the number of hidden nodes; The error of the randomly configured network is: ; In the formula R Denotes the dimension of the output vector; y represents the expected output value; Respectively represent the errors on the 1st, 2nd, …, R th output components.

[0007] Furthermore, the specific steps of step 3 are as follows: Step 3.1: Use the variational mode decomposition method to decompose the vibration signal of the unit under normal operating conditions to obtain a series of intrinsic mode components, and calculate the approximate entropy sequence of the intrinsic mode components; Step 3.2: Calculate the approximate entropy value of the intrinsic mode component, i.e., the VMD-approximate entropy feature, as the high-dimensional frequency domain feature of the unit vibration signal; Step 3.3: Use the k-means algorithm to cluster the approximate entropy features of the intrinsic mode components under different working conditions to obtain the approximate entropy feature health clustering centers of the intrinsic mode components under different working conditions .

[0008] Preferably, the specific steps of step 6 are as follows: Step 6.1: Construct the normal operation state of the unit under a certain similar working condition a sample of fault symptom indicators that are independent and identically distributed at m moments ; where respectively represent the fault symptom indicator variables at the 1st, 2nd, …, mth moments; Step 6.2: Conduct sampling with replacement on the fault symptom indicator sample obtained in Step 6.1 and independently extract Bootstrap fault symptom indicator subsamples with a capacity of ; , and are all subsample variables of the fault symptom indicators obtained by sampling. Repeat k times; Calculate the statistic of each Bootstrap fault symptom indicator subsample, that is, the mean, denoted as ; Step 6.3: For the k subsample statistics obtained in Step 6.2 , calculate its standard deviation ; It is known that under normal conditions, the overall vibration fault symptom indicators of the unit under the same working condition are approximately normally distributed , where represents the mean of the vibration fault symptom indicators of the unit under the same working condition; Then the mean of the subsamples obtained by Bootstrap sampling satisfies the normal distribution , so the standard deviation of the overall sample of the fault symptom indicators can be obtained as ; Step 6.4: Use the Gaussian threshold method to determine the upper and lower limits of the fault symptom indicators: According to the Gaussian threshold method, the probability that the value of the standard deviation of the overall sample of the fault symptom indicators falls within the interval (μ - 3s, μ + 3s) is 99.74%, where μ represents the mean. Therefore, μ + 3s and μ - 3s are respectively used as the upper and lower limit values for monitoring the normal operation of the measured quantity. By judging whether the indicator exceeds the limit, real-time early warning of the abnormal vibration of the unit is realized.

[0009] As another object of the present invention, a hydropower unit fault early warning system based on SCN and VMD - approximate entropy includes: Time - domain health value calculation module: Calculate the time - domain characteristic health value of the unit under the current working condition by using a random configuration network; Time - domain index calculation module: Used to calculate the relative error between the actual value of the time - domain characteristics of the unit vibration signal and the time - domain characteristic health value, and obtain the time - domain fault symptom index value of the unit; Time-domain diagnosis module: Set the threshold of the time-domain fault symptom index of the unit, compare the real-time time-domain fault symptom index value of the unit obtained by the time-domain index calculation module with the time-domain fault symptom index threshold. If the real-time time-domain fault symptom index value of the unit exceeds the threshold, it is determined that the unit has a fault symptom, and a unit fault warning signal is sent through the abnormal alarm module; Standard frequency-domain feature module: Combine the variational mode decomposition method and the approximate entropy algorithm to construct an intrinsic mode component-approximate entropy feature vector set under the same working condition; Obtain the central vector of the feature vector set as the standard frequency-domain feature of the unit vibration signal under the corresponding working condition; Frequency-domain index calculation module: Obtain the intrinsic mode component-approximate entropy feature vector of the unit vibration signal in the real-time monitoring sample of the unit, calculate the distance from the standard frequency-domain feature of the unit vibration signal, and obtain the frequency-domain fault symptom index value; Unit deterioration index calculation module: Integrate the frequency-domain fault symptom index and the time-domain fault symptom index to calculate the real-time unit vibration deterioration index value; Unit deterioration index threshold module: According to the fault symptom index sample data in the normal operation state of the unit, use the sampling method to obtain the standard deviation of the overall sample of the fault symptom index, and use the Gaussian threshold method to determine the upper and lower limits of the fault symptom index; Abnormal alarm module: According to the upper and lower limits of the fault symptom index output by the unit deterioration index threshold module, judge in real time whether the unit vibration deterioration index value exceeds the limit. If the unit vibration deterioration index value exceeds the upper and lower limits of the fault symptom index, a unit fault warning signal is sent, and the corresponding unit deterioration index value, the unit time-domain fault symptom index value, and the unit frequency-domain fault symptom index value are stored in the fault database as unit fault case data; Or the judgment result of the time-domain diagnosis module is that the unit has a fault symptom, and a unit fault warning signal is sent; Fault database: Used to store unit fault case data, and the unit fault case data includes unit vibration signals, unit deterioration index values, unit time-domain fault symptom index values, and unit frequency-domain fault symptom index values.

[0010] Compared with the prior art, the beneficial effects of the present invention include: 1) The present invention combines machine learning and signal decomposition theory, and proposes a fault warning method for hydropower units based on the fusion of SCN time-domain features and VMD-approximate entropy frequency-domain features. It respectively constructs a time-domain fault symptom index based on SCN and a frequency-domain fault symptom index based on VMD-approximate entropy. By using a weighted algorithm to fuse the time-domain and frequency-domain symptom indexes, a time-frequency domain fault symptom index of the unit is obtained, providing more comprehensive monitoring of the unit state. Warning is carried out by fusing multiple indexes, making the fault diagnosis more reliable and reducing the possibility of false alarms and missed alarms. The upper and lower limit values of the unit vibration deterioration index are determined by using the Gaussian threshold method, enabling the warning threshold to be dynamically adjusted according to the unit state, and improving the sensitivity and accuracy of the warning.

[0011] 2) The present invention uses historical health data to establish a time-domain health model, and at the same time uses real-time monitoring data for verification and warning, realizing the effective combination of model-driven and data-driven. Through real-time monitoring and warning, potential faults of the unit can be discovered and processed in time, which is beneficial to improving the reliability of the unit operation.

[0012] 3) The present invention decomposes the unit vibration signal into modal components in different frequency bands and calculates their approximate entropy values, realizing the refined extraction of signal features and improving the accuracy of fault identification.

[0013] 4) The present invention uses the Bootstrap method to sample the vibration deterioration index samples under normal conditions, calculates the standard deviation of each Bootstrap subsample, obtains the overall standard deviation estimator based on the normal prior distribution of the deterioration index, and then uses the Gaussian threshold method to determine the upper and lower limit values of the deterioration index. By judging whether the index exceeds the limit, real-time warning of abnormal unit vibration is realized, enabling the warning threshold to be dynamically adjusted according to the unit state, and improving the sensitivity and accuracy of the warning.

[0014] 5) The present invention can perform real-time online fast-line diagnosis and identification of the unit fault symptoms according to the time-domain fault symptom index values of the unit, which is beneficial to discovering and processing potential faults of the unit as early as possible, and improving the safety and stability of the unit operation.

[0015] 6) The present invention adopts an incremental database, and regularly retrains the hypersphere model used to determine the threshold of the time-domain fault symptom index with the unit fault case data obtained by the method of the present invention, improving the adaptability of the time-domain fault symptom index threshold to the change of the unit over time, being able to adapt to the latest changes in the unit operation conditions, and ensuring the accuracy and precision of the fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will be further described below in conjunction with the drawings and embodiments.

[0017] Figure 1Schematic flow diagram of the hydro-generator unit fault warning method according to the embodiments of the present invention.

[0018] Figure 2 Schematic flow diagram of calculating the time-domain symptom indexes of the computer set according to the embodiments of the present invention.

[0019] Figure 3 Schematic flow diagram of calculating the frequency-domain symptom indexes of the computer set according to the embodiments of the present invention.

[0020] Figure 4 Schematic flow diagram of unit fault warning according to the unit vibration deterioration index according to the embodiments of the present invention. Detailed implementation manners

[0021] As Figure 1 shown, the hydro-generator unit fault warning method based on SCN and VMD-approximate entropy includes the following steps: Step 1: Establish a time-domain health model for the hydro-generator unit, using the historical working condition data under the healthy state of the unit as the input of the model, and using the time-domain characteristics of the vibration signal as the output of the model; In the embodiment, the standard deviation and peak-to-peak value of the unit vibration signal are selected as the time-domain characteristics of the vibration signal, and a random configuration network (SCN) is used to construct a time-domain health model for the hydro-generator unit, which specifically includes: Construct a random configuration network (SCN), where the random configuration network (SCN) includes an input layer, a hidden layer, and an output layer, and initialize the weights and biases: ; (1) Among them, , respectively represent the input weight and bias of the th hidden node; represents the scaling factor, is the maximum number of hidden nodes in the hidden layer; m represents the dimension of the input vector of the hidden node; Select Sigmoid( ) as the activation function, and calculate the output of the l th hidden node. When the number of hidden layer nodes is L- 1 , the output value of the random configuration network (SCN) is: ; (2) ; (3) In the formula , respectively represent the input weight and bias of the th hidden node; is the output weight of the l th hidden node; denotes the kernel function; denotes the l output of the th hidden node; l denotes the activation function used by the th hidden node; l denotes the output weight of the th hidden node; Initial output is 0; X represents the input of the hidden node; L denotes the number of hidden nodes; The error of randomly configuring the network is: ; In the formula R denotes the dimension of the output vector; y represents the expected output value; respectively denote the errors on the 1st, 2nd,..., R th output components; Introduce a supervision mechanism and randomly assign the input weights and errors of the hidden nodes L : ; (4) In the formula denotes the error on the h th output component; denotes the upper bound of the norm of, denotes the activation function of hidden node L, , , denotes a positive real number, r denotes the regularization parameter, whose range is [0,1]; denotes the learning rate parameter of hidden node L; ; Calculate the output weights of the hidden layer using the least squares method: ; (5) Continuously increase the number of hidden layer nodes to minimize the error value and obtain the optimal model.

[0022] Step 2: Input the operating condition data in the real-time monitoring samples of the unit into the time-domain health model to obtain the time-domain characteristic health value of the vibration signal at the current moment, and calculate the relative error between the actual value and the theoretical value of the time-domain characteristics as the time-domain fault symptom index of the unit, as Figure 2 shown; Set the time-domain index threshold, and perform preliminary abnormal warning and fault diagnosis based on the real-time time-domain fault symptom index value and the time-domain index threshold; In the embodiment, the time-domain health model of the hydropower unit uses the historical operating condition sample data of the unit under normal operating conditions X ( tAs the input, time-domain feature samples Y ( t ) As the output, the expression is: ; (6) ; ; In the formula, represents the mapping function of the stochastic configuration network SCN, represents the water head, represents the active power, is the health value of the peak-to-peak value of the unit vibration signal, is the health value of the standard deviation of the unit vibration signal; t represents time; Input the real-time working condition data of the unit into the time-domain health model of the hydropower unit shown in formula (6) to obtain the time-domain feature health value under the current corresponding working condition, and calculate the difference between the time-domain feature health value and the actual value of the time-domain feature of the unit vibration signal. After weighted summation, it is used as the time-domain fault symptom index of the unit vibration. The expression of the time-domain fault symptom index is: ; (7) In the formula, is the time-domain fault symptom index of the unit vibration signal, , are the symptom indexes corresponding to the standard deviation and peak-to-peak value of the time-domain feature respectively, , are the actual values of the standard deviation and peak-to-peak value of the unit vibration signal respectively, , are the weight coefficients of the standard deviation and peak-to-peak value of the time-domain feature respectively; To avoid the phenomenon of mutual cancellation, and need to maintain the same sign, that is, it is stipulated that the direction of deviation of each time-domain feature component from the normal value remains the same at the same moment; To enhance the sensitivity of the fault symptom index to abnormal data, the calculation formula of the weight coefficient is: ; (8) ; (9).

[0023] Step 3: Decompose the unit vibration signal under the healthy state of the unit under the same working condition into modal components in different frequency bands, calculate the approximate entropy value of the intrinsic modal components, and construct the VMD-approximate entropy feature vector set under the same working condition; Use the K-means clustering algorithm to obtain the central vector of the feature vector set as the standard frequency-domain feature of the vibration signal under the corresponding working condition, as shown in Figure 3 .

[0024] In the embodiment, step 3 specifically includes the following sub-steps: Step 3.1: Decompose the vibration signal of the unit under normal operating conditions using the variational mode decomposition method to obtain a series of intrinsic mode components, and calculate the approximate entropy sequence of the intrinsic mode components; The variational mode decomposition method iteratively searches for a variational model and decomposes the signal into modal components in different frequency bands. The specific calculation method is as follows: Taking the sum of the estimated bandwidths of each modal component as the minimum objective function, the constrained variational problem obtained is: ; (10) In the formula, is the partial derivative function, and the analytic signal , , is the unit impulse function; represents the k-th intrinsic mode component; represents the central frequency of the k-th modal component; represents the original vibration signal of the unit; K is the number of intrinsic mode components; Introduce a quadratic penalty factor and a Lagrange multiplier to transform equation (10) into an unconstrained problem: ; (11) In the formula, is the penalty factor; represents the Lagrange multiplier; L ( ) is the augmented Lagrangian function; Initialize , and in equation (11), and use the alternating direction method of multipliers to solve the constrained variational problem, and iteratively update , and to find the "saddle point" of the augmented Lagrangian function; The result of the n +1-th iteration is: ; (12) ; (13) ; (14) In the formula represents the result of the (n + 1)-th iteration of the k-th modal component in the frequency domain; represents the Fourier transform of the original signal V(t); represents the Fourier transform of the Lagrange multiplier λ(t); represents the Fourier transform of the i-th intrinsic mode component; , respectively represent the nth and (n + 1)th iteration results of the Lagrange multiplier in the frequency domain; Execute the iterative process shown in formulas (12) to (14) until the condition: ; At this time, the K modal components decomposed from the unit vibration signal sequence are obtained; Step 3.2: Calculate the approximate entropy value of the intrinsic mode component, that is, the VMD-approximate entropy feature, as the high-dimensional frequency domain feature of the unit vibration signal; Step 3.3: Use the k-means algorithm to cluster the approximate entropy features of the intrinsic mode components under different working conditions to obtain the approximate entropy feature health cluster centers of the intrinsic mode components under different working conditions .

[0025] Step 4: Extract the VMD-approximate entropy feature vector of the unit vibration signal in the unit real-time monitoring sample, and calculate its Euclidean distance from the central vector of the feature vector set obtained in Step 3 as the frequency domain fault symptom index; Take the Euclidean distance between the VMD-approximate entropy feature vector of the unit real-time vibration signal and the cluster center as the frequency domain fault symptom index of the unit vibration signal, and the calculation formula is: ; (15) In the formula, is the frequency domain fault symptom index of the unit vibration signal, is the VMD-approximate entropy feature vector of the measured vibration signal at time ;

[0026] Similarly, to ensure the identity in the deterioration directions of the frequency domain and the time domain, and should have the same sign.

[0027] Step 5: Correct the sign of the frequency domain fault symptom index obtained in Step 4 to be identical to the time domain fault symptom index, and fuse the corrected frequency domain fault symptom index and the time domain fault symptom index to construct the signal time-frequency domain fault symptom index as the unit vibration deterioration index. Weightedly fuse the unit vibration signal time domain fault symptom index and the unit vibration signal frequency domain fault symptom index to obtain the overall time-frequency domain fault symptom index, as shown in Figure 4 .

[0028] The formula for the weighted fusion of the unit vibration signal time domain feature and the unit vibration signal frequency domain feature is: ; (16) wherein represents the time-frequency domain fault symptom index; and are both weight coefficients, ; (17) ; (18) Finally, a time-frequency domain fault symptom index of the unit vibration signal integrating the overall time domain and frequency domain features is obtained.

[0029] Step 6: Combine the unit vibration deterioration index, sample the unit vibration signal samples in the normal state, calculate the standard deviation of each sampled sub-sample, obtain the overall standard deviation estimator based on the normal prior distribution of the unit vibration deterioration index, and use the Gaussian threshold method to determine the upper and lower limit values of the unit vibration deterioration index. By judging whether the unit vibration deterioration index exceeds the limit, real-time early warning of unit vibration abnormality is realized.

[0030] For the time-frequency domain fault symptom index of the unit vibration signal obtained in Step 5, the overall standard deviation of this index is obtained by using the Bootstrap method. At the same time, the upper and lower limit values of this index are determined by using the Gaussian threshold method. By judging whether this index exceeds the limit, real-time early warning of unit vibration abnormality is realized.

[0031] In the embodiment, Step 6 specifically includes: Step 6.1: Construct a sample of fault symptom indexes that are independently and identically distributed at moments under the normal operating state of the unit under a certain similar working condition ; where respectively represent the fault symptom index variables at the 1st, 2nd,..., mth moments; Step 6.2: Conduct sampling with replacement on the fault symptom index sample obtained in Step 6.1, and independently extract a Bootstrap fault symptom index sub-sample with a capacity of , and are both sampled fault symptom index sub-sample variables, and repeat k times; calculate the statistic of each Bootstrap fault symptom index sub-sample, that is, the mean value, denoted as ; Step 6.3: For the k sub-sample statistics obtained in Step 6.2, calculate its standard deviation ; It is known that under the normal state, the overall fault symptom index of the unit vibration under the same working condition is approximately normally distributed where denotes the mean of the unit vibration fault symptom indicators under the same working condition; the mean of the subsamples obtained by Bootstrap sampling follows a normal distribution , so the standard deviation of the overall sample of the fault symptom indicators can be obtained as ; Step 6.4: Use the Gaussian threshold method to determine the upper and lower limits of the fault symptom indicators: According to the Gaussian threshold method, the probability that the value of the standard deviation of the overall sample of the fault symptom indicators falls within the interval (μ - 3s, μ + 3s) is 99.74%, where μ represents the mean. Therefore, μ + 3s and μ - 3s are used as the upper and lower limits for the normal operation of the monitored quantity respectively. By judging whether the indicator exceeds the limit, real-time early warning of abnormal unit vibration is realized.

[0032] The above-mentioned hydropower unit fault early warning system includes: Time-domain health value calculation module: Calculate the time-domain characteristic health value of the unit under the current working condition by using a random configuration network; Time-domain index calculation module: Used to calculate the relative error between the actual value of the time-domain characteristics of the unit vibration signal and the time-domain characteristic health value, and obtain the time-domain fault symptom index value of the unit; Time-domain diagnosis module: Set the threshold of the time-domain fault symptom index of the unit, compare the real-time time-domain fault symptom index value of the unit obtained by the time-domain index calculation module with the time-domain fault symptom index threshold. If the real-time time-domain fault symptom index value of the unit exceeds the threshold, it is judged that the unit has a fault symptom, and a unit fault early warning signal is sent through the abnormal alarm module; Standard frequency-domain characteristic module: Combine the variational mode decomposition method and the approximate entropy algorithm to construct an intrinsic mode function-approximate entropy feature vector set under the same working condition; Obtain the central vector of the feature vector set as the standard frequency-domain characteristic of the unit vibration signal under the corresponding working condition; Frequency-domain index calculation module: Obtain the intrinsic mode function-approximate entropy feature vector of the unit vibration signal in the real-time monitoring sample of the unit, calculate the distance from the standard frequency-domain characteristic of the unit vibration signal, and obtain the frequency-domain fault symptom index value; Unit deterioration index calculation module: Integrate the frequency-domain fault symptom index and the time-domain fault symptom index to calculate the real-time unit vibration deterioration index value; Unit deterioration index threshold module: According to the sample data of the fault symptom index under the normal operation state of the unit, use the sampling method to obtain the standard deviation of the overall sample of the fault symptom index, and use the Gaussian threshold method to determine the upper and lower limits of the fault symptom index; Abnormal alarm module: According to the upper and lower limits of the fault symptom indicators output by the unit deterioration index threshold module, it judges in real time whether the unit vibration deterioration index value exceeds the limit. If the unit vibration deterioration index value exceeds the upper and lower limits of the fault symptom indicators, it issues a unit fault warning signal, and stores the corresponding unit deterioration index value, unit time-domain fault symptom index value, and unit frequency-domain fault symptom index value as unit fault case data in the fault library; or the judgment result of the time-domain diagnosis module is that the unit has a fault symptom, and it issues a unit fault warning signal; Fault library: Used to store unit fault case data, and the unit fault case data includes unit vibration signals, unit deterioration index values, unit time-domain fault symptom index values, and unit frequency-domain fault symptom index values.

[0033] In the embodiment, the time-domain fault symptom index threshold of the time-domain diagnosis module is re-determined through the hypersphere model by regularly using the time-domain fault symptom index values in the unit fault case data of the fault library, so as to realize the dynamic adaptive update of the frequency-domain fault symptom index threshold.

[0034] The implementation results show that the method of the present invention has the advantages of comprehensive time-domain and frequency-domain analysis, combination of model-driven and data-driven, refined feature extraction, dynamic threshold setting, and early warning by fusing multiple indicators, etc., can improve the accuracy of fault early warning, enhance the reliability of fault diagnosis, optimize maintenance strategies, improve the operation safety of the unit, and reduce maintenance costs.

Claims

1. A hydropower unit fault early warning method based on SCN and VMD-approximate entropy, characterized in that: The following steps are involved: Step 1: Establish a time-domain health model for a hydropower unit, using historical operating data of the unit in a healthy state as input of the model, and using the time-domain characteristics of the unit vibration signal as output of the model; Step 2: Input the operating condition data in the real-time monitoring sample of the unit into the time domain health model, obtain the time domain characteristic health value of the unit vibration signal at the current moment, calculate the relative error between the actual value of the time domain characteristic and the time domain characteristic health value, and use it as the time domain fault symptom index of the unit; set the time domain index threshold, and perform preliminary abnormal warning and fault diagnosis according to the real-time time domain fault symptom index value of the unit vibration signal and the time domain index threshold; Step 3: Decompose the vibration signal of the unit under the same working condition in the healthy state into modal components of different frequency bands, calculate the approximate entropy value of the eigenmodal component, and construct the eigenmodal component-approximate entropy feature vector set under the same working condition; Using a clustering algorithm to obtain a central vector of the feature vector set as a standard frequency domain feature of the unit under the corresponding working condition; Step 4: Extract the intrinsic modal component of the unit vibration signal in the real-time monitoring sample of the unit - the approximate entropy feature vector, and calculate its Euclidean distance with the center vector of the feature vector set obtained in step 3 as the frequency domain fault symptom indicator of the unit; Step 5: Perform sign correction on the frequency domain fault symptom index obtained in step 4 to keep it identical with the time domain fault symptom index of the unit, fuse the corrected frequency domain fault symptom index with the time domain fault symptom index, and construct the signal time-frequency domain fault symptom index as the unit vibration degradation index; Step 6: In combination with the unit vibration degradation index, sample the unit vibration signal samples under normal conditions, calculate the standard deviation of each sample, obtain the overall standard deviation estimate based on the normal prior distribution of the unit vibration degradation index, determine the upper and lower limit values ​​of the unit vibration degradation index, and realize real-time warning of unit vibration abnormality by judging whether the unit vibration degradation index exceeds the limit.

2. The hydropower unit fault early warning method based on SCN and VMD-approximate entropy according to claim 1 is characterized in that: In step 1, the standard deviation and peak-to-peak value of the unit vibration signal are selected as the time domain features of the vibration signal, and a random configuration network (SCN) is used to construct a time domain health model of the hydropower unit, which specifically includes: Construct a random configuration network (SCN), which includes an input layer, a hidden layer, and an output layer, and initialize weights and biases: ；（1） in, , Respectively represent The input weights and biases of hidden nodes; represents the scale factor, is the maximum number of hidden nodes in the hidden layer; m Represents the dimension of the input vector of the hidden node; Select Sigmoid ( ) as the activation function and calculate l The output of the hidden node, the output value of the random configuration network SCN is: ；（2） ；（3） In the formula , Respectively represent The input weights and biases of hidden nodes; For the l The output weights of hidden nodes; represents the kernel function; Indicates l The output of hidden nodes; Indicates l The activation function used by the hidden nodes; Indicates l The output weights of hidden nodes; Indicates that the initial output is 0; X indicates the input of the hidden node; L represents the number of hidden nodes; The error of a randomly configured network is: ； In the formula R represents the dimension of the output vector; y represents the expected output value; Respectively represent the 1st, 2nd, ..., R The error on the output component.

3. The hydropower unit fault early warning method based on SCN and VMD-approximate entropy according to claim 2 is characterized in that: In step 1, a supervision mechanism is introduced into the random configuration network SCN to randomly assign input weights and errors of hidden nodes of the random configuration network SCN: The output weight of the hidden layer is calculated using the least squares method: ；（5） Continuously increase the number of hidden layer nodes to minimize the error value and obtain the optimal random configuration network.

4. The hydropower unit fault early warning method based on SCN and VMD-approximate entropy according to claim 1, 2 or 3, characterized in that: In step 2, the time domain health model of the hydropower unit is based on the historical operating condition sample data of the unit under normal operating conditions. X ( t ) as input, time domain feature samples Y ( t ) as output, the expression is: ；（6） ； ； In the formula, represents the mapping function of the random configuration network SCN, Indicates water head, Indicates active power, is the peak-to-peak health value of the unit vibration signal, is the healthy value of the standard deviation of the unit vibration signal; t Indicates time; The real-time operating condition data of the unit is input into the time domain health model of the hydropower unit shown in formula (6) to obtain the time domain characteristic health value under the current corresponding operating condition, and the difference between the time domain characteristic health value and the actual value of the time domain characteristic of the unit vibration signal is calculated, and the difference is taken as the unit vibration time domain fault symptom index after weighted summation. The expression of the time domain fault symptom index is: ；（7） In the formula, It is the fault symptom index of the unit vibration signal in the time domain. , They are the sign indicators corresponding to the standard deviation of time domain characteristics and the peak-to-peak value, , are the actual values ​​of the standard deviation and peak-to-peak value of the unit vibration signal, , are the weight coefficients of the standard deviation and peak-to-peak value of the time domain characteristics respectively.

5. The hydropower unit fault early warning method based on SCN and VMD-approximate entropy according to claim 4 is characterized in that: In step 2, the calculation formula of the weight coefficient is: ；（8） ； （9） In the formula , are the weight coefficients of the standard deviation and peak-to-peak value of the time domain characteristics respectively.

6. The hydropower unit fault early warning method based on SCN and VMD-approximate entropy according to claim 1, 2, 3 or 5, characterized in that: The step 3 specifically includes the following sub-steps: Step 3.1: Decompose the vibration signal of the unit under normal operating conditions using a variational mode decomposition method to obtain a series of intrinsic mode components, and calculate the approximate entropy sequence of the intrinsic mode components; Step 3.2: Calculate the approximate entropy value of the intrinsic mode component, namely the VMD-approximate entropy feature, as the high-dimensional frequency domain feature of the unit vibration signal; Step 3.3: Use the k-means algorithm to cluster the approximate entropy characteristics of the intrinsic mode components under different working conditions to obtain the healthy clustering centers of the approximate entropy characteristics of the intrinsic mode components under different working conditions. .

7. The hydropower unit fault early warning method based on SCN and VMD-approximate entropy according to claim 6 is characterized in that: In step 4, the VMD-approximate entropy feature vector of the real-time vibration signal of the unit is compared with the cluster center. The Euclidean distance between them is used as the fault symptom index of the unit vibration signal frequency domain, and the calculation formula is: ； （15） In the formula, It is the fault symptom index of the unit vibration signal frequency domain. for VMD-approximate entropy eigenvector of the vibration signal measured at the moment; It is the time domain fault symptom indicator of the unit vibration signal.

8. The hydropower unit fault early warning method based on SCN and VMD-approximate entropy according to claim 7 is characterized in that: In the step 5, the unit vibration signal time domain fault symptom index and the unit vibration signal frequency domain fault symptom index are weightedly fused to obtain an overall time-frequency domain fault symptom index; The formula for weighted fusion of unit vibration signal time domain characteristics and unit vibration signal frequency domain characteristics is: ；（16） In the formula Indicates fault symptom indicators in the time-frequency domain; ；（17） ；（18） In the formula , are all weight coefficients.

9. The hydropower unit fault early warning method based on SCN and VMD-approximate entropy according to claim 8 is characterized in that: In step 6, for the unit vibration signal time-frequency domain fault symptom index obtained in step 5, the Bootstrap method is used to obtain the overall standard deviation of the index, and the upper and lower limits of the index are determined by the Gaussian threshold method. By judging whether the index exceeds the limit, a real-time warning of abnormal unit vibration is achieved; The step 6 specifically includes: Step 6.1: Construct the normal operating state of the unit under similar working conditions Fault symptom indicator samples independent and identically distributed at each moment ;in Respectively represent the fault symptom indicator variables at the 1st, 2nd, …, mth moments; Step 6.2: Sample fault symptom indicators obtained in step 6.1 Sampling with replacement is performed, and the independent sampling capacity is Bootstrap Failure Indicator Subsample , , are all fault symptom indicator sub-sample variables obtained by sampling, repeated k times; calculate the statistic of each Bootstrap fault symptom indicator subsample, that is, the mean, recorded as ; Step 6.3: Get the value obtained in step 6.2 k Subsample statistics , calculate its standard deviation ; Under known normal conditions, the vibration fault symptom indicators of the unit under the same working condition are generally close to normal distribution ,in represents the mean of the unit vibration fault symptom index under the same working condition; then the sub-sample mean obtained by Bootstrap sampling satisfies the normal distribution , so the standard deviation of the overall sample of fault symptom indicators is ; Step 6.4: Use the Gaussian threshold method to determine the upper and lower limits of the fault symptom indicators: According to the Gaussian threshold method, the probability that the value of the overall sample standard deviation of the fault symptom indicator falls in the interval (μ-3s, μ+3s) is 99.74%, where μ represents the mean. Therefore, μ+3s and μ-3s are respectively used as the upper and lower limits of normal operation of the monitored quantity. By judging whether the indicator exceeds the limit, real-time warning of abnormal unit vibration can be achieved.

10. The system of the hydropower unit fault early warning method according to claim 1 or 2 or 3 or 5 or 7 or 8 or 9, characterized in that: The system comprises: Time domain health value calculation module: uses random configuration network calculation to obtain the time domain characteristic health value of the unit under the current operating conditions; Time domain index calculation module: used to calculate the relative error between the actual value of the time domain characteristic of the vibration signal of the computer group and the health value of the time domain characteristic, and obtain the time domain fault symptom index value of the unit; Time domain diagnosis module: sets the time domain fault symptom index threshold of the unit, compares the real-time unit time domain fault symptom index value obtained by the time domain index calculation module with the time domain fault symptom index threshold, and if the real-time unit time domain fault symptom index value exceeds the threshold, it is judged that the unit has a fault symptom, and sends a unit fault warning signal through the abnormal alarm module; Standard frequency domain feature module: Combining the variational mode decomposition method with the approximate entropy algorithm, constructing the intrinsic mode component-approximate entropy feature vector set under the same working condition; obtaining the central vector of the feature vector set as the standard frequency domain feature of the unit vibration signal under the corresponding working condition; Frequency domain index calculation module: obtain the intrinsic modal component-approximate entropy feature vector of the unit vibration signal in the real-time monitoring sample of the unit, calculate the distance from the standard frequency domain feature of the unit vibration signal, and obtain the frequency domain fault symptom index value; Unit degradation index calculation module: integrates the frequency domain fault symptom index with the time domain fault symptom index to calculate the real-time unit vibration degradation index value; Unit degradation index threshold module: Based on the fault symptom index sample data under normal operation of the unit, the standard deviation of the overall sample of the fault symptom index is obtained using the sampling method, and the upper and lower limits of the fault symptom index are determined using the Gaussian threshold method; Abnormal alarm module: According to the upper and lower limits of the fault symptom index output by the unit degradation index threshold module, it is judged in real time whether the unit vibration degradation index value exceeds the limit. If the unit vibration degradation index value exceeds the upper and lower limits of the fault symptom index, a unit fault warning signal is issued, and the corresponding unit degradation index value, unit time domain fault symptom index value and unit frequency domain fault symptom index value are stored in the fault library as unit fault case data; or the judgment result of the time domain diagnosis module is that the unit has a fault symptom, and a unit fault warning signal is issued; Fault database: used to store unit fault case data, the unit fault case data includes unit vibration signals, unit degradation index values, unit time domain fault symptom index values ​​and unit frequency domain fault symptom index values.

Images