A method for early warning of inter-turn short circuit of a turbine generator rotor winding based on cooperative perception
By combining the sliding window method and Pearson correlation coefficient quantitative analysis with the collaborative sensing method of excitation current and vibration signal, the problem of accurate diagnosis of early defects of inter-turn short circuit in turbine generator rotor winding was solved, achieving higher diagnostic accuracy and sensitivity.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANXI JINGYU POWER GENERATION CO LTD
- Filing Date
- 2022-12-21
- Publication Date
- 2026-06-02
AI Technical Summary
Existing technologies are insufficient to accurately detect early defects in inter-turn short circuits in turbine generator rotor windings, especially under complex operating conditions. The correlation analysis method between excitation current and shaft vibration signal is not sensitive or accurate enough.
The sliding window method is used to acquire the electrical and vibration signals of the DCS system. The correlation between the excitation current and the vibration signal is quantitatively analyzed by the Pearson correlation coefficient. The collaborative gain is calculated and combined with the residual between the predicted and measured values of the excitation current to achieve collaborative sensing fusion diagnosis.
This improves the accuracy and sensitivity of early warning for inter-turn short-circuit faults in the rotor windings of steam turbine generators, ensuring the safe and stable operation of the unit.
Smart Images
Figure CN115932577B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of generator technology, and in particular to a method for early warning of inter-turn short circuits in the rotor windings of a steam turbine generator based on collaborative sensing. Background Technology
[0002] Inter-turn short circuits in the rotor windings of steam turbine generators are a common fault, and in their early stages, they do not pose a threat to the generator's operational safety. However, if left unchecked, they can lead to catastrophic accidents and affect the safety of power production. Therefore, early warning of inter-turn short circuit faults in the rotor windings of steam turbine generators is of great significance.
[0003] Currently, the main methods for diagnosing inter-turn short-circuit faults in turbine generator rotor windings include the excitation current method, virtual power method, leakage flux method, shaft voltage method, and vibration signal method. Among these, the excitation current method is based on the characteristic that an inter-turn short-circuit defect in the rotor winding causes changes in the excitation current. It establishes a generator excitation current prediction model based on normal generator data and diagnoses the inter-turn short-circuit defect by analyzing the residual between the measured and predicted excitation current values. The vibration signal method diagnoses the problem based on the principle that changes in the air gap magnetic field during an inter-turn short circuit in the turbine generator rotor winding cause abnormal vibrations in the shaft. Theoretical analysis shows that the radial vibration amplitude of the shaft is positively correlated with the square of the generator's excitation current (hereinafter referred to as the excitation current square). Based on this, some scholars have proposed a "current-vibration" fusion diagnostic method for inter-turn short circuits in generator rotor windings.
[0004] However, on the one hand, the correlation characteristics of "excitation current-vibration" are not obvious enough when there are minor defects between rotor winding turns. On the other hand, the operating conditions of steam turbine generators are relatively complex, and the excitation current and shaft vibration signals fluctuate greatly during normal operation. This makes it difficult for the "current-vibration" correlation analysis method to accurately detect early defects of short circuits between rotor winding turns. Summary of the Invention
[0005] This invention provides a method for early warning of inter-turn short circuits in the rotor windings of a steam turbine generator based on collaborative sensing, the main purpose of which is to solve the problems existing in the prior art.
[0006] The present invention adopts the following technical solution:
[0007] A method for early warning of inter-turn short circuits in the rotor winding of a steam turbine generator based on collaborative sensing, characterized by the following steps:
[0008] S1. Use the sliding window method to acquire the electrical and vibration signals collected by the turbine generator DCS system over a period of time. The electrical signals include the excitation current I. f The vibration signal includes the lateral vibration amplitude V of the front bearing bush. fxLongitudinal vibration amplitude V of the front bearing fy、 Lateral vibration amplitude V of the rear bearing rx and the longitudinal vibration amplitude V of the rear bearing ry ;
[0009] S2. The correlation between current and vibration is quantitatively analyzed using the Pearson correlation coefficient:
[0010] S21. Calculate the excitation current direction I respectively. f 2 The Pearson correlation coefficients between the four vibration signals were calculated, resulting in four correlation coefficients. The maximum value was denoted as ρ. if～vib ;
[0011] S22. Calculate the Pearson correlation coefficients between the four vibration signals, obtaining six correlation coefficient calculation results. Take the minimum value among them and denote it as ρ. vib～vib ;
[0012] S23. Calculate the cooperative gain g, and thus affect the excitation current direction I. f 2 The correlation characteristics between vibration and other factors are enhanced; the formula for calculating the synergistic gain g is:
[0013]
[0014] In the formula: λ is the correlation threshold, λ∈(0,1); μ is the gain coefficient;
[0015] S3. Calculate the residual between the measured and predicted values of the turbine generator excitation current:
[0016] S31. Based on the historical DCS data of the turbine generator during normal operation, establish a turbine generator excitation current prediction model to obtain the predicted excitation current for that period.
[0017] S32. Calculate the actual excitation current I obtained in step S1. f The predicted excitation current obtained in step S31 The excitation current residual Res between the two is obtained by smoothing the excitation current residual Res to obtain Res';
[0018] S4. Calculate the synergistic gain residual Rg by combining the synergistic gain g and the smoothed excitation current residual Res'. The formula for calculating the synergistic gain residual Rg is:
[0019] Rg=g·Res′
[0020] S5. Determine whether the cooperative gain residual Rg exceeds the set cooperative gain residual threshold, thereby determining the inter-turn insulation status of the turbine generator rotor winding.
[0021] Furthermore, in step S1, the electrical signal also includes stator current I, stator voltage U, active power P, and reactive power Q; in step S32, the stator current I, stator voltage U, active power P, and reactive power Q are used as input features, and the predicted excitation current is calculated by fitting the MLP-Mixer model.
[0022] Furthermore, in step S32, the excitation current residual Res is smoothed using an SG filter to obtain Res′.
[0023] Furthermore, in step S2, the formula for calculating the Pearson correlation coefficient is as follows:
[0024]
[0025] In the formula: ρ represents the correlation coefficient, n is the total number of sample points, and X i and Y i Represents the observed values of two variables. and Then, s represents the mean of the two variables respectively. X and s Y This represents the standard deviation of each of the two variables.
[0026] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0027] This invention first utilizes the physical characteristic that the excitation current square and the radial vibration of the motor shaft are positively correlated when the turbine generator rotor winding is in a state of inter-turn short-circuit fault. Using real-time operating data collected by the turbine generator DCS system, the correlation between the excitation current square and the motor shaft vibration is calculated using the Pearson correlation coefficient. Based on this, a collaborative gain transformation method is proposed to fuse the "current-vibration" correlation coefficient. Combining this with the residual value between the predicted and measured values of the excitation current, the collaborative gain residual is further calculated and compared with a set collaborative gain residual threshold to determine the inter-turn insulation status of the turbine generator rotor winding. Compared to existing diagnostic modes based on a single variable, the "current-vibration" collaborative sensing fusion diagnostic mode provided by this invention can achieve early online warning of inter-turn short-circuit faults in the turbine generator rotor winding, exhibiting higher accuracy and sensitivity, and effectively ensuring the safe and stable operation of the unit. Attached Figure Description
[0028] Figure 1 This is a Pearson correlation coefficient curve of the training dataset in this invention.
[0029] Figure 2This is a Pearson correlation coefficient curve of the test dataset in this invention.
[0030] Figure 3 This is a Pearson correlation coefficient curve of the fault dataset in this invention.
[0031] Figure 4 This is a graph showing the excitation current residual curve of the training dataset in this invention.
[0032] Figure 5 This is a graph showing the excitation current residual curve of the test dataset in this invention.
[0033] Figure 6 This is a graph showing the excitation current residual curve of the fault dataset in this invention.
[0034] Figure 7 This is a frequency distribution diagram of the excitation current residuals of the training dataset, test dataset, and fault dataset in this invention.
[0035] Figure 8 This is a graph showing the collaborative gain residual of the training dataset in this invention.
[0036] Figure 9 This is a graph showing the collaborative gain residual of the test dataset in this invention.
[0037] Figure 10 This is a graph showing the collaborative gain residual of the fault dataset in this invention.
[0038] Figure 11 This is a frequency distribution diagram of the collaborative gain residuals of the training dataset, test dataset, and fault dataset in this invention. Detailed Implementation
[0039] The specific embodiments of the present invention will now be described with reference to the accompanying drawings.
[0040] A method for early warning of inter-turn short circuits in the rotor winding of a steam turbine generator based on collaborative sensing includes the following steps:
[0041] S1. Use the sliding window method to obtain the monitoring data collected by the steam turbine generator DCS system over a period of time. The monitoring data includes electrical signals and vibration signals. Specific data characteristic variables are shown in Table 1.
[0042] Table 1. Characteristic Variables of DCS Data
[0043]
[0044] S2. The Pearson correlation coefficient is used to quantify the correlation between current and vibration. Specifically, the formula for calculating the Pearson correlation coefficient is as follows:
[0045]
[0046] In the formula: ρ represents the correlation coefficient, n is the total number of sample points, and X i and Y i Represents the observed values of two variables. and Then, s represents the mean of the two variables respectively. X and s Y This represents the standard deviation of each of the two variables. ρ∈(0,1) describes the degree of correlation between the two variables. The larger the value, the stronger the positive correlation between the data. A threshold can be set to evaluate the magnitude of the correlation coefficient.
[0047] S21. Calculate the excitation current direction I respectively. f 2 The Pearson correlation coefficients between the four vibration signals were calculated, resulting in four correlation coefficients. The maximum value was denoted as ρ. if～vib Specifically, the four correlation coefficients are: excitation current coefficient I... f 2 Vibration signal V in the X direction of the front bearing fx Correlation coefficient between Excitation current direction I f 2 Vibration signal V in the Y direction of the front bearing fy Correlation coefficient between Excitation current direction I f 2 Vibration signal V in the X direction of the rear bearing rx Correlation coefficient between Excitation current direction I f 2 Vibration signal V in the Y direction of the rear bearing ry Correlation coefficient between
[0048] S22. Calculate the Pearson correlation coefficients between the four vibration signals, obtaining six correlation coefficient calculation results. Take the minimum value among them and denote it as ρ. vib～vib Specifically, the six correlation coefficients include: the X-axis vibration signal V of the front bearing. fx Vibration signal V in the Y direction of the front bearing fy The correlation coefficient ρ between them fx～fy The front bearing X-axis vibration signal V fx Vibration signal V in the X direction of the rear bearing rx The correlation coefficient ρ between them fx～rx The front bearing X-axis vibration signal V fx Vibration signal V in the Y direction of the rear bearing ry The correlation coefficient ρ between them fx～ry The front bearing's Y-axis vibration signal Vfy Vibration signal V in the X direction of the rear bearing rx The correlation coefficient ρ between them fy～rx The front bearing's Y-axis vibration signal V fy Vibration signal V in the Y direction of the rear bearing ry The correlation coefficient ρ between them fy～ry The X-axis vibration signal V of the rear bearing rx Vibration signal V in the Y direction of the rear bearing ry The correlation coefficient ρ between them rx～ry .
[0049] S23. Calculate the cooperative gain g, and thus affect the excitation current direction I. f 2 The correlation characteristics between vibration and other factors are enhanced; the formula for calculating the synergistic gain g is:
[0050]
[0051] In the formula: λ is the correlation threshold, λ∈(0,1); μ is the gain coefficient.
[0052] S3. Calculate the residual between the measured and predicted values of the excitation current of the steam turbine generator.
[0053] S31. Based on the historical DCS data of the turbine generator during normal operation, establish a turbine generator excitation current prediction model to obtain the predicted excitation current for that period. Preferably, in this embodiment, the stator current I, stator voltage U, active power P, and reactive power Q are used as input features, and the predicted excitation current is calculated by fitting an MLP-Mixer model. The specific method involves organizing a 5-dimensional electrical signal with a time duration of T-1 into... As input, the predicted excitation current at time T is calculated by fitting the MLP-Mixer model.
[0054] S32. Calculate the actual excitation current I obtained in step S1. f The predicted excitation current obtained in step S31 The excitation current residual Res between the two is calculated, and then smoothed to obtain Res'. Specifically, the excitation current residual... The excitation current residual Res is smoothed using an SG filter to obtain Res'.
[0055] S4. Calculate the cooperative gain residual Rg by combining the cooperative gain g and the smoothed excitation current residual Res'. Specifically, multiply the cooperative gain g by the absolute value of the excitation current residual after smoothing with the SG filter to obtain the cooperative gain residual Rg. The formula for calculating the cooperative gain residual Rg is:
[0056] Rg=g·Res′.
[0057] S5. Determine whether the cooperative gain residual Rg exceeds the set cooperative gain residual threshold, thereby determining the inter-turn insulation status of the turbine generator rotor winding. The cooperative gain residual threshold is obtained by repeatedly calculating based on the DCS historical data during normal operation of the turbine generator using the above steps.
[0058] The No. 2 generator at a power plant is a 50WT23E-138 type steam turbine generator. During offline maintenance in October 2018, an inter-turn short circuit fault was discovered in its rotor winding. To illustrate the accuracy of the early warning method provided by this invention, the following reasoning and demonstration are conducted using this generator unit as an example:
[0059] (1) Obtain the DCS historical data of the unit during normal operation, determine the training dataset and test dataset, and calculate the collaborative gain residual threshold based on the training dataset and test dataset.
[0060] (1.1) Calculate the Pearson correlation coefficient ρ between the training dataset and the test dataset using the method described in step S2 above. if～vib and ρ vib～vib The calculation results are as follows Figure 1 and Figure 2 As shown.
[0061] (1.2) Using step S3 above, calculate the excitation current residual Res for the training dataset and the test dataset. The calculation results are as follows: Figure 4 and Figure 5 As shown.
[0062] (1.3) Calculate the co-gain residual Rg of the training and test datasets using step S4 above, where λ is 0.7 and μ is 2. The calculation results are as follows: Figure 8 and Figure 9 As shown.
[0063] (1.4) Determine the threshold of the cooperative gain residual based on the cooperative gain residual Rg of the training set dataset and the test dataset, and use it as the judgment criterion for the inter-turn insulation state of the turbine generator rotor winding.
[0064] (2) Obtain historical DCS data for a period of time when the rotor winding of the unit has an inter-turn short circuit fault, and use this data as a fault dataset to test the accuracy of the diagnostic method provided by the present invention.
[0065] (2.1) Calculate the Pearson correlation coefficient ρ of the fault dataset using the method described in step S2 above. if～vib and ρ vib～vib The calculation results are as follows Figure 3 As shown.
[0066] (2.2) The excitation current residual Res of the fault dataset is calculated using step S3 above. The calculation results are as follows: Figure 6 As shown, the frequency distribution of the excitation current residual Re in the fault dataset is compared with that in the training and test datasets. Figure 7 As shown.
[0067] (2.3) The cooperative gain residual Rg of the fault dataset is calculated using step S4 above, where λ is 0.7 and μ is 2. The calculation results are as follows: Figure 10 As shown, the frequency distribution of the co-gain residual Rg of the fault dataset, training dataset, and test dataset is compared to that of the other dataset. Figure 7 As shown.
[0068] (2.4) Compare the cooperative gain residual Rg obtained from the fault dataset with the cooperative gain residual threshold. The cooperative gain residual Rg exceeds the set cooperative gain residual threshold. Therefore, the diagnosis conclusion that there is an inter-turn short circuit fault in the turbine rotor winding can be obtained. This diagnosis conclusion is correct.
[0069] To more clearly demonstrate the advantages of this invention compared to existing technologies, the following section describes the training of a diagnostic model using the excitation current method and vibration signal method mentioned in the background section, based on the aforementioned training and test datasets. Diagnostic tests were then conducted on inter-turn short-circuit faults in the rotor windings of a steam turbine generator using the aforementioned fault dataset, and the accuracy of each diagnostic method was statistically analyzed. A comparison of the diagnostic results for each method is shown in Table 2.
[0070] Table 2 Comparison of Diagnostic Results
[0071]
[0072] As shown in Table 2, when using the DCS data before maintenance (which may be in the early stages of a defect) as the fault dataset for diagnostic testing, the diagnostic accuracy of the excitation current method is 87.35%, the vibration signal method is 61.62%, while the diagnostic accuracy of the "current-vibration" collaborative sensing method provided by this invention is as high as 97.62%. Therefore, compared to single-variable diagnosis, the "current-vibration" collaborative diagnostic model, achieved through residual collaborative gain, can realize timely and accurate prediction of inter-turn short-circuit faults in the turbine generator rotor winding.
[0073] The above are merely specific embodiments of the present invention, but the design concept of the present invention is not limited thereto. Any non-substantial modifications made to the present invention using this concept shall be considered as infringing upon the protection scope of the present invention.
Claims
1. A method for early warning of inter-turn short circuits in the rotor winding of a steam turbine generator based on collaborative sensing, characterized in that: Includes the following steps: S1. Use the sliding window method to acquire the electrical and vibration signals collected by the turbine generator DCS system over a period of time. The electrical signals include the excitation current I. f The vibration signal includes the lateral vibration amplitude V of the front bearing bush. fx Longitudinal vibration amplitude V of the front bearing fy Lateral vibration amplitude V of the rear bearing rx and the longitudinal vibration amplitude V of the rear bearing ry ; S2. The correlation between current and vibration is quantitatively analyzed using the Pearson correlation coefficient: S21. Calculate the excitation current direction I respectively. f 2 The Pearson correlation coefficients between the four vibration signals were calculated, resulting in four correlation coefficients. The maximum value was denoted as ρ. if～vib ; S22. Calculate the Pearson correlation coefficients between the four vibration signals, obtaining six correlation coefficient calculation results. Take the minimum value among them and denote it as ρ. vib～vib ; S23. Calculate the cooperative gain g, and thus affect the excitation current direction I. f 2 The correlation characteristics between vibration and other factors are enhanced; the formula for calculating the synergistic gain g is: In the formula: λ is the correlation threshold, λ∈(0,1); μ is the gain coefficient; S3. Calculate the residual between the measured and predicted values of the turbine generator excitation current: S31. Based on the historical DCS data of the turbine generator during normal operation, establish a turbine generator excitation current prediction model to obtain the predicted excitation current for that period. S32. Calculate the actual excitation current I obtained in step S1. f The predicted excitation current obtained in step S31 The excitation current residual Res between the two is obtained by smoothing the excitation current residual Res to obtain Res'; S4. Calculate the synergistic gain residual Rg by combining the synergistic gain g and the smoothed excitation current residual Res'. The formula for calculating the synergistic gain residual Rg is: Rg=g·Res′ S5. Determine whether the cooperative gain residual Rg exceeds the set cooperative gain residual threshold, thereby determining the inter-turn insulation status of the turbine generator rotor winding.
2. The method for early warning of inter-turn short circuits in the rotor winding of a steam turbine generator based on collaborative sensing as described in claim 1, characterized in that: In step S1, the electrical signal further includes stator current I, stator voltage U, active power P, and reactive power Q; in step S32, the stator current I, stator voltage U, active power P, and reactive power Q are used as input features, and the predicted excitation current is calculated by fitting the MLP-Mixer model.
3. The method for early warning of inter-turn short circuits in turbine generator rotor windings based on collaborative sensing as described in claim 1, characterized in that: In step S32, the excitation current residual Res is smoothed using an SG filter to obtain Res'.
4. The method for early warning of inter-turn short circuits in the rotor winding of a steam turbine generator based on collaborative sensing as described in claim 1, characterized in that: In step S2, the formula for calculating the PearSOn correlation coefficient is: In the formula: ρ represents the correlation coefficient, n is the total number of sample points, and X i and Y i Represents the observed values of two variables. and Then, s represents the mean of the two variables respectively. X and s Y This represents the standard deviation of each of the two variables.