Signal abrupt change fault threshold calculation method and channel switching control process thereof

Abstract

The invention discloses a signal abrupt change fault threshold calculation method and a channel switching control process thereof. The method comprises the following steps: step 1, detecting the type of an input signal; step 2, scheme selection; 3, according to the variances of the signals to be detected calculated by different schemes, determining a composite variance and setting a signal abrupt change fault threshold value; and 4, according to the abrupt change threshold obtained in the step 3, when an abrupt change fault switching condition is met, stopping using the signal and switching to a standby signal channel, and completing signal abrupt change fault alarm and channel switching. The technical problem to be solved by the invention is to provide the signal abrupt change fault threshold calculation method and the channel switching control process thereof, so that the detection and identification of sensor signal abrupt change faults are realized, the inherent problems of a traditional signal abrupt change fault identification mode are solved, the system reliability and the effectiveness of output signals are ensured, and the reliability of the system is improved. And the working performance of the system is optimized.

Description

Technical Field

[0001] This invention relates to the field of water turbine technology, and in particular to a method for calculating the fault threshold of signal sudden change and its channel switching control process. Background Technology

[0002] In hydro-generator control systems, implementing abrupt fault detection functionality for critical signals can effectively improve system reliability and fault tolerance. The setting of the abrupt fault discrimination threshold is crucial in this process. However, existing methods for setting this threshold have limitations and struggle to meet the high reliability and stability requirements of the equipment.

[0003] The setting of the fault discrimination threshold for sudden changes needs to consider many influencing factors, among which signal noise is the main factor, including noise at the measurement point itself and transmission noise at the input. Under stationary conditions, Gaussian noise is a common type of transmission noise, while under non-stationary conditions, Brownian noise is a common type of transmission noise. Transmission noise sources are transmitted to the sensor of the signal under test after passing through various links in the system, which can increase the error of the signal under test and lead to signal measurement distortion.

[0004] Under the influence of transmitted noise, the setting of the fault threshold is related to the influence of each component of the system, involving the solution of system parameters and the action process, thereby obtaining the relationship between parameters and noise under the system structure, and setting the fault discrimination threshold for signal mutation accordingly.

[0005] After a signal failure due to a sudden change, the signal must be stopped and switched to a backup signal channel. This requires setting up appropriate signal failure switching logic. However, directly using an immediate unconditional switching logic that exceeds a threshold is not rigorous. Even if the signal under test changes abruptly and exceeds the threshold, it cannot be determined that the signal sensor is faulty. It is still necessary to rule out whether the change is caused by a change in the input signal of the preceding stage and design an appropriate failure switching procedure accordingly.

[0006] The existing methods for setting thresholds for signal mutation faults and their problems are as follows: (1) Upper and lower limit method. That is, the threshold is set as the upper limit and lower limit of the signal range. This method is relatively crude and the coverage of fault identification is narrow. It can only identify severe mutation faults and is suitable for sensor hardware type faults. It is not very effective for detecting mutation faults in general situations. (2) Empirical method. The fault threshold is set based on experience and lacks theoretical basis. It is difficult to ensure the effectiveness of the threshold setting and there is a hidden danger of failure in fault judgment, which affects the safe operation of the equipment. If the threshold is set too small, it will be falsely activated, and if it is set too large, it will not be activated. There are often cases where the threshold is set too large in the operating equipment. However, since the probability of equipment failure is small and there has been no signal mutation problem for many years, the problem of excessively large threshold setting is not noticed. Once a mutation problem occurs, it is impossible to make an effective judgment and take countermeasures. (3) Fixed threshold. That is, the threshold is set as a fixed value and is not related to the input signal and system parameters. The threshold does not change with the modification of system parameters and does not consider the influence of the input signal characteristics, which does not meet the requirements of actual application. Due to the lack of analysis on the impact of input signal characteristics and system parameters on the threshold, the threshold is not adjusted accordingly when the input signal characteristics or system parameters change, resulting in the failure of the threshold setting and causing missed or false alarms. (4) Single fault judgment. Fault judgment is based solely on whether the threshold is exceeded, without considering whether the sudden change is caused by a sudden change in the previous stage input, leading to misjudgment. Therefore, it is urgent to find a calculation method for the threshold of sudden change in sensor signal and design a corresponding fault switching process accordingly to improve the reliability of equipment operation.

[0007] For input signals consisting of stationary mixed Gaussian noise or non-stationary Brownian noise, this invention provides a method for calculating the fault threshold due to signal abrupt changes and its channel switching control process. This method takes into account both types of noise signals and addresses the shortcomings of existing threshold setting methods. By modeling and calculating the impact of interference noise on the system fault threshold, a system solution model is constructed. Based on this model, the fault threshold is set, and a corresponding signal abrupt change fault switching process is designed, thereby improving the operational stability of the hydro-generator unit. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to provide a method for calculating the threshold of signal sudden change fault and its channel switching control process, so as to realize the detection and identification of sensor signal sudden change fault, solve the inherent problems of traditional signal sudden change fault identification methods, ensure system reliability and the effectiveness of output signal, and optimize system performance.

[0009] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for calculating the fault threshold of signal sudden change and its channel switching control process, including the following steps: Step 1: Input signal type detection; Step 2: Solution Selection; Step 3: Based on the variance of the signal to be tested calculated according to different schemes, determine the composite variance and set the signal sudden change fault threshold; Step 4: Based on the mutation threshold obtained in Step 3, when the mutation fault switching conditions are met, the signal is deactivated and switched to the backup signal channel to complete the signal mutation fault alarm and channel switching.

[0010] Preferably, in step 1, the input noise signal is divided into two categories: the first category is stationary Gaussian white noise signal, and the second category is non-stationary noise signal. The input signal type detection is as follows: DF test for unit roots: Assume the data production model is ; but ; In the formula, : sequence in The sampled value at time; : sequence in Noise value at any given time; : It follows a mean of 0 and a variance of . The independent and identically distributed normal distribution.

[0011] The difference between adjacent time points in the sequence; Parameters to be tested; Parameters to be tested; To test the input signal To determine whether a random walk applies, the following model is established. ; in, This is the error term.

[0012] Null hypothesis (A unit root exists, and the time series is non-stationary.) Alternative Hypothesis (There is no unit root, so the time series is stationary.) By least squares estimation, we get ; in, , Given the sample size, the DF statistic is... for ; in, for Standard deviation; Given a critical value for testing ,like If the null hypothesis is rejected, the sequence is stationary; if If so, then the null hypothesis is accepted, and the sequence is non-stationary.

[0013] Preferably, for abrupt changes under stationary noise signals, when the input is a stationary mixed Gaussian noise signal, the variance of the signal to be detected is calculated as follows: Step 1: Establish the system structure and transfer function model; Step 2: Gaussian white noise parameter estimation; Step 3: Calculate the mean square spectral density of the stationary process of the signal under test; Step 4: Calculate the autocorrelation function of the signal to be detected; Step 5: Calculate the variance of the signal under test under mixed stationary noise.

[0014] Preferably, step 1 is as follows: The system has The mean square spectral densities of the input signal with stable noise are respectively .

[0015] Based on the system structure, construct a mathematical model of the system and calculate the transfer function from each input signal branch to the signal to be detected. ; Step 2 is described in detail below: No. Road noise variance estimate ; In the formula, : No. Road noise signal The sampled value at time; : No. The mean value of the road noise signal; Its autocorrelation function is ; In the formula, : Impulse function; Power spectral density is ; In the formula, : No. The power spectral density of the input noise signal; : No. The variance of the road noise signal; Step 3 is described in detail below: Calculate the mean square spectral density of each input signal branch after passing through the system and reaching the signal under test. ; In the formula, : No. The mean square spectral density of the input noise signal after passing through the system reaches the signal under test. : abbreviation; : No. The power spectral density of the input noise signal; : No. The transfer function from the input noise signal to the signal under test.

[0016] Preferably, step 4 is as follows: Calculate the autocorrelation function of the signal to be detected. ; The autocorrelation function of the output signal of the first noise signal after it passes through the system and reaches the signal to be detected; : abbreviation; Therefore ; Similarly, Therefore ; The autocorrelation function of the output signal of the second noise signal after it reaches the signal under test through the system; : abbreviation; The autocorrelation function of the superimposed mixed signals is ; In the formula, : The autocorrelation function of the superimposed mixed signals; The cross-correlation function of the input noise signal; :exist At time i, the noise signal from the i-th channel reaches the signal to be detected through the system and outputs the signal value. :exist At time i, the noise signal from the i-th channel reaches the signal to be detected through the system and outputs the signal value. The cross-correlation function of the input noise signal is: ; In the formula, Transfer function Impact response; Transfer function Impact response; : First input noise signal; : Second input noise signal; Similarly, we can conclude that ; In the formula, Transfer function Impact response; Transfer function Impact response; : The i-th input noise signal; : The j-th input noise signal; Therefore, the autocorrelation function after the superposition of mixed signals ; make ; but ; ; In the formula, : The autocorrelation function of the superimposed mixed signals; : Time and The duration of the time interval; Step 5 is described in detail below: The variance of the signal under test in mixed noise is ; In the formula, : The variance of the signal under test under mixed noise; Therefore, the signal to be detected is at an interval of time. The variance of the two sampled values ​​at time is ; From the stationary process, we get ; In the formula, : The expected value of the signal to be detected.

[0017] That is, the interval time of the signal to be detected is The variance of the two sampled values ​​at time is ; In the formula, Under stationary noise, the signal to be detected at intervals of... The variance of the two sampled values ​​at time.

[0018] Preferably, for abrupt changes in non-stationary noise signals, when the input is a non-stationary Brownian noise signal, the variance of the signal to be detected is calculated as follows: Step 1: Construct the system structure and its transfer function model; Step 2: Brownian noise parameter estimation; Step 3: Calculate the system transfer function; Step 4: Derive the variance of the detected signal based on Brownian motion under non-stationary conditions; Step 5: Calculate the variance of the signal under test under mixed non-stationary noise.

[0019] Preferably, step 1 is as follows: The system has The non-stationary noise input signal has the following corresponding Brownian noise intensities: Based on the system structure, construct a mathematical model of the system and calculate the transfer function from each input signal branch to the signal to be detected. ; Step 2 is described in detail below: No. LeBron noise intensity estimate ; In the formula, : No. Lew Brown noise intensity; : No. Road noise signal The sampled value at time; Step 3 is described in detail below: Based on the system transfer function, equivalent ordinary differential equations and boundary conditions are established.

[0020] ; In the formula, System structural parameters; Input signal parameters; The solution to the ordinary differential equation is: ; The transfer function is: ; In the formula, Solutions to ordinary differential equations; : Transfer function.

[0021] Preferably, step 4 is as follows: When the input noise signal exhibits Brownian motion characteristics, i.e., the Brownian noise is At that time, the signal to be detected The solution is ; In the formula, : The standard deviation of the i-th input noise signal; Standard Brownian motion; The formula for calculating the variance of the signal under test under input noise is: ; Integral For Ito's integral (random integral); Substituting the system parameters, the variance of the difference between the sampled values ​​at different times can be calculated as follows: ; Step 5 is described in detail below: ; In the formula, Under non-stationary noise signals, the signal to be detected at intervals of... The variance of the two sampled values ​​at time.

[0022] Preferably, the specific steps for setting the signal abrupt change fault threshold in step 3 are as follows: (1) Output noise variance Output Gaussian white noise obey Normal distribution, autocorrelation function ; In the formula, : The autocorrelation function at the output; : The noise variance at the output terminal; : Impulse function; Its power spectral density is ; In the formula, The unit impact function.

[0023] Therefore ; In the formula, Equivalent to , which is the noise variance value at the output end, is the notation used for the sake of consistency in formula expression; Input noise signal; The output noise signal at the interval is The variance of the two sampled values ​​at time is ; (2) Final composite standard deviation of the signal under test ; Right now ; In the formula, The final composite standard deviation of the signal under test; (3) Setting the mutation threshold Set the false alarm rate for sudden faults to be less than Find the threshold for sudden failure. ; Consult the normal distribution table and select the appropriate quantile. Value (choose the lower bound) such that ; In the formula, : False alarm rate for sudden faults; : Quantiles; : Sudden failure threshold; Standard normal distribution function; Solving ; The threshold for sudden failure can be set as follows: ; Preferably, step 4, signal sudden change fault alarm and channel switching, is as follows: Step 1: Read the sampled value of the input signal to be tested; Step 2: Perform a stationarity test on the input signal. If it is stationary, select a stationary model; otherwise, select a non-stationary model. Step 3: Calculate the fault threshold for signal abrupt changes; Step 4: Determine whether the absolute value of the difference between the current value of the signal under test and the sampled value of the signal under test in the previous scanning cycle is greater than the signal sudden change fault threshold. If it is false, proceed to step 5; otherwise, return to step 4. Step 5: Determine if there is a sudden fault in the first input signal. If false, proceed to step 6; otherwise, return to step 4. Step 6: Determine if there is a sudden fault in the second input signal. If it is false, proceed to step 7; otherwise, return to step 4. Step k+4: Determine if there is a sudden fault in the k-th input signal. If it is false, proceed to step k+5; otherwise, return to step 4. Step k+5: Alarm for sudden change in the signal under test, switch to backup signal.

[0024] This invention provides a method for calculating the fault threshold of a sudden signal change and its channel switching control process, which has the following beneficial effects: 1. Covers multiple noise scenarios, expanding its applicability. Existing methods do not fully consider the diversity of noise signals and are difficult to adapt to signal characteristics under different operating conditions. This invention explicitly classifies input noise signals into two categories: stationary Gaussian white noise and non-stationary Brownian noise. It accurately identifies signal stationarity through the unit root DF test and designs two variance calculation schemes (stationary mixed Gaussian noise variance model and non-stationary Brownian noise variance model), which can cover common noise interference scenarios in hydro-generator control systems and avoid threshold failure problems caused by insufficient noise type adaptation.

[0025] 2. Threshold settings based on theoretical modeling reduce the risk of false positives and false negatives. Existing methods (such as empirical methods and upper / lower limit methods) lack theoretical support and are prone to problems such as "false triggering due to excessively small thresholds, false triggering due to excessively large thresholds," or only being able to identify severe faults. This invention achieves accurate threshold calculation through systematic modeling: 1) Construct a system structure and transfer function model, and combine noise parameter estimation (Gaussian white noise variance, Brownian noise intensity), mean square spectral density calculation, autocorrelation function derivation, etc., to quantify the impact of noise on the detected signal after it is transmitted through the system. 2) Based on the composite variance and normal distribution quantile, a threshold is set, and a "false alarm rate" control index is introduced (such as setting the false alarm rate to be less than the target value). This ensures the rationality of the threshold from a theoretical perspective, significantly reduces false alarms and missed alarms caused by coarse threshold tuning, and improves the reliability of fault detection.

[0026] 3. Dynamically adapt to system and signal changes to avoid threshold "fixed failure". Existing methods often use fixed threshold values ​​that do not adjust to changes in system parameters or input signal characteristics, making them prone to failure after changes in operating conditions. This invention deeply integrates threshold calculation with system parameters (such as PID parameters and relay response time constants) and input signal characteristics (stationary / non-stationary): it dynamically selects the calculation model through stationarity testing, adapting in real-time to changes in system parameters and input signals. This ensures that the threshold always matches the actual operating conditions, avoiding missed or false alarms due to a fixed threshold, and meeting the dynamic adjustment needs of equipment during long-term operation.

[0027] 4. Multi-stage fault determination logic improves fault identification accuracy. Existing methods rely solely on "signal exceeding a threshold" as the basis for sensor fault determination, failing to exclude interference from sudden changes in upstream input signals, which can easily lead to misjudgments. This invention designs a "layered determination" channel switching logic: after detecting that the test signal exceeds a threshold, it further checks each upstream input signal for sudden faults. Only when all upstream inputs are normal is the test signal sensor determined to be faulty and switching is triggered. This effectively eliminates "signal sudden changes not caused by sensor faults" (such as upstream input interference), significantly improving the accuracy of fault determination and reducing the impact of unnecessary channel switching on system stability.

[0028] 5. Enhance system reliability and fault tolerance to ensure stable unit operation. As a specialized solution in the field of hydro-turbine technology, this invention directly improves the performance of the hydro-generator control system through a combination of "precise fault detection + rigorous channel switching": 1) Promptly identify sensor signal abrupt changes to prevent distorted signals from affecting the control logic; 2) After a fault occurs, quickly switch to the backup signal channel to ensure the effectiveness of the output signal and reduce the interference of the fault on the unit operation; 3) Ultimately, the system's reliability and fault tolerance will be optimized, reducing the operational risks of the unit due to signal failures and improving the long-term stable operation capability of the hydro-generator unit. Attached Figure Description

[0029] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart of the method according to Scheme 1 of the present invention; Figure 3 This is a flowchart of the method of Scheme 2 of the present invention; Figure 4 This is a structural diagram of the stable system of the present invention; Figure 5 This is a structural diagram of the non-stationary system of the present invention; Figure 6 This is a flowchart of the fault alarm and channel switching process of the present invention; Figure 7 This is a structural diagram of the opening control mode control system of the present invention; Figure 8 This is a simplified calculation diagram for the present invention; Figure 9 This is a flowchart illustrating the signal mutation threshold setting process of this invention. Detailed Implementation

[0030] like Figure 1 As shown, a method for calculating the fault threshold of a signal sudden change and its channel switching control process includes the following steps: Step 1: Input signal type detection; Step 2: Solution Selection; Step 3: Based on the variance of the signal to be tested calculated according to different schemes, determine the composite variance and set the signal sudden change fault threshold; Step 4: Based on the mutation threshold obtained in Step 3, when the mutation fault switching conditions are met, the signal is deactivated and switched to the backup signal channel to complete the signal mutation fault alarm and channel switching.

[0031] like Figure 2 As shown, preferably, in step 1, the input noise signal is divided into two categories: the first category is stationary Gaussian white noise signal, and the second category is non-stationary noise signal. The input signal type detection is as follows: DF test for unit roots: Assume the data production model is ; but ; In the formula, : sequence in The sampled value at time; : sequence in Noise value at any given time; : It follows a mean of 0 and a variance of . The independent and identically distributed normal distribution.

[0032] The difference between adjacent time points in the sequence; Parameters to be tested; Parameters to be tested; To test the input signal To determine whether a random walk applies, the following model is established. ; in, This is the error term.

[0033] Null hypothesis (A unit root exists, and the time series is non-stationary.) Alternative Hypothesis (There is no unit root, so the time series is stationary.) By least squares estimation, we get ; in, , Given the sample size, the DF statistic is... for ; in, for Standard deviation; Given a critical value for testing ,like If the null hypothesis is rejected, the sequence is stationary; if If so, then the null hypothesis is accepted, and the sequence is non-stationary.

[0034] Preferred, such as Figure 2 As shown, for the abrupt change case under stationary noise signal, when the input is a stationary mixed Gaussian noise signal, the variance of the signal to be detected is calculated as follows: Step 1: Establish the system structure and transfer function model; Step 2: Gaussian white noise parameter estimation; Step 3: Calculate the mean square spectral density of the stationary process of the signal under test; Step 4: Calculate the autocorrelation function of the signal to be detected; Step 5: Calculate the variance of the signal under test under mixed stationary noise.

[0035] Preferred, such as Figure 4 As shown, step 1 is as follows: The system has The mean square spectral densities of the input signal with stable noise are respectively .

[0036] Based on the system structure, construct a mathematical model of the system and calculate the transfer function from each input signal branch to the signal to be detected. ; Step 2 is described in detail below: No. Road noise variance estimate ; In the formula, : No. Road noise signal The sampled value at time; : No. The mean value of the road noise signal; Its autocorrelation function is ; In the formula, : Impact function; Power spectral density is ; In the formula, : No. The power spectral density of the input noise signal; : No. The variance of the road noise signal; Step 3 is described in detail below: Calculate the mean square spectral density of each input signal branch after passing through the system and reaching the signal under test. ; In the formula, : No. The mean square spectral density of the input noise signal after passing through the system reaches the signal under test. : abbreviation; : No. The power spectral density of the input noise signal; : No. The transfer function from the input noise signal to the signal under test.

[0037] Preferably, step 4 is as follows: Calculate the autocorrelation function of the signal to be detected. ; The autocorrelation function of the output signal of the first noise signal after it passes through the system and reaches the signal to be detected; : abbreviation; Therefore ; Similarly, Therefore ; The autocorrelation function of the output signal of the second noise signal after it reaches the signal under test through the system; : abbreviation; The autocorrelation function of the superimposed mixed signals is ; In the formula, : The autocorrelation function of the superimposed mixed signals; The cross-correlation function of the input noise signal; :exist At time i, the noise signal from the i-th channel reaches the signal to be detected through the system and outputs the signal value. :exist At time i, the noise signal from the i-th channel reaches the signal to be detected through the system and outputs the signal value. The cross-correlation function of the input noise signal is: ; In the formula, Transfer function Impact response; Transfer function Impact response; : First input noise signal; : Second input noise signal; Similarly, we can conclude that ; In the formula, Transfer function Impact response; Transfer function Impact response; : The i-th input noise signal; : The j-th input noise signal; Therefore, the autocorrelation function after the superposition of mixed signals ; make ; but ; ; In the formula, : The autocorrelation function of the superimposed mixed signals; : Time and The duration of the time interval; Step 5 is described in detail below: The variance of the signal under test in mixed noise is ; In the formula, : The variance of the signal under test under mixed noise; Therefore, the signal to be detected is at an interval of time. The variance of the two sampled values ​​at time is ; From the stationary process, we get ; In the formula, : The expected value of the signal to be detected.

[0038] That is, the interval time of the signal to be detected is The variance of the two sampled values ​​at time is ; In the formula, Under stationary noise, the signal to be detected at intervals of... The variance of the two sampled values ​​at time.

[0039] Preferred, such as Figure 3 As shown, for the abrupt change case under non-stationary noise signal, when the input is a non-stationary Brownian noise signal, the variance of the signal to be detected is calculated as follows: Step 1: Construct the system structure and its transfer function model; Step 2: Brownian noise parameter estimation; Step 3: Calculate the system transfer function; Step 4: Derive the variance of the detected signal based on Brownian motion under non-stationary conditions; Step 5: Calculate the variance of the signal under test under mixed non-stationary noise.

[0040] Preferred, such as Figure 5 As shown, step 1 is as follows: The system has The non-stationary noise input signal has the following corresponding Brownian noise intensities: Based on the system structure, construct a mathematical model of the system and calculate the transfer function from each input signal branch to the signal to be detected. ; Step 2 is described in detail below: No. LeBron noise intensity estimate ; In the formula, : No. Lew Brown noise intensity; : No. Road noise signal The sampled value at time; Step 3 is described in detail below: Based on the system transfer function, equivalent ordinary differential equations and boundary conditions are established.

[0041] ; In the formula, System structural parameters; Input signal parameters; The solution to the ordinary differential equation is: ; The transfer function is: ; In the formula, Solutions to ordinary differential equations; : Transfer function.

[0042] Preferably, step 4 is as follows: When the input noise signal exhibits Brownian motion characteristics, i.e., the Brownian noise is At that time, the signal to be detected The solution is ; In the formula, : The standard deviation of the i-th input noise signal; Standard Brownian motion; The formula for calculating the variance of the signal under test under input noise is: ; Integral For Ito's integral (random integral); Substituting the system parameters, the variance of the difference between the sampled values ​​at different times can be calculated as follows: ; Step 5 is described in detail below: ; In the formula, Under non-stationary noise signals, the signal to be detected at intervals of... The variance of the two sampled values ​​at time.

[0043] Preferably, the specific steps for setting the signal abrupt change fault threshold in step 3 are as follows: (1) Output noise variance Output Gaussian white noise obey Normal distribution, autocorrelation function ; In the formula, : The autocorrelation function at the output; : The noise variance at the output terminal; : Impact function; Its power spectral density is ; In the formula, The unit impact function.

[0044] Therefore ; In the formula, Equivalent to , which is the noise variance value at the output end, is the notation used for the sake of consistency in formula expression; Input noise signal; The output noise signal at the interval is The variance of the two sampled values ​​at time is ; (2) Final composite standard deviation of the signal under test ; Right now ; In the formula, The final composite standard deviation of the signal under test; (3) Setting the mutation threshold Set the false alarm rate for sudden faults to be less than Find the threshold for sudden failure. ; Consult the normal distribution table and select the appropriate quantile. Value (choose the lower bound) such that ; In the formula, : False alarm rate for sudden faults; : Quantiles; : Sudden failure threshold; Standard normal distribution function; Solving ; The threshold for sudden failure can be set as follows: ; Preferred, such as Figure 6 As shown, step 4, signal sudden change fault alarm and channel switching, is specifically as follows: Step 1: Read the sampled value of the input signal to be tested; Step 2: Perform a stationarity test on the input signal. If it is stationary, select a stationary model; otherwise, select a non-stationary model. Step 3: Calculate the fault threshold for signal abrupt changes; Step 4: Determine whether the absolute value of the difference between the current value of the signal under test and the sampled value of the signal under test in the previous scanning cycle is greater than the signal sudden change fault threshold. If it is false, proceed to step 5; otherwise, return to step 4. Step 5: Determine if there is a sudden fault in the first input signal. If false, proceed to step 6; otherwise, return to step 4. Step 6: Determine if there is a sudden fault in the second input signal. If it is false, proceed to step 7; otherwise, return to step 4. Step k+4: Determine if there is a sudden fault in the k-th input signal. If it is false, proceed to step k+5; otherwise, return to step 4. Step k+5: Alarm for sudden change in the signal under test, switch to backup signal.

[0045] In this embodiment, the unit equipment is a control system device operating under grid-connected operation control mode, which involves determining sudden changes in the input signals of the hydropower unit's control system. In this embodiment, the unit's control system has four noise input signals.

[0046] Unit root DF tests were performed on the frequency difference signal, guide vane setpoint signal, and relay position setpoint signal, respectively. The conclusion was that the frequency difference signal and guide vane setpoint signal were stationary sequences, while the relay position setpoint signal was a non-stationary sequence. Variance calculations were performed using Scheme 1 and Scheme 2 respectively to determine the mutation threshold.

[0047] Option 1: Variance Calculation under Mixed Stationary Noise Signals For the case of sudden changes under stable noise signals, the example focuses on the guide vane opening signal under grid-connected opening control mode. That is, when the input is a stable mixed Gaussian noise signal, the fault threshold of the guide vane opening signal sudden change is calculated and tuned, and the channel switching control process is set.

[0048] Step 1, the unit is in grid-connected state, in the opening control mode, such as Figure 7 The diagram shows the control system structure under the grid-connected operation control mode of the generator unit. The system has two main input signals and four noise input signals, where the main input signals are frequency difference signals, etc. and guide vane given signal The noise input signals are the noise signals at the frequency difference. Noise signal at the given signal of the guide vane Noise signal at the given position signal of the relay device and noise at the guide vane opening signal .

[0049] noise signal and noise signals After passing through the PID control stage and then the servo system of the relay, the signal reaches the guide vane opening output. (Noise signal) After passing through the servo system of the relay, it reaches the guide vane opening output point.

[0050] Noise sources are transmitted to the sensor of the signal under test after passing through various links in the system, which can increase the error of the signal under test and cause signal measurement distortion.

[0051] When the unit is in grid-connected state and in opening control mode, the system transfer function is: ; In the formula, The derivative coefficients of the PID parameters in a control system; The integral coefficients of the PID parameters in the control system; The proportional gain of the PID parameters in the control system; : Laplace operator; Permanent slip coefficient; Laplace transform of the frequency difference function; Laplace transform of the given function of the guide vane; Laplace transform of the guide vane opening function; : servo system gain; : Response time constant of the main relay; : Response time constant of the main pressure regulating valve; Depend on The above formula is approximately equal to ; Right now ; In the formula, : Transfer function of guide vane opening with respect to frequency difference; : The transfer function of the guide vane opening with respect to a given guide vane; in, ; Step 2, mean square spectral density of the guide vane opening during the steady-state process Depend on ; ; Therefore ; ; Step 3: Autocorrelation function of guide vane opening ; Depend on ; have to ; Therefore ; Similarly, ; Therefore ; Compiled from relevant sources ; ; Therefore ; ; ; ; Step 4: Variance of guide vane opening signal under mixed noise ; Therefore ; steady process ; Therefore ; Right now ; Option 2: Variance Calculation under Mixed Non-Stationary Noise Signals In this embodiment, the relay position is given as a non-stationary signal, representing a sudden change under non-stationary noise signal. Therefore, Scheme 2 is selected for implementation. That is, when the input is a non-stationary Brownian noise signal, the fault threshold of the guide vane opening signal sudden change in the grid-connected opening control mode of the unit is calculated, adjusted, and the channel switching control process is set.

[0052] Step 1: Construct the system structure and its transfer function model The control system structure under the grid-connected operation control mode of the unit is shown in the figure. Figure 7 The simplified calculation diagram is shown in Figure 8. The control system structure of the servo system in the grid-connected operation control mode of the unit includes one main input signal and two noise input signals. The main input signal is the main servo position setpoint signal. The noise input signal at this signal point is The other noise input signal is the noise at the guide vane opening signal. Noise signal After passing through the servo system of the relay, it reaches the guide vane opening output point.

[0053] Noise sources are transmitted to the sensor of the signal under test after passing through various links in the system, which can increase the error of the signal under test and cause signal measurement distortion.

[0054] in, ; In the formula, : The response time constant of the main relay after gain reduction; In grid-connected opening control mode, the system transfer function is: ; In the formula, Guide vane opening affects input noise The transfer function; Therefore ; Input noise The Laplace transform of; Depend on Approximately ; Depend on ,have to ; Depend on Approximately ; achievable ; Step 2: Calculate the system transfer function The homogeneous differential equation is ; Solving ; The transfer function is ; Step 3: Derive the guide vane opening variance based on Brownian motion under non-stationary conditions Since the frequency around 50Hz can be considered as Brownian motion in the no-load state, it will be transmitted to the guide vane opening.

[0055] when The solution at that time is ; but ; ; Right now ; The variance of the difference between the sampled values ​​at different times can be obtained as follows: ; in, .

[0056] Then, the threshold for the sudden change in the guide vane opening signal is calculated. (1) Output noise variance Output Gaussian white noise obey Normal distribution, autocorrelation function ; Its power spectral density is ; In the formula, The unit impact function.

[0057] Therefore ; The output noise signal at the interval is The variance of the two sampled values ​​at time is ; (2) Final composite standard deviation of the signal under test ; Right now ; (3) Setting the mutation threshold like Figure 9 As shown, the false alarm rate for sudden faults is set to be less than Find the threshold for sudden failure. .

[0058] Consult the normal distribution table and select the appropriate quantile. Value (choose the lower bound) such that ; Solving ; The threshold for sudden failure can be set as follows: ; The specific algorithm is as follows: enter:

[0059] : The variance of the noise signal; : Sampling signal interval time; : The normal distribution quantile corresponding to the frequency of fault misjudgment; Output:

[0060]

[0061]

[0062]

[0063]

[0064]

[0065] ; End Source code for the mutation failure threshold algorithm: function pushbutton1_Callback(hObject, eventdata, handles) %%%Read setting value Ts = get(handles.edit_T,'String'); bps = get(handles.edit_bp,'String'); as = get(handles.edit_a,'String'); Tys = get(handles.edit_Ty,'String'); Ty1s = get(handles.edit_Ty1,'String'); % T= str2num(Ts); bp = str2num(bps); a = str2num(as); Ty= str2num(Tys); Ty1= str2num(Ty1s); Kps = get(handles.edit_Kp,'String'); Kis = get(handles.edit_Ki,'String'); Kds = get(handles.edit_Kd,'String'); Kms = get(handles.edit_Km,'String'); % Kp= str2num(Kps); Ki = str2num(Kis); Kd = str2num(Kds); Km = str2num(Kms); % N1s = get(handles.edit_N1,'String'); N2s = get(handles.edit_N2,'String'); N3s = get(handles.edit_N3,'String'); Nys = get(handles.edit_Ny,'String'); % N1= str2num(N1s); N2 = str2num(N2s); N3 = str2num(N3s); Ny = str2num(Nys); % % Calculate the quantile of the normal distribution fib = 1 - a / 2; Norm_fib = 0; set(handles.edit_fib,'String',num2str(fib,18)); Norm_fib = function_invernormcdf(fib); % Inverse of the normal distribution set(handles.edit_b,'String',num2str(Norm_fib,18)); bs = get(handles.edit_b,'String'); b = str2num(bs); Ks = get(handles.edit_K,'String'); K = str2num(Ks); x1 = N1 * Km^2 * Kp^2 * T / (Ty^2); x2 = N2 * bp^2 * Ki^2 * T^2 / ((Km + Ty * bp * Ki) * 2 * Ty); x3 = N3 * (exp(-T * Km / Ty) - exp(-T * Km / Ty1))^2 * T^3 * (1 + 0.5 * Km * T / Ty) * (0.5 + Km * T / (3 * Ty)); x4 = 2 * Ny; sigma = sqrt(x1 + x2 + x3 + x4); set(handles.edit_sigma,'String',num2str(sigma,18)); Bm = b * sigma; ​​​​​​​​​​​​​ else b_low = b_middle; end end fib = b_middle; end This invention discloses a method for calculating the fault threshold of signal abrupt changes and its channel switching control process. The method sequentially executes the steps of input signal type detection, scheme selection, threshold setting, and channel switching control. Input signal type detection uses the unit root DF test to classify noise signals into stationary Gaussian white noise signals and non-stationary noise signals. Then, a corresponding stationary mixed Gaussian noise variance calculation scheme or a non-stationary Brownian noise variance calculation scheme is selected. The variance of the signal to be detected is obtained by constructing a system structure and transfer function model, estimating noise parameters, and calculating the mean square spectral density and autocorrelation function. The abrupt change fault threshold is set by combining the composite variance and the normal distribution quantile, and the false positive rate is controlled. Finally, when the detected signal exceeds the threshold, and no abrupt change fault is found in the preceding input signals, an alarm for the detected signal abrupt change fault is triggered, and the system switches to a backup signal channel. This method overcomes the limitations of existing methods such as upper and lower limit methods, empirical methods, and fixed thresholds. It can cover multiple noise scenarios, reduce the risk of false positives and false negatives, dynamically adapt to system and signal changes, improve fault identification accuracy, enhance the reliability and fault tolerance performance of the hydro-generator control system, and ensure stable unit operation.

Claims

1. A signal mutation fault threshold value calculation method and a path switching control flow thereof, characterized by, The method comprises the following steps: Step 1: input signal type detection; Step 2: scheme selection; Step 3: calculating the variance of the signal to be detected according to different schemes, determining the composite variance and setting the signal mutation fault threshold; Step 4: when the mutation fault switching condition is met according to the mutation threshold obtained in step 3, the signal is disabled and switched to the standby signal channel, and the signal mutation fault alarm and channel switching are completed.

2. The signal abrupt change fault threshold value calculation method and channel switching control flow thereof according to claim 1, characterized in that, In step 1, the input noise signal is divided into two categories, the first category is stationary Gaussian white noise signal, and the second category is non-stationary noise signal; The input signal type detection is specifically as follows: Unit root DF test: Assume that the production model of data is ； Then ； In the formula, : sequence at time of sampling value; : sequence at time of noise value; : The observations are subject to an independent normal distribution with mean 0 and variance 1. : difference of the sequence at adjacent time instants; : parameter to be tested; : parameter to be tested; To test the input signal whether it obeys a random walk, the following model is established ； wherein is an error term; Original hypothesis ; alternative hypothesis ; From the least square estimation, we have ； wherein , is the sample size, then the DF statistic is ； wherein is the standard deviation of the values of the parameter; Given test critical value , if , reject the null hypothesis, the series is stationary; if , accept the null hypothesis, the series is not stationary.

3. The signal abrupt change fault threshold value calculation method and channel switching control flow thereof according to claim 1, wherein, For the mutation case of stationary noise signal, when the input is stationary mixed Gaussian noise signal, the variance of the signal to be detected is calculated as follows: Step 1: establish the system structure and transfer function model; Step 2: Gaussian white noise parameter estimation; Step 3: calculate the mean square spectral density of the stationary process of the signal to be detected; Step 4: calculate the autocorrelation function of the signal to be detected; Step 5: calculate the variance of the signal to be detected under mixed stationary noise.

4. The signal abrupt change fault threshold value calculation method and channel switching control flow thereof according to claim 3, characterized in that, The step 1 is specifically as follows: The system has The road flat noise input signal, corresponding mean square spectral density is respectively; According to the system structure, a system mathematical model is constructed and a transfer function of each input signal branch to the signal to be detected is calculated ; The step 2 is specifically as follows: road noise variance estimation​ ； In the formula, : the road noise signal sampling value at the moment : second mean of the road noise signal; The autocorrelation function is ； In the formula, : impact function; The power spectral density is ； In the formula, : the power spectral density of the road input noise signal; : 1st : variance of the road noise signal; The step 3 is specifically as follows: Calculate the mean square spectral density of each input signal branch after passing through the system to the signal to be detected ； In the formula, : the The mean square spectral density of the input noise signal after passing through the system to the signal to be detected; : : abbreviation for : 1st power spectral density of the input noise signal; : 1st The transfer function of the path inputting the noise signal to the signal to be detected.

5. The signal abrupt change fault threshold value calculation method and channel switching control flow thereof according to claim 4, characterized in that, The step 4 is specifically as follows: Calculate the autocorrelation function of the signal to be detected ； : autocorrelation function of the output signal of the first path noise signal passing through the system to the signal under test : abbreviations; Therefore ； Similarly, Therefore ； : autocorrelation function of the output signal of the second path noise signal passing through the system to the signal under test : abbreviations; The autocorrelation function of the mixed signal superposition is ； In the formula, : autocorrelation function after mixed signal superposition : cross-correlation function of the input noise signals; : In The i-th channel noise signal at time t is outputted after passing through the system to the signal to be detected. : At the moment the ith channel noise signal passes through the system to the signal to be detected and outputs a signal value; The cross-correlation function of the input noise signal is ； In the formula, : transfer function : impact response; : transfer function : impulse response; : 1st input noise signal; : 2nd input noise signal; Similarly, we have ； In the formula, : transfer function : impact response; : transfer function : impulse response; : the ith input noise signal; : jth input noise signal; Therefore, the autocorrelation function of the mixed signal superposition is ； Let ； Then ； ； In the formula, : autocorrelation function after mixed signal superposition : time interval time interval The step 5 is specifically as follows: The variance of the signal to be detected under mixed noise is ； In the formula, : variance value of the signal to be detected under mixed noise; The variance of the two sample values of the signal to be detected at the interval time of is ； From the stationary process, we have ； In the formula, : expected value of the signal to be tested; The variance of the two sample values of the signal to be detected at the interval time of is ； In the formula, : The variance of two sampling values of the signal to be detected at the interval time of under the stationary noise signal.

6. The signal mutation fault threshold calculation method and channel switching control process according to claim 1, characterized in that For the mutation case of non-stationary noise signal, when the input is non-stationary Brown noise signal, the variance of the signal to be detected is calculated as follows: Step 1: construct the system structure and transfer function model; Step 2: Brown noise parameter estimation; Step 3: calculate the system transfer function; Step 4: derive the variance of the signal to be detected based on Brown motion under non-stationary conditions; Step 5: calculate the variance of the signal to be detected under mixed non-stationary noise.

7. The signal abrupt change fault threshold value calculation method and channel switching control flow thereof according to claim 6, wherein, The step 1 is specifically as follows: The system has The corresponding Brown noise intensity of the non-stationary noise input signal is According to the system structure, a mathematical model of the system is constructed, and transfer functions of branches of the input signals to the to-be-detected signal are calculated ； The step 2 is specifically as follows: First Road Brownian noise intensity Estimation ； In the formula, : the road Brownian noise intensity; : first road noise signal sample value at the time instant The step 3 is specifically as follows: From the system transfer function, an equivalent ordinary differential equation and boundary conditions are established; ； In the formula, : system structure parameters; : input signal parameters; The solution of the ordinary differential equation is: ； The transfer function is: ； In the formula, : solution of ordinary differential equation; : transfer function.

8. The signal-mutation fault threshold value calculation method and channel switching control flow thereof according to claim 7, characterized in that, The step 4 is specifically as follows: When the input noise signal has the Brownian motion characteristics, that is, the Brown noise is , the solution of the signal to be detected is ； In the formula, : standard deviation of the i-th input noise signal; : standard Brownian motion; The variance calculation formula of the signal to be detected under input noise signal is ； Integral Ito integral; Substitute the system parameters, so the variance of the difference of the sampling values at different times can be calculated as ; The step 5 is specifically as follows: ； In the formula, : The variance of two sampling values of the signal to be detected at the interval time of when the non-stationary noise signal is 9. The signal abrupt change fault threshold value calculation method and channel switching control flow thereof according to claim 1, wherein, The step 3 of setting the signal mutation fault threshold is specifically as follows: (1) the output end noise variance Output end Gaussian white noise Subjection Normal distribution, autocorrelation function ； In the formula, : autocorrelation function of the output : noise variance value of output end; : impact function; The power spectral density is ； In the formula, Unit impact function; Therefore ； In the formula, : is equivalent to , i.e. the noise variance value of the output end, the notation adopted for the unity of formula expression; : input noise signal; The variance of the two sample values of the output noise signal at the interval time of is ； (2) the final composite standard deviation of the signal to be detected ； That is ； In the formula, : final composite standard deviation of the signal under test; (3) mutation threshold setting Set the mutation fault misjudgment rate to be less than , find the mutation fault threshold ; Consult a normal distribution table to select the appropriate quantile value such that ； In the formula, : false alarm rate of a sudden failure : quantile; : mutation fault threshold : standard normal distribution function; Solving the equation, we have ； The mutation fault threshold setting is 。 10. The signal abrupt change fault threshold value calculation method and channel switching control flow thereof according to claim 1, wherein, The step 4 of signal mutation fault alarm and channel switching is specifically as follows: Step 1: read the input signal sampling value; Step 2: perform stationary test on the input signal, if it is stationary, select the stationary model; otherwise, select the non-stationary model; Step 3: calculate the signal mutation fault threshold; Step 4: judge whether the absolute value of the difference between the current value of the signal to be detected and the sampling value of the signal to be detected in the last scanning period is greater than the signal mutation fault threshold, if false, turn to step 5; otherwise return to step 4; Step 5: judge whether the first input signal has mutation fault, if false, turn to step 6; otherwise return to step 4; Step 6: judge whether the second input signal has mutation fault, if false, turn to step 7; otherwise return to step 4; Step k+4: judge whether the kth input signal has mutation fault, if false, turn to step k+5; otherwise return to step 4; Step k+5: alarm the mutation fault of the signal to be detected, switch to the standby signal.

Images