A method for early warning of sub-working condition measurement of hydroelectric generating unit based on probability function

Abstract

The application discloses a kind of early warning methods of water turbine unit sub-working condition measured value based on probability function, wherein the method includes: collecting the historical time domain waveform vibration data of each component of water turbine unit, and eliminating noise data;Based on historical time domain waveform vibration data, construct and train neural network model, and predict the working condition determination result of water turbine unit based on the trained neural network model;Based on probability function and working condition determination result, calculate the alarm threshold of each component of water turbine unit;Based on alarm threshold, carry out out-of-limit alarm detection and shutdown protection determination to the real-time vibration monitoring value of each component of water turbine unit.The application uses reasonable probability function to determine the calculation method of swing early warning threshold, determines the shutdown protection logic under different working conditions, and guarantees the accuracy of early warning signal output.

Description

Technical Field

[0001] This invention relates to the field of early warning technology for hydropower unit faults, and in particular to an early warning method for hydropower unit operating condition measurements based on probability functions. Background Technology

[0002] Vibration is a common phenomenon in hydropower units. Strong vibrations can affect the normal operation of the unit and reduce the service life of the unit and some components. Therefore, vibration signals are an important indicator for evaluating the operating status of hydropower units and a basis for fault diagnosis. Existing national and industry standards have issued various industry rules and power plant operation procedures to determine the monitoring and early warning thresholds for the units. However, the thresholds for some indicators change with operating conditions, and these are not directly given in the relevant regulations. Furthermore, monitoring values ​​are significantly affected by factors such as the monitoring installation location and different operating conditions of the unit. Therefore, relying solely on industry standards cannot effectively determine the early warning thresholds for various locations within the unit.

[0003] To address this issue, current methods for determining the oscillation warning threshold in hydropower stations primarily rely on historical data thresholds from the generating units themselves and expert experience to set different alarm values ​​or trip values ​​for different operating conditions. Experts determine the alarm setpoint coefficients for each operating condition based on experience, and then multiply them by the peak-to-peak values ​​of the hydropower unit's historical data to obtain the graded alarm thresholds for each condition. Currently, to better consider the actual conditions of the generating units and the massive amount of accumulated historical data, some scholars have proposed using probability function methods suitable for normal distributions, such as the 3σ criterion and Gaussian distribution, to calculate the unit monitoring warning values ​​by analyzing the distribution of the unit monitoring samples. However, because hydropower unit monitoring data exhibits a skewed distribution rather than a perfectly normal distribution, the warning values ​​output by existing schemes are not always accurate. Summary of the Invention

[0004] In view of this, the purpose of this invention is to propose an early warning method for hydropower unit operating condition measurements based on probability functions. This method can use reasonable probability functions to clarify the calculation method of the oscillation early warning threshold, determine the shutdown protection logic under different operating conditions, and ensure the accuracy of the early warning signal output.

[0005] According to one aspect of the present invention, an early warning method for hydroelectric generator component operating condition measurements based on probability functions is provided, comprising:

[0006] Historical time-domain waveform vibration data of each component of the hydropower unit were collected, and noise data was removed.

[0007] A neural network model is constructed and trained based on historical time-domain waveform vibration data, and the operating condition judgment result of the hydropower unit is predicted based on the trained neural network model.

[0008] The alarm thresholds for each component of the hydropower unit are calculated based on the probability function and the operating condition determination results.

[0009] Based on alarm thresholds, the real-time vibration monitoring values ​​of each component of the hydropower unit are used to perform over-limit alarm detection and shutdown protection determination.

[0010] In the aforementioned technical solution, historical data is used to construct and extrapolate the operating conditions of hydropower units. This is because the operating conditions of hydropower units are quite complex. Therefore, this case does not completely abandon the traditional method of extrapolating the oscillation warning threshold based on expert experience and historical data. Instead, this case first utilizes the traditional method of extrapolating the oscillation warning threshold based on expert experience and historical data to construct a portion of the analysis work, which allows for faster completion of subsequent work and facilitates the adaptation of different hydropower unit solutions. Furthermore, after obtaining the operating condition results using the traditional method, the concept of a probability function is introduced. This probability function is used to calculate the alarm threshold, which is then used for over-limit alarm detection and shutdown protection determination. The advantage of using a probability function in this case is that, since the monitoring data of hydropower units has statistical characteristics that follow a skewed distribution rather than a perfectly normal distribution, the Pearson Type III frequency curve, as a probability function suitable for skewed distributions, can better fit the monitoring signals and improve the accuracy and reliability of environmental assessments. Therefore, this case uses the Pearson Type III frequency curve to clarify the calculation method of the oscillation warning threshold, determine the shutdown protection logic under different operating conditions, and ensure the accuracy of the warning signal output.

[0011] In some embodiments, historical time-domain waveform vibration data of various components of the hydropower unit are collected, specifically:

[0012] The radial time-domain waveform vibration data of the main shaft was measured using an eddy current sensor.

[0013] Low-frequency velocity sensors were used to measure the time-domain waveform vibration data of each supporting component of the turbine unit.

[0014] The purpose of this setup in the above technical solution is to collect vibration data more accurately by setting different vibration monitoring sensors according to different locations.

[0015] In some embodiments, noisy data is removed, specifically:

[0016] Calculate the mean and standard deviation of each historical time-domain waveform vibration data segment, and use the Raida criterion to extract the noise data it contains.

[0017] The reason for choosing the Raida criterion in the above technical solution is that it is simple to use, and the time-domain waveform data is approximately normally distributed, allowing for the rapid removal of outliers. Although the Raida criterion requires a sufficient amount of measurement data to remove outliers, otherwise the error will be large, most hydropower units have a large amount of historical data, so this situation will not occur.

[0018] In some embodiments, a neural network model is constructed and trained based on historical time-domain waveform vibration data, and the operating condition judgment result of the hydropower unit is predicted based on the trained neural network model. Specifically:

[0019] The time-domain waveform data is divided into equal-length segments and then into multiple groups according to the rotation period. The time-domain characteristic value of each group is calculated.

[0020] The calculated time-domain feature values ​​are used as the training and test sets, and a neural network model is constructed and trained. The trained neural network model is then used to predict the operating condition of the hydropower unit.

[0021] In the above technical solution, since waveform data has many time-domain feature values, such as maximum value, maximum absolute value, minimum value, average value, peak-to-peak value, absolute average value, root mean square value, root mean square amplitude, standard deviation, kurtosis, skewness, margin, waveform, impulse, and peak value, a neural network model is adopted. Leveraging its strong information synthesis capabilities, it can simultaneously handle quantitative and qualitative aspects and effectively coordinate the relationships between different input feature values. Compared to traditional machine learning solutions, neural networks offer better self-learning, self-organization, and adaptability.

[0022] In some embodiments, alarm thresholds for each component of the hydropower unit are calculated based on a probability function and the operating condition determination results. Specifically:

[0023] Based on the operating condition determination results, the time-domain waveform vibration data of the hydropower unit components were selected and arranged in descending order of amplitude value, and the modulus ratio and dispersion coefficient of each component of the hydropower unit were calculated.

[0024] Set the deviation coefficients for each component of the hydropower unit, plot the theoretical frequency curves of Pearson-III type, fit the corresponding theoretical frequency curves according to the curves, and calculate the empirical cumulative probability and modulus coefficient of each component of the hydropower unit.

[0025] The maximum amplitude value of each component of the hydropower unit is calculated based on the modulus ratio coefficient, and the alarm threshold of each component of the hydropower unit is set according to the maximum amplitude value.

[0026] The purpose of this setup in the above technical solution is as follows: The advantage of using a probability function in this case is that, since the monitoring data of hydropower units exhibits statistical characteristics that follow a skewed distribution rather than a perfectly normal distribution, the Pearson Type III frequency curve, as a probability function suitable for skewed distributions, can better fit the monitoring signals and improve the accuracy and reliability of environmental assessments. Therefore, this case utilizes the Pearson Type III frequency curve to clarify the calculation method for the oscillation warning threshold, determine the shutdown protection logic under different operating conditions, and ensure the accuracy of the warning signal output.

[0027] In some embodiments, over-limit alarm detection and shutdown protection determination are performed on the real-time vibration monitoring values ​​of each component of the hydropower unit based on alarm thresholds. Specifically:

[0028] i. Over-limit alarm detection

[0029] If the real-time vibration monitoring values ​​of each component of the hydropower unit meet the over-limit alarm conditions, the corresponding level of alarm signal will be released. The alarm signal includes a first-level alarm signal and a second-level alarm signal.

[0030] The criteria for issuing a Level 1 alarm signal are: the Level 2 alarm value for the operating condition is greater than or equal to the real-time vibration monitoring value for the operating condition, which is greater than or equal to the Level 1 alarm value for the operating condition.

[0031] The criteria for issuing a Level 2 alarm signal are: the real-time vibration monitoring value under this working condition is ≥ the Level 2 alarm value under this working condition;

[0032] ii. Shutdown Protection Judgment

[0033] When a hydropower unit releases two or more level-two alarm signals, if the signals are located at two monitoring points of the same hydropower unit component, the hydropower unit will be shut down for protection. If the signals are located at monitoring points of different components, the unit will not be shut down for protection.

[0034] In the above technical solution, the purpose of this setup is to use the maximum amplitude value obtained from the probability function to set the primary and secondary alarm thresholds for each component of the hydropower unit. By using the primary and secondary alarm thresholds as the judgment criteria, the probability function yields the maximum amplitude value under different cumulative probabilities, which has the advantage of more fully considering historical data. Based on this, using empirical cumulative probabilities of 0.1 and 0.01 as indicators to set the primary and secondary alarm thresholds establishes a reasonable shutdown logic, helping to ensure the accuracy of the early warning signal output.

[0035] According to another aspect of the present invention, an early warning device for measuring the operating conditions of a hydropower unit based on a probability function is proposed. Based on the aforementioned early warning method for measuring the operating conditions of a hydropower unit based on a probability function, the device comprises a data acquisition module, a neural network module, a probability function module, and a judgment module, which are connected in sequence.

[0036] The acquisition module is used to acquire historical time-domain waveform vibration data of each component of the hydropower unit and remove noise data.

[0037] The neural network module is used to construct and train a neural network model based on historical time-domain waveform vibration data, and to predict the operating condition judgment result of the hydropower unit based on the trained neural network model.

[0038] The probability function module is used to calculate the alarm thresholds of each component of the hydropower unit based on the probability function and the operating condition judgment results.

[0039] The determination module is used to perform over-limit alarm detection and shutdown protection determination on the real-time vibration monitoring values ​​of each component of the hydropower unit based on alarm thresholds.

[0040] In the above technical solution, to better utilize the method, different steps are sequentially built into different modules, and these modules are linked together, enabling more efficient training and use of the method. It should be noted that the principles and effects of each step have already been described above and will not be elaborated upon here.

[0041] According to another aspect of the present invention, an early warning device for measuring the operating conditions of a hydroelectric generator based on a probability function is provided, comprising:

[0042] At least one processor; and,

[0043] A memory communicatively connected to the at least one processor; wherein,

[0044] The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform the aforementioned early warning method for measuring the operating conditions of hydropower components based on probability functions.

[0045] In the above technical solution, to better operate and process the method, the method is stored in memory, and the processor executes the stored method. It should be noted that the principle and effect of each step have been described above and will not be elaborated upon here.

[0046] According to another aspect of the present invention, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the above-described early warning method for hydroelectric component operating condition measurements based on a probability function.

[0047] In the above technical solution, to better operate and use the method, the method is stored in a computer-readable storage medium and implemented using a processor. It should be noted that the principle and effect of each step have been described above and will not be elaborated upon here. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a flowchart illustrating one embodiment of the early warning method for hydroelectric generator component operating condition measurement based on probability functions according to the present invention.

[0050] Figure 2 This is a flowchart illustrating the implementation of a specific embodiment of the early warning method for hydroelectric components based on probability functions according to the present invention.

[0051] Figure 3 This is a specific embodiment of the early warning method for hydroelectric generator component operating condition measurement based on probability function of the present invention, showing the time-domain vibration waveforms of the stator core horizontal and vertical vibrations;

[0052] Figure 4 This is a flowchart illustrating the calculation of the oscillation warning threshold in a specific embodiment of the early warning method for hydroelectric generator components based on probability functions according to the present invention. Detailed Implementation

[0053] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be particularly noted that the following embodiments are for illustrative purposes only and do not limit the scope of the invention. Similarly, the following embodiments are only some, not all, embodiments of the present invention, and all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] This invention provides an early warning method for hydropower unit component operating condition measurements based on probability functions. It can use reasonable probability functions to clarify the calculation method of the oscillation early warning threshold, determine the shutdown protection logic under different operating conditions, and ensure the accuracy of the early warning signal output.

[0055] Please see Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the early warning method for hydroelectric generator component operating condition measurements based on probability functions according to the present invention. It should be noted that if substantially the same result is obtained, the method of the present invention is not necessarily identical. Figure 1 The illustrated process sequence is limited. For example... Figure 1 As shown, the method includes the following steps:

[0056] S101. Collect historical time-domain waveform vibration data of each component of the hydropower unit and remove noise data;

[0057] In this embodiment, historical time-domain waveform vibration data of each component of the hydropower unit are collected. Specifically, an eddy current sensor is used to measure the radial time-domain waveform vibration data of the main shaft; and a low-frequency velocity sensor is used to measure the time-domain waveform vibration data of each supporting component of the hydropower unit.

[0058] In this embodiment, setting different vibration monitoring sensors according to different locations can collect vibration data more accurately.

[0059] In this embodiment, it should be understood that, in order to facilitate data processing, the data type of the collected data is first located as waveform data.

[0060] In this embodiment, noisy data is removed, specifically:

[0061] Calculate the mean and standard deviation of each historical time-domain waveform vibration data segment, and use the Raida criterion to extract the noise data it contains.

[0062] In this embodiment, the Raida criterion is chosen because it is simple to use, and the time-domain waveform data is approximately normally distributed, allowing for the rapid removal of outliers. Although the Raida criterion requires a sufficient amount of measurement data to remove outliers, otherwise the error will be large, most hydroelectric generating units have a large amount of historical data, so this situation will not occur.

[0063] In this embodiment, it should be understood that this embodiment is only used to illustrate the distance based on the Raida criterion. This is a preferred solution adopted to closely match the processing data of this case and facilitate subsequent processing. Other elimination schemes may also be adopted in a targeted manner.

[0064] S102. Construct and train a neural network model based on historical time-domain waveform vibration data, and predict the operating condition judgment result of the hydropower unit based on the trained neural network model.

[0065] In this embodiment, a neural network model is constructed and trained based on historical time-domain waveform vibration data, and the operating condition judgment result of the hydropower unit is predicted based on the trained neural network model. Specifically:

[0066] The time-domain waveform data is divided into equal-length segments and then into multiple groups according to the rotation period. The time-domain characteristic value of each group is calculated.

[0067] The calculated time-domain feature values ​​are used as the training and test sets, and a neural network model is constructed and trained. The trained neural network model is then used to predict the operating condition of the hydropower unit.

[0068] In this embodiment, since waveform data has many time-domain feature values, such as maximum value, maximum absolute value, minimum value, average value, peak-to-peak value, absolute average value, root mean square value, root mean square amplitude, standard deviation, kurtosis, skewness, margin, waveform, impulse, and peak value, a neural network model is adopted. Leveraging its strong information synthesis capabilities, it can simultaneously handle quantitative and qualitative aspects, effectively coordinating the relationships between different input feature values. Compared to traditional machine learning schemes, neural networks offer better self-learning, self-organization, and adaptability.

[0069] In this embodiment, it should be noted that this case only limits the use of a neural network; other methods can also be used to complete this embodiment without limitation. It should be understood that the model used in these steps needs to meet the following two points:

[0070] 1. Big Data Processing

[0071] 2. Coordination and rapid processing of multiple feature inputs.

[0072] S103. Calculate the alarm thresholds of each component of the hydropower unit based on the probability function and the operating condition judgment results.

[0073] In this embodiment, the alarm thresholds for each component of the hydropower unit are calculated based on the probability function and the operating condition determination results. Specifically:

[0074] Based on the operating condition determination results, the time-domain waveform vibration data of the hydropower unit components were selected and arranged in descending order of amplitude value, and the modulus ratio and dispersion coefficient of each component of the hydropower unit were calculated.

[0075] Set the deviation coefficients for each component of the hydropower unit, plot the theoretical frequency curves of Pearson-III type, fit the corresponding theoretical frequency curves according to the curves, and calculate the empirical cumulative probability and modulus coefficient of each component of the hydropower unit.

[0076] The maximum amplitude value of each component of the hydropower unit is calculated based on the modulus ratio coefficient, and the alarm threshold of each component of the hydropower unit is set according to the maximum amplitude value.

[0077] In this embodiment, the advantage of using a probability function is that, since the monitoring data of hydropower units exhibits a skewed distribution rather than a perfectly normal distribution, the Pearson Type III frequency curve, as a probability function suitable for skewed distributions, can better fit the monitoring signals and improve the accuracy and reliability of environmental assessments. Therefore, this embodiment uses the Pearson Type III frequency curve to clarify the calculation method for the oscillation warning threshold, determine the shutdown protection logic under different operating conditions, and ensure the accuracy of the warning signal output.

[0078] S104. Based on alarm thresholds, perform over-limit alarm detection and shutdown protection determination on the real-time vibration monitoring values ​​of each component of the hydropower unit.

[0079] In this embodiment, based on alarm thresholds, the real-time vibration monitoring values ​​of each component of the hydropower unit are used for over-limit alarm detection and shutdown protection determination. Specifically:

[0080] i. Over-limit alarm detection

[0081] If the real-time vibration monitoring values ​​of each component of the hydropower unit meet the over-limit alarm conditions, the corresponding level of alarm signal will be released. The alarm signal includes a first-level alarm signal and a second-level alarm signal.

[0082] The criteria for issuing a Level 1 alarm signal are: the Level 2 alarm value for the operating condition is greater than or equal to the real-time vibration monitoring value for the operating condition, which is greater than or equal to the Level 1 alarm value for the operating condition.

[0083] The criteria for issuing a Level 2 alarm signal are: the real-time vibration monitoring value under this working condition is ≥ the Level 2 alarm value under this working condition;

[0084] ii. Shutdown Protection Judgment

[0085] When a hydropower unit releases two or more level-two alarm signals, if the signals are located at two monitoring points of the same hydropower unit component, the hydropower unit will be shut down for protection. If the signals are located at monitoring points of different components, the unit will not be shut down for protection.

[0086] In this embodiment, the purpose of this setup is to use the maximum amplitude value obtained from the probability function to set the primary and secondary alarm thresholds for each component of the hydropower unit. By using the primary and secondary alarm thresholds as the judgment criteria, the probability function yields the maximum amplitude value under different cumulative probabilities, which has the advantage of more fully considering historical data. Based on this, using empirical cumulative probabilities of 0.1 and 0.01 as indicators to set the primary and secondary alarm thresholds establishes a reasonable shutdown logic, which helps ensure the accuracy of the early warning signal output.

[0087] Based on the method described in one of the embodiments, please refer to Figure 2 , Figure 3 as well as Figure 4 The specific methods and steps for predicting and warning the division of labor of a hydroelectric power unit are as follows:

[0088] Step 1: Install vibration sensors at various components of the hydroelectric generator to collect and upload vibration data from the monitoring points. The collected vibration data is time-domain waveform data, such as... Figure 3 As shown.

[0089] The present invention is further configured such that the oscillation data of the monitoring points in step 1) are respectively: upper guide XY oscillation, lower guide XY oscillation, and water guide XY oscillation.

[0090] The XYZ sway of the upper frame, the XY sway of the stator frame, the XYZ sway of the lower frame, the XYZ sway of the top cover, the horizontal and vertical vibration of the stator core, and the key phase signal.

[0091] Among them, the vibration sensor type for measuring the radial vibration of the main shaft (upper guide, lower guide, water guide) is an eddy current sensor, and the sensor type for measuring the vibration of each supporting component of the turbine unit (upper frame, lower frame, top cover) is a low-frequency velocity sensor.

[0092] The present invention is further configured such that: the time-domain waveform data in step 1) refers to the waveform of the vibration signal changing over time, representing the vibration and oscillation data of the hydropower unit under various operating conditions, such as... Figure 2 As shown.

[0093] Step 2: Calculate the average value and standard deviation of each waveform data segment, and remove noise data from the monitoring points.

[0094] The present invention is further configured such that: in step 2), the average value and standard deviation of each waveform data segment are calculated, and noise data from the monitoring points are removed; specifically,

[0095] 2-1) Starting with the bond phase signal, select data containing n rotation cycles as a calculation sample (n≥2).

[0096] 2-2) Calculate the mean μ and standard deviation σ of the sample, perform confidence analysis, and remove noise data that exceeds the range [μ-3σ, μ+3σ] in the waveform sample data.

[0097] 2-3) In the next calculation interval, shift right by 1 rotation cycle to obtain the next data sample containing n rotation cycles. Repeat step 2-2) to finally obtain the time-domain waveform data after removing data noise.

[0098] Step 3: Determine operating conditions.

[0099] The present invention is further configured such that: the operating condition determination in step 3) includes pumping start-up condition determination, power generation start-up condition determination, pumping phase adjustment condition determination, shutdown condition determination, and steady-state condition determination. The current operating condition status of the unit can be obtained through a one-dimensional convolutional neural network method, specifically as pumping start-up condition, power generation start-up condition, pumping phase adjustment condition, shutdown condition, and steady-state condition.

[0100] 3-1) Divide the time-domain waveform data obtained in step two into equal-length segments, grouping them according to the rotation period. Calculate the time-domain feature values ​​for each group, which will serve as input to the convolutional neural network model. The time-domain feature values ​​include: maximum value, maximum absolute value, minimum value, average value, peak-to-peak value, absolute average value, root mean square value, root mean square amplitude, standard deviation, kurtosis, skewness, margin, waveform, impulse, and peak value.

[0101] 3-2) Label the time-domain feature values ​​with the corresponding working condition categories and save them in CSV format as the experimental dataset.

[0102] 3-3) The experimental dataset was divided into a training set and a test set, with a ratio of 7:3.

[0103] 3-4) Construct a one-dimensional convolutional neural network model. The network model structure consists of 7 layers, specifically:

[0104] C1 & C2: One-dimensional convolutional layers with a kernel length of 64 and the ReLU activation function;

[0105] C3: Maximum pooling layer with a unit length of 2;

[0106] C4 & C5: One-dimensional convolutional layers with a kernel length of 256 and the ReLU activation function;

[0107] C6: To prevent overfitting, a Dropout layer with a retention probability of 0.5 is set.

[0108] C7: Fully connected layer, the output layer uses the Softmax activation function.

[0109] 3-5) Input the training set into the model for training until the model converges and the performance meets the requirements, and then obtain the working condition judgment model based on a one-dimensional convolutional neural network.

[0110] 3-6) Collect the vibration and sway data of the hydropower unit in real time, input it into the working condition judgment model trained in step 3-5), and output the working condition judgment result.

[0111] Step 4: Use probability functions to calculate the first and second level alarm thresholds for each component of the hydropower unit.

[0112] The present invention is further configured such that: in step 4), a probability function is used to calculate the primary and secondary alarm thresholds for each component of the hydropower unit, such as... Figure 3 The diagram shows how to use a Pearson-III type probability function to plot the frequency curve of the amplitude under this operating condition, and determine the first and second level alarm thresholds for each component of the hydropower unit's oscillation system. Specifically:

[0113] 4-1) Under the judgment operating conditions, the vibration and oscillation data of the hydropower unit components are selected and arranged in descending order of amplitude value.

[0114] 4-2) Calculate the average amplitude value from the pendulum data. Modulus coefficient K i The coefficient of variation C v The calculation formula is:

[0115]

[0116]

[0117] In formula ①, x i The amplitude data is arranged in descending order for the i-th index. This represents the average amplitude value of the amplitude data;

[0118] In formula ②, K i is the modulus ratio coefficient of the i-th amplitude data, and n is the total number of oscillation data.

[0119] 4-3) Set the deviation coefficient C s (0 <C s ≤3), respectively plot the Pearson-III type theoretical frequency curves, and analyze the theoretical frequency curve with the best fitting effect based on the least squares method. Then, the skewness coefficient C corresponding to this curve is... s As an estimate of the population parameter.

[0120] 4-4) Calculate the empirical cumulative probability f of the amplitude data. The calculation formula is:

[0121]

[0122] In formula ③, m is the m-th amplitude data arranged in descending order, and n is the total number of oscillation data.

[0123] 4-5) Based on the deviation coefficient C s Consult the table of deviation coefficients for Pearson-III type curves to find the deviation coefficients when f is 0.1 and 0.01. Calculate the modulus coefficient K for different f values. i The calculation formula is:

[0124]

[0125] 4-6) Based on the modulus ratio coefficient K in step 4-5) i Calculate the maximum amplitude value P corresponding to f being 0.1 and 0.01. i These are set as Level 1 and Level 2 alarm thresholds, used as the criteria for detecting over-limit alarms. The calculation formula is:

[0126]

[0127] 4-7) Select the horizontal or vibration signals of each component of the hydropower unit, and repeat steps 4-1) to 4-6) to calculate the first-level and second-level alarm thresholds of each component of the hydropower unit under different operating conditions.

[0128] Step 5: Perform over-limit alarm detection and shutdown protection determination.

[0129] The present invention is further configured such that: in step 5), the over-limit alarm detection means that if the monitored value of the oscillating system meets the over-limit alarm conditions, then an alarm signal of the corresponding level is released. The determination condition for issuing a level one alarm signal is: the level two alarm value of the operating condition ≥ the monitored value of the oscillating system under the operating condition ≥ the level one alarm value of the operating condition; the determination condition for issuing a level two alarm signal is: the monitored value of the oscillating system under the operating condition ≥ the level two alarm value of the operating condition.

[0130] The shutdown protection determination in step 5) refers to the following: when the unit releases two or more level-two alarm signals, if the signals are located at two monitoring points of the same component, the shutdown protection logic is satisfied, the shutdown protection signal is activated, and the unit mechanically trips. If the signals are located at monitoring points of different components, the shutdown protection logic determination is not satisfied.

[0131] Example 2

[0132] An early warning device for measuring the operating conditions of a hydropower unit based on a probability function, and an early warning method for measuring the operating conditions of a hydropower unit based on a probability function as described in one embodiment, comprising a data acquisition module, a neural network module, a probability function module, and a judgment module connected in sequence:

[0133] The acquisition module is used to acquire historical time-domain waveform vibration data of each component of the hydropower unit and remove noise data.

[0134] The neural network module is used to construct and train a neural network model based on historical time-domain waveform vibration data, and to predict the operating condition judgment result of the hydropower unit based on the trained neural network model.

[0135] The probability function module is used to calculate the alarm thresholds of each component of the hydropower unit based on the probability function and the operating condition judgment results.

[0136] The determination module is used to perform over-limit alarm detection and shutdown protection determination on the real-time vibration monitoring values ​​of each component of the hydropower unit based on alarm thresholds.

[0137] In this embodiment, to better utilize the method described in one of the embodiments, different modules are sequentially established for different steps, and these modules are connected in series, enabling more efficient training and use of the method. It should be noted that the principle and effect of each step have been described above and will not be elaborated upon here.

[0138] Example 3

[0139] An early warning device for measuring the operating conditions of hydroelectric generator components based on probability functions includes:

[0140] At least one processor; and,

[0141] A memory communicatively connected to the at least one processor; wherein,

[0142] The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform the early warning method for measuring the operating conditions of hydropower components based on a probability function, as described in one embodiment.

[0143] In this embodiment, to better run and process the method, the above method is stored in memory, and the stored method is executed using a processor. It should be noted that the principle and effect of each step have been described above and will not be elaborated upon here.

[0144] Example 4

[0145] A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned early warning method for hydropower unit operating condition measurements based on a probability function.

[0146] In this embodiment, to better operate and use the method described in one embodiment, the method is stored in a computer-readable storage medium and implemented using a processor. It should be noted that the principle and effect of each step have been described above and will not be elaborated further here.

[0147] The above description is only a part of the embodiments of the present invention and does not limit the scope of protection of the present invention. Any equivalent device or equivalent process transformation made based on the content of the present invention specification and drawings, or direct or indirect application in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for early warning of hydropower unit operating condition measurements based on probability functions, characterized in that, include: Historical time-domain waveform vibration data of each component of the hydropower unit were collected, and noise data was removed. A neural network model is constructed and trained based on historical time-domain waveform vibration data, and the operating condition judgment result of the hydropower unit is predicted based on the trained neural network model. The alarm thresholds for each component of the hydropower unit are calculated based on the probability function and the operating condition determination results. Based on alarm thresholds, the real-time vibration monitoring values ​​of each component of the hydropower unit are used to perform over-limit alarm detection and shutdown protection determination.

2. The early warning method for hydropower unit operating condition measurements based on probability functions as described in claim 1, characterized in that, Collect historical time-domain waveform vibration data of various components of the hydropower unit, specifically: The radial time-domain waveform vibration data of the main shaft was measured using an eddy current sensor. Low-frequency velocity sensors were used to measure the time-domain waveform vibration data of each supporting component of the turbine unit.

3. The early warning method for hydropower unit component operating condition measurement based on probability function as described in claim 1, characterized in that, To remove noisy data, specifically: Calculate the mean and standard deviation of each historical time-domain waveform vibration data segment, and use the Raida criterion to remove the noise data it contains.

4. The early warning method for hydropower unit operating condition measurement based on probability function as described in claim 1, characterized in that, A neural network model is constructed and trained based on historical time-domain waveform vibration data. The trained neural network model is then used to predict the operating conditions of the hydropower unit. Specifically: The time-domain waveform data is divided into equal-length segments and then into multiple groups according to the rotation period. The time-domain characteristic value of each group is calculated. The calculated time-domain feature values ​​are used as the training and test sets, and a neural network model is constructed and trained. The trained neural network model is then used to predict the operating condition of the hydropower unit.

5. The early warning method for hydropower unit component operating condition measurement based on probability function as described in claim 1, characterized in that, The alarm thresholds for each component of the hydropower unit are calculated based on probability functions and operating condition determination results. Specifically: Based on the operating condition determination results, the time-domain waveform vibration data of the hydropower unit components were selected and arranged in descending order of amplitude value, and the modulus ratio and dispersion coefficient of each component of the hydropower unit were calculated. Set the deviation coefficients for each component of the hydropower unit, plot the theoretical frequency curves of Pearson-III type, fit the corresponding theoretical frequency curves according to the curves, and calculate the empirical cumulative probability and modulus coefficient of each component of the hydropower unit. The maximum amplitude value of each component of the hydropower unit is calculated based on the modulus ratio coefficient, and the alarm threshold of each component of the hydropower unit is set according to the maximum amplitude value.

6. The early warning method for hydropower unit component operating condition measurement based on probability function as described in claim 5, characterized in that, Based on alarm thresholds, the real-time vibration monitoring values ​​of various components of the hydropower unit are used to perform over-limit alarm detection and shutdown protection determination. Specifically: i. Over-limit alarm detection If the real-time vibration monitoring values ​​of each component of the hydropower unit meet the over-limit alarm conditions, the corresponding level of alarm signal will be released. The alarm signal includes a first-level alarm signal and a second-level alarm signal. The criteria for issuing a Level 1 alarm signal are: the Level 2 alarm value for the operating condition is greater than or equal to the real-time vibration monitoring value for the operating condition, which is greater than or equal to the Level 1 alarm value for the operating condition. The criteria for issuing a Level 2 alarm signal are: the real-time vibration monitoring value under this working condition is ≥ the Level 2 alarm value under this working condition; ii. Shutdown Protection Judgment When a hydropower unit releases two or more level-two alarm signals, if the signals are located at two monitoring points of the same hydropower unit component, the hydropower unit will be shut down for protection. If the signals are located at monitoring points of different components, the unit will not be shut down for protection.

7. An early warning device for measuring the operating conditions of hydroelectric generator components based on probability functions, characterized in that, A method for early warning of hydropower unit operating condition measurements based on a probability function, according to any one of claims 1-6, comprises a data acquisition module, a neural network module, a probability function module, and a judgment module connected in sequence: The acquisition module is used to acquire historical time-domain waveform vibration data of each component of the hydropower unit and remove noise data. The neural network module is used to construct and train a neural network model based on historical time-domain waveform vibration data, and to predict the operating condition judgment result of the hydropower unit based on the trained neural network model. The probability function module is used to calculate the alarm thresholds of each component of the hydropower unit based on the probability function and the operating condition judgment results. The determination module is used to perform over-limit alarm detection and shutdown protection determination on the real-time vibration monitoring values ​​of each component of the hydropower unit based on the alarm threshold.

8. An early warning device for measuring the operating conditions of hydroelectric generator components based on probability functions, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform an early warning method for measuring the operating conditions of hydropower components based on a probability function as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the early warning method for measuring the operating conditions of hydropower units based on probability functions, as described in any one of claims 1 to 6.

Images