Water turbine top cover bolt fault alarm system and alarm method for pumped storage power station

Abstract

The invention discloses a pumped storage power station water turbine top cover bolt fault alarm system and an alarm method, relates to the technical field of fault monitoring, and is used for monitoring stress and temperature data of top cover bolts through grating stress sensors in the top cover bolts. The method comprises the following steps: performing feature extraction on stress and temperature data to obtain the feature quantity of each bolt, and storing the feature quantity of each bolt to a current cache region Ai of a feature quantity cache region; the working condition of each top cover bolt and the working condition of the unit in the cache region A (i-1) are judged according to the bolt characteristic quantity data of the current cache region Ai and the adjacent cache region A (i-1) and cache region A (i-2), the working state of each unit is judged according to the bolt characteristic quantity data and the working condition data in the cache region A (i-1), and an alarm signal is sent out. And meanwhile, a threshold prediction model is established according to the convolutional neural network to output the optimized latest alarm threshold and send the optimized latest alarm threshold to an early warning rule base.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault monitoring, in particular to a pumped storage power station water turbine top cover bolt fault alarm system and method. BACKGROUND

[0002] The water turbine of a pumped storage power station is one of the core devices of this type of power station, mainly used for regulating the load of the power grid, playing a role in balancing power demand. A pumped storage power station usually includes two reservoirs: an upper reservoir and a lower reservoir, and energy storage and release are achieved through the bidirectional operation of the water turbine and the generator set;

[0003] In the prior art, the top cover bolt of the water turbine of the pumped storage power station fixes the top cover of the unit, but it has been reported that there have been safety accidents of water flooding the plant due to the fracture of the top cover bolt. The working conditions of the pumped storage unit are numerous, and the stress on the bolt is complex. In the prior art, the determination of the state of the bolt is usually based on the stable working state of the unit, or without considering the working condition, and is determined based on the maximum stress of the bolt. It is difficult to realize real-time monitoring of the top cover bolt, evaluate the safety state of the bolt, and ensure the stable and safe operation of the power station.

[0004] In view of the above technical defects, the present application provides a solution. SUMMARY

[0005] The purpose of the present application is to monitor the state of the top cover bolt of the water turbine in real time, avoid water turbine failure, and adapt to the performance difference caused by aging and environmental factors through an updated alarm threshold.

[0006] In order to achieve the above purpose, the present application adopts the following technical scheme: a pumped storage power station water turbine top cover bolt fault alarm system, comprising a data acquisition module, a data processing module, a warning module and a threshold optimization module, wherein:

[0007] The data acquisition module is used to monitor the stress and temperature data of the top cover bolt through the grating stress sensor inside each top cover bolt and send them to the data processing module.

[0008] The data processing module includes a bolt characteristic quantity calculation unit and a bolt and unit working condition calculation unit, wherein:

[0009] The bolt characteristic quantity calculation unit obtains data by accessing the double buffer area of the data acquisition module, obtains each bolt characteristic quantity by feature extraction on stress and temperature data, and saves each bolt characteristic quantity to the current buffer area Ai of the characteristic quantity buffer area. The bolt characteristic quantity includes dynamic pre-tightening force, static pre-tightening force and dynamic stress main frequency ratio. The characteristic quantity buffer area adopts a multi-buffer circular structure, and each characteristic quantity buffer area includes a bolt characteristic quantity data buffer area, a bolt working condition buffer area and a unit working condition buffer area.

[0010] The bolt and unit working condition calculation unit is used to complete the bolt characteristic quantity calculation of the three cache areas, and the working conditions of each head cover bolt in the adjacent cache area A(i-1) are determined through the bolt characteristic quantity data of the current cache area Ai and the cache areas A(i-1) and A(i-2), and the determination results of the working conditions of each bolt are saved to the cache area A(i-1), and the cache area A(i-1) data is saved to the characteristic quantity database;

[0011] The early warning module is used to acquire the bolt characteristic quantity and the determination results of the bolt working conditions in the cache area A(i-1) and process the determination results, call the early warning rule library, determine the unit working condition through the bolt working condition determination results, and issue an alarm signal, and alarm and remind through the sound and light alarm and the remote monitoring end;

[0012] The threshold optimization module is used to establish a threshold prediction model according to the convolutional neural network, train the model by acquiring the historical operation data of the water turbine from the characteristic quantity database, output the latest optimized alarm threshold, and send the latest optimized alarm threshold to the early warning rule library.

[0013] Further, the data acquisition module is a single thread in the program, adopts a double-buffer structure, realizes continuous data acquisition, and realizes data sharing through the double-buffer and the data processing module.

[0014] Further, the specific process of extracting the characteristic quantity of the head cover bolt is as follows:

[0015] S101, acquire cache data, and the cache data includes strain wavelength data and temperature wavelength data;

[0016] S102, calculate the strain data ε according to the following formula: ε = ((λ-λ0)-(λ T -λ T0 )*K T ) / K,

[0017] Wherein: λ is the test wavelength, λ0 is the zero wavelength, λ T is the temperature compensation wavelength, λ T0 is the temperature compensation zero wavelength, k is the sensitivity coefficient, and k T is the temperature compensation coefficient;

[0018] S102, calculate the bolt stress data according to the strain data, and based on the correlation between stress and strain through the elastic modulus of the material itself, calculate the stress according to the following formula: σ = E*ε, wherein σ is the stress, ε is the strain, and E is the elastic modulus of the material itself;

[0019] S103, performing FFT calculation on each bolt data array, obtaining each bolt frequency domain data, and then obtaining characteristic quantity data such as main frequency and secondary main frequency, performing mean value calculation, maximum value calculation and minimum value calculation on each bolt data one by one, obtaining maximum value characteristic quantity, minimum value characteristic quantity and average value characteristic quantity, and storing them in the characteristic quantity buffer area, i.e. the current buffer area Ai.

[0020] Further, the bolt working condition and the unit working condition both include the following working condition states:

[0021] Specifically, the working condition states include shutdown working condition, power generation start working condition, power generation working condition, power generation shutdown transient working condition, pumping start working condition, pumping and phase modulation working condition, pumping working condition and pumping shutdown transient working condition.

[0022] Further, the evaluation rules of each working condition are as follows:

[0023] A) For the pumping and phase modulation stable working condition, the static pretightening force F0 in the current buffer area Ai, the static pretightening force F1 in the buffer area A(i-1) and the static pretightening force F2 in the buffer area A(i-2) are obtained.

[0024] The dynamic pretightening force Fˊ0 in the current buffer area Ai, the dynamic pretightening force Fˊ1 in the buffer area A(i-1) and the dynamic pretightening force Fˊ2 in the buffer area A(i-2) are obtained.

[0025] A preset judgment standard threshold H is obtained, and if the following conditions are met:

[0026] The working condition judgment result is pumping and phase modulation stable, and if the above conditions are not met, the existing working condition is maintained.

[0027] B) For the power generation stable working condition, a preset judgment standard threshold H2 is obtained, and if the following conditions are met:

[0028] The working condition judgment result is power generation stable, and if the above conditions are not met, the existing working condition is maintained.

[0029] G) For the shutdown working condition, the dynamic pretightening force Fˊ0 in the current buffer area Ai, the dynamic pretightening force main frequency proportion K0 in the current buffer area Ai and the dynamic pretightening force Fˊ1 in the buffer area A(i-1) are obtained.

[0030] A preset judgment standard threshold H is obtained, and if the following conditions are met:

[0031] The working condition judgment result is shutdown stable, and if the above conditions are not met, the existing working condition is maintained.

[0032] H) For pumping stable working condition: obtain the static pre-tightening force F0 in the current buffer zone Ai, the dynamic pre-tightening force F'0 in the current buffer zone Ai and the dynamic pre-tightening force F'2 in the buffer zone A(i-2);

[0033] Obtain the preset judgment standard threshold H and the targeted judgment threshold H1 and the targeted judgment threshold H2, if the following conditions are met:

[0034] The working condition judgment result is pumping stable, if the above conditions are not met, the existing working condition is maintained;

[0035] I) For power generation stable working condition: obtain the static pre-tightening force F0 in the current buffer zone Ai, the static pre-tightening force F1 in the buffer zone A(i-1) and the dynamic pre-tightening force F'2 in the buffer zone A(i-2);

[0036] Obtain the preset judgment standard threshold H, if the following conditions are met:

[0037] The working condition judgment result is power generation stable, if the above conditions are not met, the existing working condition is maintained;

[0038] J) For power generation shutdown transient stable working condition: obtain the static pre-tightening force F0 in the current buffer zone Ai, the static pre-tightening force F1 in the buffer zone A(i-1) and the dynamic pre-tightening force F'2 in the buffer zone A(i-2);

[0039] Obtain the preset judgment standard threshold H, if the following conditions are met:

[0040] The working condition judgment result is power generation shutdown transient stable, if the above conditions are not met, the existing working condition is maintained.

[0041] Further, the specific process of issuing an alarm signal is as follows:

[0042] Obtain a pre-warning rule library, the pre-warning rule library includes bolt standard data and unit standard data, the bolt standard data specifically is bolt standard stress data and bolt abnormal number threshold;

[0043] According to the bolt standard data, the bolt working condition is judged, the bolt working condition judgment result is obtained, the bolt working condition judgment result includes bolt abnormal working condition type and bolt abnormal number, if the bolt abnormal number is greater than or equal to the bolt abnormal number threshold, and the corresponding bolt abnormal working condition type is taken as the working condition judgment basis, if the unit and the unit standard data difference value is greater than the preset judgment threshold, an alarm signal is issued.

[0044] Further, the specific process of threshold optimization is as follows:

[0045] S201, obtain historical operation data of the water turbine as a training sample, the historical operation data including historical working condition data and a standard judgment threshold, the generated sample being divided into a training set, a validation set and a test set in proportions of 6:2:2, and a threshold prediction model being established based on a convolutional neural network;

[0046] S202, download a weight file and load it to a corresponding network for initializing migration network parameters;

[0047] S203, modify a last fully connected layer of the network, keep the input unchanged, set the output as a latest alarm threshold value, initialize weights of the last layer, use a gradient descent algorithm for learning, adopt fixed step attenuation to optimize training parameters, retrain the entire network, and obtain the threshold prediction model;

[0048] S204, randomly extract small batches of samples from the training set for training during the training process, all training pictures being extracted as a training cycle, the training being completed after iteration to a certain cycle, and the threshold prediction model being obtained;

[0049] S205, input real-time working condition data of the water turbine into the threshold prediction model to obtain an optimized latest alarm threshold value.

[0050] The application also provides a pump storage power station water turbine top cover bolt fault alarm method, comprising the following steps:

[0051] Step one, monitor stress and temperature data of the top cover bolts through grating stress sensors inside the top cover bolts, and obtain bolt characteristic quantities by performing feature extraction on the stress and temperature data;

[0052] Step two, obtain characteristic quantity data in a characteristic quantity cache area, the current characteristic quantity cache area being Ai, and the characteristic quantity data of each top cover bolt in the characteristic quantity cache area A(i-1), working condition data of each top cover bolt and unit working condition being obtained;

[0053] Step three, according to the working condition of each top cover bolt in the cache area A(i-1), call evaluation rules of corresponding working conditions in a pre-warning rule library to determine whether the working state of the top cover bolt is abnormal, if not, maintain the existing working condition, and if so, generate an alarm signal;

[0054] Step four, obtain the alarm signal and perform alarm reminding through an audible and visual alarm and a remote monitoring end.

[0055] As described above, due to adoption of the above technical solutions, the application has the following beneficial effects:

[0056] The pumped storage power station water turbine top cover bolt fault alarm system and alarm method, the stress and temperature data of the top cover bolt are monitored through the grating stress sensor inside each top cover bolt, the bolt characteristic quantity is obtained by feature extraction on the stress and temperature data, and the bolt characteristic quantity is saved to the current cache area Ai of the characteristic quantity cache area, when the bolt characteristic quantity calculation of three cache areas is completed, the working condition of each top cover bolt in the cache area A(i-1) and the unit working condition are determined through the bolt characteristic quantity data of the current cache area Ai and the adjacent cache area A(i-1) and cache area A(i-2), the working state of each unit is determined according to the bolt characteristic quantity and working condition data in the cache area A(i-1), and an alarm signal is sent out, and a threshold prediction model is established according to the convolutional neural network to output the optimized latest alarm threshold and send it to the early warning rule library, which can monitor the state of the top cover bolt of the water turbine in real time, avoid water turbine failure, and through the updated alarm threshold, it can be suitable for water turbine aging and performance difference caused by environmental factors. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 The overall external structure schematic diagram of the present application is shown;

[0058] Figure 2 The characteristic quantity cache area structure schematic diagram of the present application is shown;

[0059] Figure 3 The single cache area structure schematic diagram of the present application is shown;

[0060] Figure 4 The method flow schematic diagram of the present application is shown. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0062] Embodiment 1:

[0063] As shown in Figures 1-3 The pumped storage power station water turbine top cover bolt fault alarm system comprises a data acquisition module for monitoring the stress and temperature data of the top cover bolt through the grating stress sensor inside each top cover bolt, and sending it to the data processing module.

[0064] The high-frequency strain signals of each top cover bolt are collected in real time and in parallel by the grating stress sensor, a double buffer structure is used for the collection thread, and a single buffer area stores the buffered data within a preset time (8 seconds), that is, when the last storage area after receiving the high-frequency strain signal of the grating stress sensor is 1, the data is sent to the data buffer area 2 for storage, and when the initial or last storage area after receiving the high-frequency strain signal of the grating stress sensor is 2, the data is sent to the data buffer area 1 for storage;

[0065] The buffer area structure is shown in Figure 2 The characteristic data are sequentially and circularly stored in the buffer areas A1, A2, A3,..., and the current buffer area number buffer stores the number of the current buffer area.

[0066] The data processing module includes a bolt characteristic quantity calculation unit and a bolt and unit operating condition calculation unit, wherein:

[0067] The bolt characteristic quantity calculation unit obtains data by accessing the double buffer area of the data acquisition module, obtains each bolt characteristic quantity by performing feature extraction on the stress and temperature data, and stores each bolt characteristic quantity in the current buffer area Ai of the characteristic quantity buffer area, the bolt characteristic quantity includes dynamic pre-tightening force, static pre-tightening force and dynamic stress main frequency ratio, and the characteristic quantity buffer area adopts a multi-buffer circular structure, each characteristic quantity buffer area includes a bolt characteristic quantity data buffer area, a bolt operating condition buffer area and a unit operating condition buffer area.

[0068] The specific process of feature extraction to obtain the characteristic quantity of the top cover bolt is as follows:

[0069] S101, obtain buffer data, the buffer data including strain wavelength data and temperature wavelength data;

[0070] S102, calculate the strain data ε according to the following formula: ε = ((λ-λ0)-(λ T -λ T0 )*K T ) / K,

[0071] Wherein: λ is the test wavelength, λ0 is the zero wavelength, λ T is the temperature compensation wavelength, λ T0 is the temperature compensation zero wavelength, k is the sensitivity coefficient, and k T is the temperature compensation coefficient.

[0072] S102, calculate the stress data of each bolt according to the strain data, based on the fact that there is a correlation between stress and strain through the elastic modulus of the material itself, calculate the stress according to the following formula: σ = E*ε, wherein σ is the stress, ε is the strain, and E is the elastic modulus of the material itself.

[0073] S103, performing FFT calculation on each bolt data array to obtain bolt frequency domain data, and then obtaining characteristic quantity data such as main frequency and secondary main frequency, and performing mean value calculation, maximum value calculation and minimum value calculation on each bolt data to obtain maximum value characteristic quantity, minimum value characteristic quantity and average value characteristic quantity, and storing them in a characteristic quantity buffer area, i.e. the current buffer area Ai.

[0074] The bolt working condition and the unit working condition both include the following working condition states:

[0075] Specifically, the working conditions include shutdown condition, power generation start condition, power generation condition, power generation shutdown transient condition, pumping start condition, pumping and phase modulation condition, pumping condition and pumping shutdown transient condition.

[0076] When the bolt characteristic quantity calculation unit of the three buffer areas is used to complete the bolt characteristic quantity calculation, the working condition of each top cover bolt in the buffer area A(i-1) is determined through the bolt characteristic quantity data of the current buffer area Ai and the adjacent buffer areas A(i-1) and A(i-2), and the determination results of each bolt working condition are saved to the buffer area A(i-1), and the data of the buffer area A(i-1) is saved to the characteristic quantity database.

[0077] The evaluation rules of each working condition are as follows:

[0078] A) For the pumping and phase modulation stable working condition, the static pretightening force F0 in the current buffer area Ai, the static pretightening force F1 in the buffer area A(i-1) and the static pretightening force F2 in the buffer area A(i-2) are obtained.

[0079] The dynamic pretightening force Fˊ0 in the current buffer area Ai, the dynamic pretightening force Fˊ1 in the buffer area A(i-1) and the dynamic pretightening force Fˊ2 in the buffer area A(i-2) are obtained.

[0080] A preset judgment standard threshold H is obtained, and if the following conditions are met:

[0081] The working condition determination result is pumping and phase modulation stability, and if the above conditions are not met, the existing working condition is maintained.

[0082] B) For the power generation stable working condition, a preset judgment standard threshold H2 is obtained, and if the following conditions are met:

[0083] The working condition determination result is power generation stability, and if the above conditions are not met, the existing working condition is maintained.

[0084] K) For the shutdown working condition, the dynamic pretightening force Fˊ0 in the current buffer area Ai, the dynamic pretightening force Fˊ1 in the buffer area A(i-1) and the dynamic pretightening force Fˊ1 in the buffer area A(i-1) are obtained.

[0085] acquire a preset judgment criterion threshold value H, if the following conditions are met:

[0086] the working condition judgment result is stable shutdown, if the above conditions are not met, the existing working condition is maintained;

[0087] L) for pumping stable working condition: acquire the static pretightening force F0 in the current cache area Ai, the dynamic pretightening force F'0 in the current cache area Ai and the dynamic pretightening force F'2 in the cache area A(i-2);

[0088] acquire a preset judgment criterion threshold value H and a targeted judgment threshold value H1 and a targeted judgment threshold value H2, if the following conditions are met:

[0089] the working condition judgment result is pumping stable, if the above conditions are not met, the existing working condition is maintained;

[0090] M) for generating stable working condition: acquire the static pretightening force F0 in the current cache area Ai and the static pretightening force F1 in the cache area A(i-1) and the dynamic pretightening force F'2 in the cache area A(i-2);

[0091] acquire a preset judgment criterion threshold value H, if the following conditions are met:

[0092] the working condition judgment result is generating stable, if the above conditions are not met, the existing working condition is maintained;

[0093] N) for generating stable working condition: acquire the static pretightening force F0 in the current cache area Ai and the static pretightening force F1 in the cache area A(i-1) and the dynamic pretightening force F'2 in the cache area A(i-2);

[0094] acquire a preset judgment criterion threshold value H, if the following conditions are met:

[0095] the working condition judgment result is generating stable, if the above conditions are not met, the existing working condition is maintained.

[0096] The early warning module is used to acquire the bolt characteristic quantity in the cache area A(i-1) and the judgment result of the bolt working condition and processing, calling the early warning rule library, judging the unit working condition through the bolt working condition judgment result, and issuing an alarm signal, at the same time, through the sound and light alarm and the remote monitoring end, the alarm is reminded;

[0097] The specific process of issuing an alarm signal is as follows:

[0098] acquire the early warning rule library, the early warning rule library includes bolt standard data and unit standard data, the bolt standard data is specifically bolt standard stress data and bolt abnormal number threshold value;

[0099] According to the bolt standard data, the bolt working condition is judged, and the bolt working condition judgment result is obtained, the bolt working condition judgment result includes bolt abnormal working condition type and bolt abnormal number, if the bolt abnormal number is greater than or equal to the bolt abnormal number threshold, and the corresponding bolt abnormal working condition type is taken as the working condition judgment basis, if the unit and unit standard data difference value is greater than the preset judgment threshold, an alarm signal is sent out;

[0100] The threshold optimization module is used for establishing a threshold prediction model according to the convolutional neural network, training the model by obtaining the historical operation data of the water turbine from the feature database, and outputting the optimized latest alarm threshold and sending it to the early warning rule library.

[0101] The specific process of threshold optimization is as follows:

[0102] S201, obtain the historical operation data of the water turbine as the training sample, the historical operation data includes the historical working condition data and the standard judgment threshold, the generated sample is divided into training set, validation set and test set according to the proportion 6:2:2, and the threshold prediction model is established based on the convolutional neural network;

[0103] S202, download the weight file and load it to the corresponding network for initializing the migration network parameters;

[0104] S203, modify the last fully connected layer of the network, keep the input unchanged, set the output to the latest alarm threshold, initialize the weight of the last layer, use gradient descent algorithm for learning, and use fixed step attenuation to optimize the training parameters, retrain the whole network, and get the threshold prediction model;

[0105] S204, during the training process, a small batch of samples are randomly extracted from the training set for training, all training pictures are extracted as a training period, the training is completed after a certain period of iteration, and the threshold prediction model is obtained;

[0106] S205, input the real-time working condition data of the water turbine into the threshold prediction model to obtain the optimized latest alarm threshold.

[0107] The application monitors stress and temperature data of the top cover bolts through the grating stress sensors inside the bolts, obtains bolt characteristic quantities by feature extraction on the stress and temperature data, and saves the bolt characteristic quantities to the current cache area Ai of the characteristic quantity cache area. When the bolt characteristic quantity calculation of the three cache areas is completed, the working conditions of each top cover bolt and the unit working condition in the cache area A(i-1) are determined through the bolt characteristic quantity data of the current cache area Ai and the adjacent cache areas A(i-1) and A(i-2), the working state of each unit is determined according to the bolt characteristic quantity and working condition data in the cache area A(i-1), and an alarm signal is sent. At the same time, a threshold prediction model is established according to the convolutional neural network to output the optimized latest alarm threshold and send it to the early warning rule library. The state of the top cover bolts of the water turbine can be monitored in real time to avoid water turbine failure, and the updated alarm threshold can be suitable for the performance difference caused by the aging and environmental factors of the water turbine.

[0108] Embodiment 2

[0109] As shown in Figure 4 The pump storage power station water turbine top cover bolt fault alarm method comprises the following steps:

[0110] Step one, monitor the stress and temperature data of the top cover bolts through the grating stress sensors inside the bolts, and obtain the bolt characteristic quantities by feature extraction on the stress and temperature data;

[0111] Step two, obtain the characteristic quantity data in the characteristic quantity cache area, and obtain the top cover bolt characteristic quantity data, top cover bolt working condition data and unit working condition of the cache area A(i-1) in the current characteristic quantity cache area Ai;

[0112] Step three, according to the working condition of each top cover bolt in the cache area A(i-1), call the evaluation rule of the corresponding working condition in the early warning rule library to determine whether the working state of the top cover bolt is abnormal. If there is no abnormality, the existing working condition is maintained, and if there is an abnormality, an alarm signal is generated;

[0113] Step four, obtain the alarm signal and alarm and remind through the audible and visual alarm and the remote monitoring end.

[0114] The setting of the interval and the size of the threshold is to facilitate comparison. The size of the threshold depends on the number of sample data and the base number set by the person skilled in the art for each group of sample data. As long as it does not affect the proportional relationship between the parameters and the quantized values.

[0115] The above formulas are dimensionless values, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the nearest real situation. The preset parameters in the formula are set by the person skilled in the art according to the actual situation;

[0116] The above merely describes preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art, according to the technical solution and inventive concept of the present application, makes equivalent replacement or change within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A fault alarm system for the top cover bolts of a pumped storage power station turbine, characterized in that, It includes a data acquisition module, a data processing module, an early warning module, and a threshold optimization module, among which: The data acquisition module is used to monitor the stress and temperature data of the top cover bolts through the grating stress sensor inside each top cover bolt, and send the data to the data processing module. The data processing module includes a bolt characteristic quantity calculation unit and a bolt and unit operating condition calculation unit, wherein: The bolt feature quantity calculation unit obtains data by accessing the dual buffer area of ​​the data acquisition module, extracts features from stress and temperature data to obtain the bolt feature quantities, and saves each bolt feature quantity to the current buffer area Ai of the feature quantity buffer area. The bolt feature quantities include dynamic preload, static preload, and dynamic stress frequency ratio. The feature quantity buffer area adopts a multi-buffer loop structure, and each feature quantity buffer area includes a bolt feature quantity data buffer area, a bolt operating condition buffer area, and a unit operating condition buffer area. When the bolt and unit operating condition calculation unit is used to complete the bolt feature quantity calculation of the three buffer zones, it determines the operating condition of each top cover bolt in buffer zone A(i-1) by using the bolt feature quantity data of the current buffer zone Ai and the adjacent buffer zones A(i-1) and A(i-2), and saves the judgment result of each bolt operating condition to buffer zone A(i-1), and saves the data of buffer zone A(i-1) to the feature quantity database. The early warning module is used to obtain and process the bolt feature quantity and bolt condition judgment result in the buffer A(i-1), call the early warning rule library, judge the unit condition based on the bolt condition judgment result, and issue an alarm signal. At the same time, it provides alarm reminders through the audible and visual alarm and the remote monitoring terminal. The threshold optimization module is used to build a threshold prediction model based on a convolutional neural network. It obtains historical operating data of the turbine from the feature database to train the model, outputs the latest optimized alarm threshold, and sends it to the early warning rule base.

2. The pumped storage power station turbine top cover bolt fault alarm system according to claim 1, characterized in that, The data acquisition module is a separate thread in the program, and adopts a dual-buffer structure to achieve continuous data acquisition. It also shares data with the data processing module through the dual buffer.

3. The pumped storage power station turbine top cover bolt fault alarm system according to claim 1, characterized in that, The specific process of feature extraction to obtain the feature quantities of the top cover bolts is as follows: S101. Obtain cached data, wherein the cached data includes strain wavelength data and temperature wavelength data; S102. Calculate the strain data ε according to the following formula: ε=((λ-λ0)-(λ T -λ T0 )*K T ) / K, Where: λ is the test wavelength, λ0 is the zero-point wavelength, and λ T For temperature-compensated wavelength, λ T0 The temperature-compensated zero-point wavelength is given, and k is the sensitivity coefficient. T This is the temperature compensation coefficient; S102. Calculate the stress data of each bolt based on the strain data. Since there is a correlation between stress and strain through the elastic modulus of the material itself, calculate the stress according to the following formula: σ=E*ε, where σ is stress, ε is strain, and E is the elastic modulus of the material itself. S103. Perform FFT calculation on each bolt data array to obtain the frequency domain data of each bolt, and then obtain the main frequency, secondary main frequency and other feature data. Perform mean, maximum and minimum value calculation on each bolt data one by one to obtain the maximum value feature, minimum value and other feature, and average value and other feature, and store them in the feature buffer, i.e. the current buffer Ai.

4. The pumped storage power station turbine top cover bolt fault alarm system according to claim 1, characterized in that, Both the bolt operating condition and the unit operating condition include the following operating states: Specifically, these include shutdown operation, power generation start-up operation, power generation operation, power generation shutdown transient operation, pumping start-up operation, pumping phase adjustment operation, pumping operation, and pumping shutdown transient operation.

5. The pumped storage power station turbine top cover bolt fault alarm system according to claim 1, characterized in that, The evaluation rules for each working condition are as follows: A) For the stable operation of pumping phase adjustment: obtain the static preload F0 in the current buffer zone Ai, the static preload F1 in the buffer zone A(i-1), and the static preload F2 in the buffer zone A(i-2); Get the dynamic preload Fˊ0 in the current buffer Ai, the dynamic preload Fˊ1 in buffer A(i-1), and the dynamic preload Fˊ2 in buffer A(i-2); Obtain the preset judgment standard threshold H. If the following conditions are met: If the operating condition is determined to be stable, the pumping phase adjustment is stable. If the above conditions are not met, the original operating condition is maintained. B) For stable power generation conditions: Obtain the preset judgment threshold H2. If the following conditions are met: If the operating condition is determined to be stable, then the power generation will remain stable. If the above conditions are not met, the power generation will remain stable. C) For shutdown conditions, obtain the dynamic preload Fˊ0 in the current buffer Ai, the dynamic preload frequency ratio K0 in the current buffer Ai, and the dynamic preload Fˊ1 in the buffer A(i-1); Obtain the preset judgment standard threshold H. If the following conditions are met: If the operating condition is determined to be stable after shutdown, the original operating condition will be maintained if the above conditions are not met. D) For stable pumping conditions: obtain the static preload F0 in the current buffer zone Ai, the dynamic preload Fˊ0 in the current buffer zone Ai, and the dynamic preload Fˊ2 in the buffer zone A(i-2); Obtain the preset judgment standard threshold H, and the targeted judgment thresholds H1 and H2. If the following conditions are met: If the operating condition is determined to be stable, the pumping is stable. If the above conditions are not met, the original operating condition is maintained. E) For stable power generation conditions: obtain the static preload F0 in the current buffer zone Ai, the static preload F1 in the buffer zone A(i-1), and the dynamic preload Fˊ2 in the buffer zone A(i-2); Obtain the preset judgment standard threshold H. If the following conditions are met: If the operating condition is determined to be stable, the power generation is stable. If the above conditions are not met, the original operating condition is maintained. F) For the transient stable condition of power generation shutdown: obtain the static preload F0 in the current buffer zone Ai, the static preload F1 in the buffer zone A(i-1), and the main frequency ratio K0 of the dynamic preload in the previous buffer zone Ai; Obtain the preset judgment standard threshold H. If the following conditions are met: The operating condition judgment result is that the generator shutdown is transiently stable. If the above conditions are not met, the original condition is maintained.

6. The pumped storage power station turbine top cover bolt fault alarm system according to claim 1, characterized in that, The specific process for issuing an alarm signal is as follows: Obtain the early warning rule base, which includes bolt standard data and unit standard data. The bolt standard data specifically includes bolt standard stress data and bolt abnormality number threshold. The bolt condition is judged based on the bolt standard data, and the bolt condition judgment result is obtained. The bolt condition judgment result includes the bolt abnormal condition type and the number of bolt abnormalities. If the number of bolt abnormalities is greater than or equal to the bolt abnormality number threshold, and the corresponding bolt abnormal condition type is used as the condition judgment basis, if the difference between the unit and the unit standard data is greater than the preset judgment threshold, an alarm signal is issued.

7. The pumped storage power station turbine top cover bolt fault alarm system according to claim 1, characterized in that, The specific process of threshold optimization is as follows: S201. Obtain historical operating data of the water turbine as training samples. The historical operating data includes historical operating condition data and standard judgment thresholds. The generated samples are divided into training set, validation set and test set in a ratio of 6:2:

2. A threshold prediction model is established based on a convolutional neural network. S202. Download the weight file and load it onto the corresponding network to initialize the migration network parameters; S203. Modify the last fully connected layer of the network, keep the input unchanged, set the output to the latest alarm threshold, initialize the weights of the last layer, use the gradient descent algorithm for learning, and use fixed step size decay to optimize the training parameters. Retrain the entire network to obtain the threshold prediction model. S204. During the training process, small batches of samples are randomly and non-repeatedly drawn from the training set for training. One training cycle is completed after all training images are drawn. The training is completed after a certain number of cycles, and the threshold prediction model is obtained. S205. Input the real-time operating data of the turbine into the threshold prediction model to obtain the optimized latest alarm threshold.

8. A fault alarm method for the top cover bolts of a pumped storage power station turbine, characterized in that, Includes the following steps: Step 1: Monitor the stress and temperature data of the top cover bolts using grating stress sensors inside each top cover bolt, and extract the feature quantities of each bolt by performing feature extraction on the stress and temperature data. Step 2: Obtain feature data in the feature buffer. The current feature buffer is Ai. Obtain the feature data of each top cover bolt, the operating condition data of each top cover bolt, and the unit operating condition of the feature buffer A(i-1). Step 3: Based on the working conditions of each top cover bolt in the cache area A(i-1), call the evaluation rules of the corresponding working conditions in the early warning rule base to determine whether the working status of the top cover bolt is abnormal. If there is no abnormality, maintain the existing working conditions; if there is an abnormality, generate an alarm signal. Step 4: Obtain the alarm signal and send alarm reminders through the audible and visual alarm and the remote monitoring terminal.

Images