Fault detection method and detection device for pumped storage unit

By acquiring and integrating fault signals under multiple operating conditions, identifying the fault mode and severity of the pumped storage unit, the one-sided problem of data acquisition in the prior art is solved, and the accuracy of fault detection and operation and maintenance support is improved.

CN118965114APending Publication Date: 2024-11-15이너 몽골리아 일렉트릭 파워 그룹 컴퍼니 리미티드 이너 몽골리아 일렉트릭 파워 리서치 인스티튜트 브랜치
View PDF 0 Cites 12 Cited by

Patent Information

Application Number
CN202411032417.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-30
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The data collection in the existing pumped storage unit fault detection methods is relatively one-sided, which makes it difficult to expose faults under different working conditions, reducing the accuracy of fault detection.

Method used

Provide a method and device for fault detection of pumped storage units. By obtaining multiple component fault signals under multiple operating conditions, extracting and fusing characteristic parameters, identifying fault patterns, and conducting severity assessment and development trend prediction, generating detailed faults. Diagnostic report.

Benefits of technology

It improves the accuracy and reliability of fault detection, avoids the one-sidedness of a single working condition, significantly improves the exposure rate of invisible faults, and provides multi-dimensional fault information to support operation and maintenance decisions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118965114A_ABST
    Figure CN118965114A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of pumped storage, and provides a fault detection method and device for a pumped storage unit. The fault detection method for the pumped storage unit comprises the following steps: acquiring a rotating part fault signal, a bearing part fault signal, a water guide part fault signal, a sealing part fault signal and an embedded part fault signal of the pumped storage unit under various working conditions, and extracting and fusing characteristic parameters of the fault signals; and identifying a fault mode and evaluating severity based on the fused feature parameters, predicting a fault development trend and a key time point through historical data and confidence analysis, and generating a diagnosis report containing the fault mode, the severity level and the trend. According to the fault detection method and device for the pumped storage unit, the defects that in an existing fault detection method, data collection is one-sided, faults generated under different working conditions are not prone to exposure, and consequently the fault detection accuracy is low are overcome.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of pumped storage, and in particular to a pumped storage unit fault detection method and detection device. Background Art

[0002] Pumped storage units refer to synchronous motors that have two working modes: pumping and generating electricity. They can use water energy for energy conversion. They include water pumps, turbines, generators, and control systems. As key electrical equipment in hydropower stations, pumped storage units can start and stop quickly, and can adjust loads quickly and conveniently. They undertake a large number of peak-shaving, frequency-regulating, voltage-regulating, and emergency standby tasks in the power grid. However, as the operating conditions of the units switch more frequently, the risk of operating failures also increases. Therefore, in order to ensure the safe and stable operation of the units and improve their reliability and efficiency, fault detection devices are required.

[0003] During the operation of the pumped storage unit, due to the various operating conditions of the unit, when the unit is generating electricity, pumping water, power generation phase adjustment and water pump phase adjustment, the specific operating conditions of each part are also different. In the prior art, data capture of a single operating condition of the unit is prone to one-sided data collection, resulting in failures caused by different operating conditions not being easily exposed, increasing the difficulty of data collection and reducing the accuracy of fault detection. Summary of the invention

[0004] The present invention provides a pumped storage unit fault detection method and detection device, which are used to solve the defects of the existing pumped storage unit fault detection method, that is, data collection is relatively one-sided, faults caused by different working conditions are not easy to expose, resulting in low fault detection accuracy.

[0005] On one hand, the present invention provides a method for detecting faults of a pumped storage unit, comprising the following steps.

[0006] Obtain the rotating part fault signal, bearing part fault signal, water guide part fault signal, sealing part fault signal and embedded part fault signal of the pumped storage unit when it is operating under power generation condition, pumping condition, power generation phase adjustment condition and water pump phase adjustment condition.

[0007] The characteristic parameters of the rotating part fault signal, the bearing part fault signal, the water guide part fault signal, the sealing part fault signal and the embedded part fault signal are extracted.

[0008] The characteristic parameters of the rotating part fault signal, the bearing part fault signal, the water guide part fault signal, the sealing part fault signal and the embedded part fault signal are fused to determine the fused characteristic parameters.

[0009] Based on the fused characteristic parameters, the failure mode of the pumped-storage unit is identified, and the failure mode of the pumped-storage unit and the confidence level corresponding to the failure mode are determined.

[0010] Based on the failure mode of the pumped storage unit, the characteristic parameters corresponding to the failure mode, and the degree of deviation from the preset normal operating parameter range, the severity of the identified failure mode is evaluated to determine the severity level of the failure.

[0011] Based on historical fault signals, the fault mode of the pumped storage unit and its confidence, and the severity level of the pumped storage unit fault, the fault development trend is predicted, and the predicted fault development trend and critical time point are determined.

[0012] Based on the failure mode of the pumped-storage unit and its confidence, the severity level of the failure, the predicted failure development trend and critical time point, a failure diagnosis report is generated, wherein the failure diagnosis report includes the failure mode of the pumped-storage unit, the severity level of the failure and the failure development trend.

[0013] On the other hand, the present invention provides a pumped-storage unit fault detection device, including a rotating part detection module, a bearing part detection module, a water guiding part detection module, a sealing part detection module, an embedded part detection module and a data integration and analysis module, wherein the rotating part detection module, the bearing part detection module, the water guiding part detection module, the sealing part detection module and the embedded part detection module are all communicatively connected with the data integration and analysis module.

[0014] The rotating part detection module, the bearing part detection module, the water guiding part detection module, the sealing part detection module and the embedded part detection module are respectively used to obtain the rotating part fault signal, the bearing part fault signal, the water guiding part fault signal, the sealing part fault signal and the embedded part fault signal when the pumped-storage unit is operating in power generation conditions, pumping conditions, power generation phase adjustment conditions and water pump phase adjustment conditions.

[0015] The data integration and analysis module is used to: extract characteristic parameters of the rotating part fault signal, the bearing part fault signal, the water guide part fault signal, the sealing part fault signal and the embedded part fault signal; fuse the characteristic parameters of the rotating part fault signal, the bearing part fault signal, the water guide part fault signal, the sealing part fault signal and the embedded part fault signal to determine the fused characteristic parameters; based on the fused characteristic parameters, identify the fault mode of the pumped-storage unit, determine the fault mode of the pumped-storage unit and the confidence corresponding to the fault mode; based on the fault mode of the pumped-storage unit, the characteristic parameters corresponding to the fault mode, and the confidence corresponding to the fault mode The method comprises the following steps: according to the degree of deviation from the preset normal operating parameter range, performing a severity assessment on the identified fault mode, and determining the severity level of the fault; based on the historical fault signals, the fault mode of the pumped-storage unit and its confidence level, and the fault severity level of the pumped-storage unit, predicting the fault development trend, and determining the predicted fault development trend and critical time point; based on the specific fault mode of the pumped-storage unit and its confidence level, the fault severity level, the predicted fault development trend and critical time point, generating a fault diagnosis report, wherein the fault diagnosis report includes the fault mode of the pumped-storage unit, the fault severity level and the fault development trend.

[0016] The pumped storage unit fault detection method provided by the present invention can provide fault diagnosis reports under different operating conditions, avoiding the one-sidedness of capturing the unit status under only one operating condition. The data capture of multiple operating conditions makes it easier to expose invisible faults, improves the accuracy of fault detection, and is conducive to the detection and corresponding maintenance of unit faults. The method solves the defects of the existing pumped storage unit fault detection method that data collection is relatively one-sided, faults caused by different operating conditions are not easy to expose, resulting in low fault detection accuracy.

[0017] The pumped storage unit fault detection device provided by the present invention can obtain the fault signals corresponding to the rotating part, bearing part, water guide part, sealing part and embedded part of the pumped storage unit when the pumped storage unit is operating in power generation condition, pumping condition, power generation phase adjustment condition and water pump phase adjustment condition respectively through the rotating part detection module, bearing part detection module, water guide part detection module, sealing part detection module and embedded part detection module, and comprehensively analyze the above fault signals through the data integration analysis module to identify the fault mode and / or fault development trend of the pumped storage unit. Similarly, the defects of the existing pumped storage unit fault detection method that the data collection is relatively one-sided, the faults caused by different working conditions are not easy to expose, and the fault detection accuracy is low are solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 Schematic diagram of a pumped storage unit provided in an embodiment of the present invention.

[0019] Figure 2It is a flow chart of a pumped storage unit fault detection method provided in an embodiment of the present invention.

[0020] Figure 3 It is a schematic diagram of a pumped storage unit fault detection device provided in an embodiment of the present invention.

[0021] Figure 4 It is a schematic diagram of a rotating part detection module in a pumped storage unit fault detection device provided in an embodiment of the present invention.

[0022] Figure 5 It is a schematic diagram of a bearing detection module in a pumped storage unit fault detection device provided in an embodiment of the present invention.

[0023] Figure 6 It is a schematic diagram of a water guide detection module in a pumped storage unit fault detection device provided in an embodiment of the present invention.

[0024] Figure 7 It is a schematic diagram of a sealing part detection module in a pumped storage unit fault detection device provided in an embodiment of the present invention.

[0025] Figure 8 It is a schematic diagram of an embedded detection module in a pumped storage unit fault detection device provided in an embodiment of the present invention.

[0026] Fig. 9 It is a schematic diagram of an operation recording module in a pumped storage unit fault detection device provided in an embodiment of the present invention.

[0027] Fig.10 It is a schematic diagram of the position of the sensor unit of the rotating part in the pumped storage unit fault detection device provided by an embodiment of the present invention.

[0028] Fig.11 It is a schematic diagram of the position of the sensor unit of the water guide part in the pumped storage unit fault detection device provided in an embodiment of the present invention.

[0029] Reference numerals: 1. Rotating part; 2. Bearing part; 3. Water guiding part; 4. Sealing part; 5. Embedded part; 6. Vibration energy converter; 7. Frequency monitoring sensor; 8. Acceleration sensor; 9. Ammeter; 10. Voltmeter; 11. Sound sensor; 12. Vibration sensor; 13. Thermal imager; 14. Fiber optic sensor; 15. Flow meter; 16. pH meter; 17. Hygrometer; 18. Strain sensor; 19. Pressure sensor; 20. Lubricant distributor; 21. Water level sensor; 22. Flow meter. DETAILED DESCRIPTION

[0030] See also Figure 1As shown, before describing the pumped storage unit fault detection method and detection device provided by the present invention, the structure of the pumped storage unit is briefly described.

[0031] Pumped storage units are usually built underground or in underground chambers below the ground. The rotating part 1 includes rotating parts such as a turbine shaft and an impeller, which are arranged inside the pumped storage unit and connected to the generator through a bearing part 2; the bearing part 2 is located at both ends of the rotating part 1 and is used to support and fix the rotating shaft; the water guide part 3 is connected to the turbine, and the sealing part 4 is arranged outside the water guide part 3 to prevent water leakage inside the turbine; the buried part 5 is located inside the underground chamber.

[0032] See also Figure 2 As shown, the pumped storage unit fault detection method provided by the embodiment of the present invention includes the following steps.

[0033] S210, obtaining a rotating part fault signal, a bearing part fault signal, a water guiding part fault signal, a sealing part fault signal and an embedded part fault signal when the pumped-storage unit is operating in a power generation condition, a pumping condition, a power generation phase adjustment condition and a water pump phase adjustment condition.

[0034] S220, extracting characteristic parameters of the rotating part fault signal, the bearing part fault signal, the water guide part fault signal, the sealing part fault signal and the embedded part fault signal.

[0035] S230, fusing characteristic parameters of the rotating part fault signal, the bearing part fault signal, the water guide part fault signal, the sealing part fault signal and the embedded part fault signal to determine the fused characteristic parameters.

[0036] S240. Based on the fused characteristic parameters, the failure mode of the pumped-storage unit is identified to determine the failure mode of the pumped-storage unit and the confidence level corresponding to the failure mode.

[0037] S250: Based on the failure mode of the pumped-storage unit, the deviation between the characteristic parameters corresponding to the failure mode and the preset normal operating parameter range, a severity assessment is performed on the identified failure mode to determine the severity level of the failure.

[0038] Among them, the degree of deviation refers to the difference between the characteristic parameters corresponding to the fault mode and the preset normal operating parameter range. Specifically, if the characteristic parameters before fusion are used, the degree of deviation is the difference between the characteristic parameters of each component and its corresponding normal operating parameter range. If the characteristic parameters after fusion are used, the degree of deviation is the difference between the fused characteristic parameters and the preset normal operating parameter range. The purpose of calculating the degree of deviation is to quantify the distance between the fault characteristic parameters and the normal state, so as to evaluate the severity of the fault. This degree of deviation can be calculated in a variety of ways, such as: absolute difference, relative percentage difference, standard deviation unit, and custom deviation measurement function.

[0039] S260, based on historical fault signals, the fault mode of the pumped storage unit and its confidence, and the fault severity level of the pumped storage unit, predict the fault development trend, and determine the predicted fault development trend and critical time point.

[0040] S270. Generate a fault diagnosis report based on the specific fault mode of the pumped-storage unit and its confidence level, the fault severity level, the predicted fault development trend and the critical time point. The fault diagnosis report includes the fault mode of the pumped-storage unit, the fault severity level and the fault development trend.

[0041] The present invention provides a comprehensive, accurate and reliable method for fault detection of pumped storage units. By acquiring multi-component fault signals under multiple working conditions and combining characteristic parameter fusion technology, the accuracy and reliability of fault identification are significantly improved. This method not only realizes the quantitative evaluation of the fault state, but also provides an intuitive fault description through the classification of severity levels. The trend prediction function based on historical data and current status provides forward-looking guidance for equipment maintenance. The comprehensive fault diagnosis report finally generated contains multi-dimensional information such as fault mode, severity and development trend, providing comprehensive support for operation and maintenance decisions.

[0042] According to some embodiments of the present invention, obtaining a rotor fault signal of a pumped-storage unit when operating in a power generation condition, a pumping condition, a power generation phase adjustment condition, and a water pump phase adjustment condition includes the following steps.

[0043] The sound signal, amplitude signal, frequency signal, rotation speed signal and vibration direction signal of the rotor of the rotating part 1 are obtained, and the rotating part fault signal is determined based on the sound signal, amplitude signal, frequency signal, rotation speed signal and vibration direction signal of the rotor.

[0044] According to some embodiments of the present invention, determining a rotor fault signal based on a sound signal of a rotor of the rotor 1 specifically includes the following steps.

[0045] A fault identification model is established to classify and store the sound frequency characteristics of the pumped-storage unit in normal operation and various fault states.

[0046] The sound frequency features extracted from the sound signal of the rotor of the rotating part 1 are compared with the sound spectrum features in the fault identification model.

[0047] The fault signal of the rotating part is determined based on the comparison result of the sound frequency characteristics extracted from the sound signal and the sound spectrum characteristics in the fault identification model.

[0048] The specific processing process is as follows.

[0049] Firstly, the sound signal of the main shaft rotor in the rotating part 1 of the pumped storage unit during operation is collected by the sensor, and the features thereof are extracted, including frequency components, amplitude, time domain and frequency domain information.

[0050] Signal processing algorithms (such as Fast Fourier Transform FFT, Short-time Fourier Transform STFT, and Wavelet Transform) are used to analyze sound signals and extract useful frequency features from them. These algorithms can transform sound signals from the time domain to the frequency domain to reveal their frequency components and energy distribution.

[0051] Perform fault pattern recognition. This includes: establishing a fault recognition model to classify and store the sound spectrum characteristics under normal operation and various fault conditions. The model can be built based on historical data, experimental data and expert knowledge through machine learning, statistical analysis or pattern recognition.

[0052] In actual operation, the sound signal of the main shaft rotor in the rotating part 1 of the pumped storage unit is collected and analyzed in real time, and the extracted frequency features are compared with the features in the fault identification model. If the monitored sound signal features match certain fault features in the model, it is determined that the main shaft rotor in the rotating part 1 of the pumped storage unit has a corresponding fault.

[0053] For example, when the main shaft rotor of a pumped storage unit is operating normally, it will produce a specific sound spectrum. This spectrum contains different frequency components and their amplitude distribution. When the rotor has faults such as wear, cracks or imbalance, the spectrum of the sound signal will shift (for example, the intensity of certain frequency components increases or decreases, new frequency components appear, etc.). By analyzing and comparing these changes, the fault identification model can determine whether there is a fault in the rotor and its type.

[0054] According to some embodiments of the present invention, determining a rotor fault signal based on an amplitude signal, a frequency signal, a rotation speed signal and a vibration direction signal of a rotor of the rotor 1 specifically includes the following steps.

[0055] Based on the amplitude signal, the frequency signal and the rotation speed signal of the rotor of the rotating part 1 , the vibration energy of the rotor of the rotating part 1 is calculated to obtain the vibration energy data of the rotor of the rotating part 1 .

[0056] Based on the vibration energy data of the rotor of the rotating part 1 and the vibration direction signal of the rotor of the rotating part 1, the source and characteristics of the vibration are determined to determine the fault signal of the rotating part.

[0057] Specifically, the amplitude, frequency, rotation speed and vibration direction of the rotor can be collected by a vibrometer or a vibration sensor 12, and the vibration energy of the rotor can be calculated according to the vibration energy formula to obtain vibration energy data, and the vibration energy data can be output. The rotation speed is expressed in revolutions per minute (RPM) or angular velocity (radians per second), and may include the linear velocity of the rotor (the distance the outer edge of the rotor moves per unit time) or the angular velocity (the angle at which the rotor rotates per unit time). The vibration direction refers to the specific direction of the vibration in space, which may be along different axes (such as the X-axis, Y-axis, and Z-axis in a spatial rectangular coordinate system) or in any direction between these axes. By determining the vibration direction, the source and characteristics of the vibration can be analyzed more accurately, thereby diagnosing and handling faults.

[0058] According to some embodiments of the present invention, obtaining a bearing fault signal of a pumped-storage unit when the unit is operating in a power generation condition, a pumping condition, a power generation phase adjustment condition, and a water pump phase adjustment condition includes the following steps.

[0059] The temperature distribution signal, deformation signal, stress state signal and vibration state signal of the bearing part 2 are obtained, and the bearing part fault signal is determined based on the temperature distribution signal, deformation signal, stress state signal and vibration state signal of the bearing part 2.

[0060] According to some embodiments of the present invention, determining a bearing fault signal based on a temperature distribution signal of the bearing 2 specifically includes the following steps.

[0061] Based on the temperature distribution signal of the bearing portion 2, a temperature distribution image of the bearing portion 2 is determined.

[0062] Based on the temperature distribution image of the bearing portion 2, an abnormal temperature region of the bearing portion 2 is determined to determine a bearing portion fault signal.

[0063] Specifically, infrared thermal imaging technology (such as thermal imager 13) can be used to detect the temperature distribution of the bearing part 2 in real time. Through infrared thermal imaging technology, the temperature distribution image of the surface of the bearing part 2 can be obtained non-contactly and analyzed to detect abnormal temperature areas, thereby identifying potential thermal faults.

[0064] According to some embodiments of the present invention, determining a bearing part fault signal based on a deformation signal and a stress state signal of the bearing part 2 includes the following steps.

[0065] An optical signal of a fiber optic sensor 14 located in the bearing part 2 or in a supporting structure of the bearing part 2 is acquired.

[0066] Based on the change of the optical signal, the deformation signal and the stress state signal of the bearing part 2 are determined to determine the bearing part fault signal.

[0067] Specifically, the optical fiber sensor 14 can be embedded in the bearing part 2 or the supporting structure of the bearing part 2 to detect the deformation information and stress state of the bearing part 2 in real time. By detecting the change of the optical signal of the optical fiber sensor 14, the force condition of the bearing part 2 can be obtained in real time, and the working state of the bearing part 2 can be evaluated according to the deformation information and stress distribution condition.

[0068] According to some embodiments of the present invention, determining a bearing fault signal based on a vibration state signal of the bearing 2 includes the following steps.

[0069] The vibration state signal of the bearing part 2 is subjected to time domain analysis and frequency domain analysis to determine the operating frequency, vibration amplitude and vibration mode of the bearing part 2 .

[0070] Based on the operating frequency, vibration amplitude and vibration mode of the bearing part 2, the vibration state of the bearing part 2 is determined to determine a bearing part fault signal.

[0071] Specifically, sensing technology and algorithms can be used to detect the working status and health status of the bearing part 2 in real time. By collecting vibration, sound, current and other signals of the bearing part 2 and combining them with signal processing algorithms, the operation status of the bearing part 2 can be analyzed in real time to identify possible fault characteristics.

[0072] The sensors and algorithms used above include the following steps.

[0073] The vibration signal of the bearing 2 is detected by using the acceleration sensor 8, and the vibration spectrum is analyzed by fast Fourier transform FFT to identify the abnormal vibration mode.

[0074] A microphone or an acoustic sensor is used to capture the sound signal generated when the bearing part 2 is in operation, and a short-time Fourier transform STFT or a wavelet transform is used to perform frequency analysis to detect abnormal sound characteristics.

[0075] The motor current signal is monitored by the current sensor, and the time series analysis and autoregressive model (AR model) are applied to analyze the change trend and abnormal fluctuation of the current signal.

[0076] The data collected by the above sensors are input into a machine learning model (such as a support vector machine (SVM) or a deep neural network (DNN). The trained model can automatically identify and classify different types of bearing part 2 faults.

[0077] Further, according to some embodiments of the present invention, obtaining a bearing fault signal during operation of the pumped-storage unit further includes the following steps.

[0078] A fault prediction model for the bearing part 2 is established.

[0079] The temperature distribution signal, deformation signal, stress state signal and vibration state signal of the bearing part 2 are input into the fault prediction model of the bearing part 2 .

[0080] Based on the output result of the fault prediction model of the bearing part 2, the fault occurrence time of the bearing part 2 is determined and a warning signal is issued. The output result of the prediction model includes a fault occurrence time prediction signal, a fault warning signal and a fault type identification signal.

[0081] Specifically, a prediction model for bearing part 2 failure is established, and historical data is used for training to predict the occurrence time of bearing part 2 failure and issue a warning signal. Based on a machine learning algorithm or a deep learning model, the fault prediction unit can use historical data and real-time detection data to accurately predict the occurrence time of bearing part 2 failure and issue an alarm in time so that maintenance personnel can take corresponding maintenance measures to avoid production interruption and equipment damage caused by the failure.

[0082] During operation, the support vector machine algorithm can be used to classify and predict the bearing part 2 faults. By training the SVM model, features such as the rotation temperature and rotation speed of the bearing part 2 and the stress distribution of the bearing part 2 during rotation can be extracted from the bearing part 2 detection data and classified to identify the normal working state and different types of fault states of the bearing part 2. At the same time, the SVM model can be used to perform regression analysis on the fault data of the bearing part 2 to predict the remaining life of the bearing part 2 or the time when a fault occurs.

[0083] According to some embodiments of the present invention, obtaining a water guide fault signal when a pumped-storage unit is operating in a power generation condition, a pumping condition, a power generation phase adjustment condition, and a water pump phase adjustment condition includes the following steps.

[0084] The running state data of the water guiding part 3 is acquired, and based on the running state data of the water guiding part 3, a water guiding part fault signal is determined, wherein the running state data of the water guiding part 3 includes water flow characteristic data and pressure data.

[0085] Specifically, the water flow characteristic data of the water guide 3 is obtained in the following ways: a flow sensor is used to measure the flow rate of the water flow of the water guide 3, that is, the amount of water passing through a certain cross section per unit time. A flow velocity sensor is used to measure the velocity of the water flow of the water guide 3. An ultrasonic flow meter is used to measure the flow rate and flow velocity by using the change in the propagation velocity of ultrasonic waves in water. An electromagnetic flow meter is used to measure the flow rate by detecting the induced electromotive force of the conductive liquid in the magnetic field.

[0086] The pressure data of the water guide 3 is obtained in the following ways: The pressure sensor 19 is installed on the pipe or structure of the water guide 3 to monitor the pressure of the water flow in real time. The differential pressure sensor (not shown in the figure) is used to measure the pressure difference between two points and can provide more accurate fluid dynamics information.

[0087] According to some embodiments of the present invention, the water flow characteristic data includes a water flow pressure signal, a water flow velocity signal, and a water flow hydraulic characteristic signal of the water guide portion 3. That is, the water flow pressure signal, the water flow velocity signal, and the water flow hydraulic characteristic signal of the water guide portion 3 can be used to determine whether there is an abnormality in the water guide pipe.

[0088] Further, according to some embodiments of the present invention, the running state data of the water guide 3 also includes structural state data, and the structural state data includes strain data and vibration data of the water guide 3. Therefore, it is possible to determine whether the water guide 3 has a structural fault according to the structural state data of the water guide 3.

[0089] Specifically, the structural status data of the water guide 3 can be obtained in the following ways: a strain gauge is installed on the structural surface of the water guide 3 to measure the structural deformation and stress (strain data) of the water guide 3. An acceleration sensor 8 is used to monitor the vibration and dynamic response of the water guide 3 and evaluate the structural health status. A laser displacement sensor (not shown in the figure) is used to detect the micro displacement and deformation of the structural surface of the water guide 3. A temperature sensor is used to monitor the temperature change of the water guide 3, which helps to analyze the influence of thermal expansion and contraction on the water guide 3.

[0090] From the above description, it can be known that the operation status of the water guide 3 can be fully monitored through various types of sensor data. By integrating the water flow characteristic data, pressure data and structural status data, the fault and health status of the water guide 3 can be judged more accurately.

[0091] Further, according to some embodiments of the present invention, the operation status data of the water guide part 3 also includes environmental status data and water quality status data.

[0092] Specifically, the environmental status data can be combined with humidity sensors and data analysis algorithms to obtain detection parameters around the water pipe, including soil humidity, surface water level and groundwater level, and identify the location and extent of water leakage in real time based on the detection parameters, and issue an alarm in advance. The water quality status data includes the water quality parameters in the water pipe, and the water quality parameters are used to determine whether there is corrosion, pollution or biological growth inside the water pipe.

[0093] According to some embodiments of the present invention, obtaining a seal fault signal of a pumped-storage unit when the unit is operating in a power generation condition, a pumping condition, a power generation phase adjustment condition, and a water pump phase adjustment condition includes the following steps.

[0094] A pressure change signal of the sealing portion 4 is acquired, and a sealing portion failure signal is determined based on the pressure change signal.

[0095] By monitoring the pressure changes at the sealing part, signs of seal failure can be detected in time. For example, if the system detects that the pressure at the sealing part 4 continues to drop, this may be due to wear or damage of the sealing material. This timely fault detection allows the maintenance team to take quick action to replace the damaged seal and prevent possible leakage or other related problems.

[0096] Specifically, a pressure sensor 19 can be installed at a set position of the sealing part 4. The pressure sensor 19 can detect the pressure condition of the sealing part 4 in real time. Using the data acquisition system, the pressure sensor 19 transmits the collected pressure data to the background system for analysis. In the background system, a data processing algorithm, such as a sliding average method or Fourier transform, can be used to process and analyze the pressure data to identify whether there is abnormal pressure or leakage in the sealing part 4, so as to take corresponding maintenance measures in time to ensure the sealing performance of the equipment.

[0097] Furthermore, according to some embodiments of the present invention, the following steps are also included.

[0098] The lubrication data of the sealing portion 4 is acquired.

[0099] Based on the lubrication data of the sealing portion 4 , and the temperature distribution signal, deformation signal, stress state signal, and vibration state signal of the bearing portion 2 , the usage amount and usage frequency of the lubricant are adjusted.

[0100] The lubrication condition of the sealing parts is detected in real time, and the amount and frequency of lubricant use are adjusted according to the actual working conditions. Specifically, by installing a lubrication condition sensor or a lubricating oil flow sensor, the lubrication condition of the sealing parts is detected in real time. At the same time, combined with the working condition and environmental conditions of the equipment, the intelligent lubrication system can automatically adjust the amount and frequency of lubricant use to ensure that the sealing parts are always in the best lubrication state.

[0101] As an example, algorithms that can be used include PID control algorithms based on feedback control, which perform real-time detection and adjustment of lubrication status data obtained by sensors so that the amount and frequency of lubricant use can be adaptively adjusted according to the working status of the equipment, thereby achieving the best lubrication effect.

[0102] According to some embodiments of the present invention, obtaining an embedded part fault signal of a pumped-storage unit when it is operating in a power generation condition, a pumping condition, a power generation phase adjustment condition, and a water pump phase adjustment condition includes the following steps.

[0103] The water level signal and flow signal of the buried part 5 are obtained, and the buried part fault signal is determined by using the fuzzy logic system and the obtained water level signal and flow signal, and the changes in the water level and flow of the buried pipes and pre-buried pipes of the buried part 5 are predicted.

[0104] Specifically, the fuzzy logic system can be used to predict the changes in the water level and flow of buried pipes and pre-buried pipes. By installing a water level sensor 21 and a flow sensor (flow meter 22), the changes in the water level and flow can be detected in real time in the pipe system. With the help of the fuzzy logic system, the water level and flow of the pipe can be predicted and analyzed based on the data obtained by the above-mentioned water level sensor 21 and flow sensor, combined with pre-set rules and fuzzy reasoning methods. The fuzzy control algorithm can be used to realize automatic control of the water pump according to the changes in the water level and flow, so as to maintain the stable operation state of the pipe system.

[0105] The pumped storage unit fault detection method provided by the present invention also includes the following steps.

[0106] Based on the fault diagnosis report of the pumped storage unit, the treatment suggestions and maintenance measures of the pumped storage unit are determined.

[0107] Specifically, intelligent decision support systems can be set up to generate specific treatment suggestions and maintenance measures based on the identified failure modes and trends using built-in rule engines or expert systems. These rules or knowledge bases are pre-set by experienced engineers and maintenance experts. Treatment suggestions include immediate shutdown for maintenance, adjustment of operating parameters, replacement of damaged parts, etc.

[0108] As an example, the failure modes and / or failure development trends of the pumped storage unit include: the temperature increase trend of the bearing part 2, which specifically refers to the data analysis showing that the temperature of the bearing part 2 has gradually increased in the past few days. The abnormal vibration frequency mode of the rotating part 1, which specifically refers to the detection that the vibration frequency of the rotating part 1 regularly exceeds the normal range. The abnormal flow fluctuation mode of the water guide part 3, which specifically refers to the identification that the flow fluctuation of the water guide part 3 exceeds the normal threshold.

[0109] Detailed fault handling and maintenance recommendations are generated based on the identified fault modes and predicted fault development trends. These recommendations include specific operating steps, required tools and materials, and estimated downtime. The system also prioritizes maintenance recommendations based on the severity and urgency of the fault. For example, when the temperature of the bearing part 2 rises, the system recommends immediate shutdown to check the lubrication system, add or replace the lubricant, and check whether the bearing part 2 is damaged. When the vibration frequency of the rotating part 1 is abnormal, the system recommends removing weight or adding counterweights to the rotor, and checking and adjusting the automatic dynamic balancing system. When the flow of the water guide part 3 fluctuates, the system recommends checking the integrity of the water guide part 3, adjusting the flow control system, and replacing or repairing the flow regulating equipment.

[0110] A rule engine is a software system that uses predefined logical rules to simulate the decision-making process. These rules are usually developed by domain experts (such as mechanical engineers, maintenance technicians, etc.) based on years of experience and industry standards.

[0111] The functions of the rule engine include the following.

[0112] Rule storage: All decision rules are stored in a database or similar storage system. These rules are usually in the form of "if-then" statements, for example, "if the temperature of the bearing 2 exceeds the set threshold, then it is recommended to shut down for inspection."

[0113] Rule matching: When fault data is received from the data integration analysis module, the rule engine checks which rules match the current situation. This process involves comparing the input data (such as temperature, vibration, etc.) with the rule conditions.

[0114] Execute actions: Once a matching rule is found, the rule engine will execute the corresponding action. These actions may be to generate maintenance recommendations, send alarms, or adjust the operating parameters of the unit.

[0115] An expert system is a computer system that simulates the decision-making ability of human experts and uses a knowledge base and inference engine to solve complex problems. In the context of fault detection and maintenance, the workflow of an expert system is as follows.

[0116] Knowledge base: contains a large amount of professional knowledge about unit operation, fault diagnosis and maintenance strategies. This knowledge may exist in the form of rules, facts and relationships.

[0117] Inference engine: It is the core of the expert system and is responsible for deriving solutions to specific problems from the knowledge base. Inference engines usually use logical reasoning methods, such as forward reasoning (deriving new facts from known facts) or backward reasoning (finding facts that meet the conditions from the expected results).

[0118] User interface: allows the system to interact with the user. The user can enter fault data through this interface, and the system will provide maintenance suggestions.

[0119] The pumped-storage unit fault detection device provided by the present invention is described below. The pumped-storage unit fault detection device described below and the pumped-storage unit fault detection method described above can be referenced to each other.

[0120] See also Figure 3 As shown, the pumped storage unit fault detection device provided by the embodiment of the present invention includes a rotating part detection module, a bearing part detection module, a water guiding part detection module, a sealing part detection module, an embedded part detection module and a data integration and analysis module. The rotating part detection module, the bearing part detection module, the water guiding part detection module, the sealing part detection module and the embedded part detection module are all communicatively connected with the data integration and analysis module.

[0121] The rotating part detection module, the bearing part detection module, the water guiding part detection module, the sealing part detection module and the embedded part detection module are respectively used to obtain the rotating part fault signal, the bearing part fault signal, the water guiding part fault signal, the sealing part fault signal and the embedded part fault signal of the pumped-storage unit when the unit is operating under power generation conditions, pumping conditions, power generation phase adjustment conditions and water pump phase adjustment conditions.

[0122] The data integration and analysis module is used to: extract characteristic parameters of the rotating part fault signal, the bearing part fault signal, the water guide part fault signal, the sealing part fault signal and the embedded part fault signal; fuse the characteristic parameters of the rotating part fault signal, the bearing part fault signal, the water guide part fault signal, the sealing part fault signal and the embedded part fault signal to determine the fused characteristic parameters; identify the fault mode of the pumped-storage unit based on the fused characteristic parameters, determine the fault mode of the pumped-storage unit and the confidence level corresponding to the fault mode; based on the fault mode of the pumped-storage unit, the characteristic parameters corresponding to the fault mode, and the confidence level corresponding to the preset positive The fault severity level is determined by evaluating the degree of deviation from the normal operating parameter range, evaluating the severity of the identified fault mode, and determining the severity level of the fault; based on historical fault signals, the fault mode of the pumped-storage unit and its confidence level, and the fault severity level of the pumped-storage unit, predicting the fault development trend, and determining the predicted fault development trend and critical time point; based on the specific fault mode of the pumped-storage unit and its confidence level, the fault severity level, the predicted fault development trend and critical time point, generating a fault diagnosis report, which includes the fault mode of the pumped-storage unit, the fault severity level and the fault development trend.

[0123] The pumped storage unit fault detection device provided by the present invention can obtain the fault signals corresponding to the rotating part 1, the bearing part 2, the water guiding part 3, the sealing part 4 and the embedded part 5 of the pumped storage unit when the pumped storage unit is operating in the power generation condition, the pumping condition, the power generation phase adjustment condition and the water pump phase adjustment condition through the rotating part detection module, the bearing part detection module, the water guiding part detection module, the sealing part detection module and the embedded part detection module, respectively, and comprehensively analyze the above fault signals through the data integration analysis module to identify the fault mode and / or fault development trend of the pumped storage unit. Similarly, the defects of the existing pumped storage unit fault detection method that the data collection is relatively one-sided, the faults caused by different working conditions are not easy to be exposed, and the fault detection accuracy is low are solved.

[0124] As an example, the failure type of the pumped storage unit is bearing failure, the failure mode is pitting corrosion on the inner ring of the bearing, and the possible causes of the failure are insufficient lubrication, overload operation, material fatigue, etc.

[0125] The following describes the specific processing process of the data integration analysis module.

[0126] 1. Data reception and preprocessing

[0127] Specifically, the specific processing process of data reception and preprocessing includes the following steps.

[0128] In the data receiving step, original fault signals are received from each detection module (rotating part 1, bearing part 2, water guide part 3, sealing part 4 and embedded part 5). These original fault signals may include various types of data such as vibration, temperature, pressure, flow, etc.

[0129] In the signal classification step, the received signals are classified according to their sources and types. For example, vibration signals, temperature signals, pressure signals, etc. are classified separately. The output of this step is the classified signal data set, which is prepared for subsequent targeted processing.

[0130] In the filtering step, according to the characteristics of each type of signal, an appropriate filter (such as a low-pass, high-pass or band-pass filter) is selected for processing. This step directly uses the output of the signal classification step to remove irrelevant frequency components in the signal. The filtered signal will be used as the input of the next denoising step.

[0131] In the denoising step, the filtered signal is denoised. Depending on the signal characteristics, a suitable denoising method is selected, such as wavelet transform or empirical mode decomposition (EMD). This step directly depends on the output of the filtering process and aims to further improve the signal-to-noise ratio of the signal. The denoised signal will be used in the subsequent normalization process.

[0132] In the normalization step, the denoised signals of different dimensions are unified to the same scale. This step uses the output of the denoising step to ensure that signals from different sources and types can be fairly compared in subsequent analysis. The methods used include min-max normalization or Z-score normalization.

[0133] In the quality check step, the normalized signal is quality checked to ensure that the quality of the pre-processed signal meets the requirements of subsequent analysis. This step uses the output of the normalization processing step to check the integrity, continuity and validity of the signal.

[0134] In the data integration step, various preprocessed signals that have passed quality inspection are integrated into a structured data set. This is the last step of the preprocessing process, which integrates the results of all previous steps to provide high-quality input data for subsequent feature extraction and analysis.

[0135] Example 1: For the vibration signal of the bearing part 2, first use a bandpass filter (e.g., 500Hz-5kHz) to remove low-frequency interference and high-frequency noise, then apply the wavelet threshold denoising method to reduce random noise, and finally normalize the signal amplitude to the range of 0-1. Example 2: For the pressure signal of the sealing part 4, use a moving average filter to remove instantaneous fluctuations, and then use the Z-score normalization method for normalization. Example 3: For the flow signal of the water guide part 3, first use a median filter to remove outliers, then use Fourier transform to filter out power frequency interference, and finally perform maximum and minimum value normalization.

[0136] 2. Feature extraction

[0137] Specifically, the feature extraction process includes the following steps.

[0138] In the data loading step, the preprocessed signal data is loaded. These data are the output results of the data preprocessing in the previous stage, including various types of signals that have been filtered, denoised, and normalized. The output of this step is high-quality signal data ready for analysis, which provides the basic input for all subsequent feature extraction steps.

[0139] In the time domain feature extraction step, the loaded signal data is analyzed in the time domain to calculate statistical features. This step directly uses the output of the data loading step to calculate features such as mean, standard deviation, peak, RMS value, kurtosis, and skewness. The output of the time domain feature extraction will be a set of parameters that describe the time domain characteristics of the signal, which will form the final feature set together with the features of the subsequent steps.

[0140] In the frequency domain analysis step, the time domain signal is transformed into the frequency domain using the fast Fourier transform (FFT), and then the frequency domain features are extracted based on the transformation results. This step still directly uses the output of the data loading step to convert the time domain signal into a frequency domain representation and calculate features such as the main frequency, frequency center, frequency band energy, and power spectrum density. The output of the frequency domain feature extraction will be a set of parameters that describe the frequency domain characteristics of the signal, which will form the final feature set together with the time domain features and subsequent time-frequency domain features.

[0141] In the time-frequency domain analysis step, the original signal is transformed into the time-frequency domain. Short-time Fourier transform (STFT) or wavelet transform can be used, and then the time-frequency domain features are extracted based on the transformation results. This step again uses the output of the data loading step to convert the signal into a time-frequency domain representation and calculate features such as wavelet energy, wavelet entropy, and time-frequency spectrum moment. The output of the time-frequency domain feature extraction will be a set of parameters that describe the time-frequency characteristics of the signal, which will form the final feature set together with the previously extracted time-domain and frequency-domain features.

[0142] In the feature integration and optimization step, time domain features, frequency domain features, and time-frequency domain features are integrated to form a preliminary feature set. If the number of features is large, feature selection can be performed to select the most representative features through methods such as correlation analysis or principal component analysis. Finally, the selected features are standardized to ensure that features of different dimensions can be fairly compared in subsequent analysis. This step combines the outputs of all previous feature extraction steps and ultimately generates an optimized, standardized feature vector or feature matrix, which provides a comprehensive and effective feature representation for subsequent fault diagnosis and status assessment.

[0143] Example 1: For the vibration signal of the rotating part 1, extract the time domain features such as the root mean square value, peak factor, margin factor, frequency domain features such as characteristic frequency amplitude, frequency band energy ratio, and time-frequency domain features such as wavelet packet energy. Example 2: For the acoustic emission signal of the bearing part 2, calculate the time domain statistics such as kurtosis and skewness, frequency domain features such as power spectral density, and time-frequency domain features such as wavelet entropy. Example 3: For the temperature signal of the sealing part 4, extract the time domain features such as mean and variance, frequency domain features such as main frequency component, and time-frequency domain features such as Hilbert spectrum.

[0144] 3. Multi-source data fusion

[0145] The input of this step is multi-source feature data from different detection modules. The processing involves data alignment, feature selection and information fusion. Data alignment ensures that data from different sources are synchronized in time; feature selection is used to remove redundant and irrelevant features; information fusion uses methods such as Dempster-Shafer evidence theory and fuzzy integral to integrate multi-source information. The processing is based on the correlation and complementarity between different component faults. The output result is a fused and more comprehensive description of the fault characteristics.

[0146] Specifically, the specific processing process of multi-source data fusion includes the following steps: In the data alignment step, data from different sources are synchronized in time and the sampling rate is unified. The output is a multi-source dataset that is fully aligned in time. For example, all data is adjusted to 10 sampling points per second. This time-aligned dataset provides the basis for subsequent feature selection, so that features from different sources can be compared and analyzed on the same time scale.

[0147] In the feature selection step, the most relevant and discriminative features are selected from the aligned data using methods such as correlation analysis, principal component analysis (PCA), and information gain evaluation. The output is a reduced feature set containing the most important fault indicators. For example, key features such as the peak frequency of bearing vibration and the rate of change of stator temperature may be selected. These selected features will be standardized in the next step in preparation for subsequent fusion.

[0148] In the data normalization step, the selected features are normalized so that data of different dimensions can be compared on the same scale. The output is a standardized feature dataset. For example, all feature values ​​are converted to the range of 0-1. This standardized dataset is directly used for the next step of information fusion, ensuring that different types of data can participate in the fusion process fairly.

[0149] In the information fusion step, select an appropriate fusion method (such as Dempster-Shafer evidence theory or fuzzy integral) and assign weights to information from different sources. Using the selected fusion method and weights, perform specific fusion calculations on the standardized multi-source data. The output is a comprehensive fault indicator or probability distribution, which represents the fault status assessment after comprehensive consideration of all information.

[0150] In the result interpretation and output generation step, the fused results are converted into understandable fault feature descriptions and the final multi-source data fusion report is generated. The output is a complete fault diagnosis report, including possible fault types, severity and credibility, which can be directly used for decision support and subsequent maintenance plan formulation.

[0151] Example 1: The vibration characteristics of the rotating part 1 and the bearing part 2 are integrated, and the principal component analysis (PCA) is used for feature selection. Then, the Dempster-Shafer evidence theory is applied to perform information fusion to obtain a more reliable shaft fault judgment. Example 2: Combine the pressure fluctuation of the water guide part 3 and the leakage of the sealing part 4, and use the fuzzy integral method to fuse information to improve the accuracy of hydraulic system fault diagnosis. Example 3: The stress data of the embedded part 5 and the displacement data of the rotating part are integrated, and the Kalman filter is used for data fusion to more accurately evaluate the overall operating status of the unit.

[0152] 4. Failure mode identification

[0153] The input of this step is the fused feature data. The processing process includes feature matching and pattern classification. Feature matching compares the current feature with the pre-established fault mode library; pattern classification uses machine learning algorithms such as support vector machine (SVM), random forest, etc. to distinguish the fault type. The processing is based on the characteristic performance of different fault modes. The output result is the specific fault mode identified.

[0154] Specifically, the processing of fault mode identification includes the following steps.

[0155] In the feature data loading step, the fused feature data is loaded. These data are the output results of the multi-source data fusion in the previous stage, and contain the comprehensive feature information of each part of the equipment. The output of this step is a standardized feature vector or matrix, which serves as the basic input for all subsequent steps.

[0156] In the fault mode library preparation step, load the pre-established fault mode library. This library contains the characteristic patterns of various known fault types. The quality and completeness of the fault mode library directly affect the accuracy of subsequent matching and classification. The output of this step is structured fault mode data, which will be used in the next step of feature matching.

[0157] In the feature matching step, the loaded current feature data is compared with the patterns in the fault mode library. This step uses the feature data in the feature data loading step and the fault mode library in the fault mode library preparation step to find the known fault mode that is closest to the current feature by calculating similarity or distance. The matching result will be used as one of the inputs for the next machine learning classification to provide a preliminary judgment of the fault type.

[0158] In the machine learning model loading step, load pre-trained machine learning models, such as support vector machines (SVMs) or random forests. These models are trained based on historical data and known fault cases. The loaded models will be used in the next step to accurately classify fault types.

[0159] In the fault type classification step, the feature data is classified using the loaded machine learning model. This step uses the feature data in the feature data loading step and the matching results in the feature matching step as input, and uses the model loaded in the machine learning model loading step to identify the fault type. The classification results include the fault type and its probability, which will serve as the main component of the final fault pattern recognition result.

[0160] In the result synthesis and verification, the results of feature matching and machine learning classification are integrated to obtain the final fault mode recognition conclusion. This step can also include cross-validation of the results, such as comparing the consistency of the results of different classification methods, or comparing with historical cases. The output of this step is a verified and highly credible fault mode recognition result.

[0161] In the uncertainty assessment step, the uncertainty of the recognition results is assessed. Based on factors such as the confidence of the classification and the similarity of the matching, the credibility assessment of the fault pattern recognition results is given. This provides important reference information for subsequent decision-making.

[0162] In the result output step, the final fault mode identification report is generated. The report includes the specific fault modes identified, the corresponding confidence levels, possible alternative fault modes, and uncertainty assessment results. This comprehensive report will serve as an important basis for subsequent fault diagnosis and maintenance decisions.

[0163] Example 1: Using the SVM algorithm, based on the bearing vibration characteristics (such as characteristic frequency amplitude, envelope spectrum, etc.), the bearing fault is classified into inner ring fault, outer ring fault, rolling element fault or cage fault. Example 2: Using deep learning methods such as convolutional neural network (CNN), the pressure fluctuation spectrum of the water guide 3 is analyzed to identify fault modes such as cavitation, water hammer or flow channel blockage. Example 3: Using the decision tree algorithm, combined with the rotor vibration characteristics and the axis trajectory, it is determined whether there are fault modes such as misalignment, imbalance or shaft bending.

[0164] 5. Fault severity assessment

[0165] Specifically, the process of fault severity assessment includes the following steps.

[0166] In the data input step, the identified fault modes and characteristic parameters related to the fault modes are loaded. These data are the output results of the fault mode identification in the previous stage, including the fault type and key characteristic parameters. The output of this step is structured fault information and characteristic data, which serve as the basic input for all subsequent evaluation steps.

[0167] In the normal value benchmark setting step, the normal operating range or benchmark value is set for each key characteristic parameter based on the technical specifications and historical operating data provided by the equipment manufacturer. The output of this step is a set of standardized reference values, which will be used to calculate the degree of deviation in the deviation calculation step.

[0168] In the deviation calculation step, the degree of deviation between the characteristic parameters related to the failure mode and the normal value benchmark set in the normal value benchmark setting step is calculated. This step directly uses the characteristic parameters related to the failure mode in the data input step and the benchmark value in the normal value benchmark setting step to calculate the degree of deviation of each key characteristic parameter, that is, the deviation degree. The output is a set of quantitative deviation indicators, which will be used for subsequent severity quantification.

[0169] In the severity quantification step, the severity of the fault is quantified based on the deviations calculated in the deviation calculation step. This step involves a weighted combination of the deviations of different parameters, where the weights can be based on the importance or sensitivity of the parameters. The output is a composite severity score that provides the basis for the grading step.

[0170] In the evaluation rule loading step, predefined severity evaluation rules are loaded. These rules can come from expert experience or statistical analysis based on historical data, and define the levels corresponding to different severity scores. The output of the rules will be used for level classification in the level classification step.

[0171] In the grading step, the severity scores obtained in the severity quantification step are mapped to specific severity levels using the evaluation rules loaded in the evaluation rule loading step. This step can use fuzzy logic methods to handle the uncertainty in severity assessment. The output is a preliminary fault severity level.

[0172] In the expert system verification step, the severity levels initially classified in the classification step are verified and adjusted using the expert system or knowledge base. This step takes into account factors such as the fault type, equipment importance, and operating environment, and may fine-tune the preliminary levels. The output is the verified final severity level.

[0173] In the uncertainty assessment step, uncertainty analysis is performed on the severity assessment results of the expert system verification step. This can include sensitivity analysis of key parameter measurement errors, or probability distribution calculation based on fuzzy set theory. The output is the severity assessment result with uncertainty indicators.

[0174] In the result output step, the final fault severity assessment report is generated. The report includes the fault type, severity level, deviation of key parameters, uncertainty assessment results of the uncertainty assessment step, and possible development trend prediction. This comprehensive report will serve as an important basis for subsequent maintenance decisions and resource allocation.

[0175] Example 1: For bearing inner ring faults, based on the increase in vibration amplitude, the fault severity is divided into three levels: slight (amplitude increase <50%), medium (amplitude 50%-100%) and severe (amplitude>100%). Example 2: For cavitation faults in water guides, the fuzzy logic system is used to divide the fault severity into five levels: negligible, slight, medium, severe and dangerous, taking into account the intensity and duration of pressure pulsation. Example 3: For sealing system leakage, based on the change trend and cumulative amount of leakage rate, the expert system method is used to evaluate the severity of the fault and give corresponding treatment suggestions.

[0176] In one example, using the characteristic parameters before fusion, the fault severity assessment process will first analyze each component (such as the rotating part, bearing part, etc.) separately. For each component, compare its characteristic parameters with the preset normal operating parameter range, and calculate the degree of deviation of the characteristic parameters of each component. Then, according to the identified fault mode, determine the importance weight of the deviation of the characteristic parameters of each component. Consider the degree of deviation and weight of each component comprehensively to evaluate the overall fault severity. Finally, based on the predefined evaluation criteria, map the comprehensive evaluation results to a specific fault severity level. For the "vibration and seal leakage caused by water-guided bearing wear" fault mode, this embodiment considers the characteristic parameters of four related components: the vibration amplitude of the rotating part shaft (45μm, normal range 0-30μm), the bearing temperature of the bearing part (75°C, normal range 40-70°C), the water-guide clearance of the water-guide part (0.65mm, normal range 0.2-0.5mm) and the leakage rate of the sealing part (3.5L / min, normal range 0-2L / min). First, the deviation of each component parameter is calculated: 50% for the rotating part, 7.14% for the bearing part, 30% for the water guide part, and 75% for the sealing part. Then, the weights are determined according to the fault mode characteristics: 0.3 for the rotating part, 0.2 for the bearing part, 0.2 for the water guide part, and 0.3 for the sealing part. The weighted average deviation is calculated comprehensively: 0.3 × 50% + 0.2 × 7.14% + 0.2 × 30% + 0.3 × 75% = 44.43%. Based on the predefined criteria (<20% minor, 20-40% moderate, >40% severe), the fault severity is evaluated as "severe". This method retains the specific deviation of each component, which is conducive to more detailed fault analysis and location.

[0177] In another example, the fault severity level assessment process will directly use the fused feature parameter set. These fused feature parameters are compared with the preset normal operating parameter range to calculate the overall deviation of the fused feature parameters. Next, the importance weights of different fused features are determined based on the identified fault mode. The overall fault severity is assessed by comprehensively considering the deviation and weight of the fused feature parameters. Finally, the assessment results are mapped to specific severity levels based on the predefined assessment criteria. For the same "vibration and seal leakage caused by water-guided bearing wear" fault mode, this embodiment designs a comprehensive health index (CHI) to fuse the parameters of all relevant components. The normal range of CHI is 0-100, the higher the better, and the calculation formula is: CHI = 100 - (30 × deviation of the rotating part + 20 × deviation of the bearing part + 20 × deviation of the water-guide part + 30 × deviation of the seal part). Using the previous data, we calculated: CHI = 100 - (30 × 0.5 + 20 × 0.0714 + 20 × 0.3 + 30 × 0.75) = 55.57. Comparing this CHI value with the preset normal range (CHI>80 is normal), we calculated the deviation: (80-55.57) / 80 = 30.54%. Since CHI has incorporated all relevant features, its weight is 1. According to the predefined criteria (<20% mild, 20-40% moderate, >40% severe), the fault severity is assessed as "moderate".

[0178] In practical applications, the above two methods can be combined to obtain a comprehensive and accurate fault severity level assessment.

[0179] 6. Fault development trend analysis

[0180] Specifically, the processing of fault development trend analysis includes the following steps.

[0181] In the data collection and preprocessing step, historical fault data and current fault status information (such as specific fault modes and their confidence levels, fault severity levels) are collected. The data is cleaned, standardized, and time-aligned. The output of this step is a structured, time-series data set that provides the basis for all subsequent analysis steps.

[0182] In the time series feature extraction step, key time series features such as trends, periodicity, seasonality, etc. are extracted from the preprocessed data. This step directly uses the output of the data collection and preprocessing steps to generate a set of characteristic parameters that describe the fault development mode, providing a basis for subsequent model selection and training.

[0183] In the model selection step, based on the results of time series feature extraction, select an appropriate analysis model. You can choose to include ARIMA model (suitable for linear trends) or LSTM network (suitable for complex nonlinear trends). The results of model selection will directly affect the next step of model training process.

[0184] In the model training step, the selected model is trained using historical data. This step uses the output of the data collection and preprocessing step as training data and the results of the model selection step as the model architecture. The output of the training process is a predictive model that captures the failure development pattern.

[0185] In the model validation step, a portion of historical data that was not involved in the training is used to validate the trained model and evaluate the model's prediction accuracy. This step uses the output of model training and some reserved validation data. The validation results will determine whether it is necessary to return to the model selection or model training step for adjustment.

[0186] In the trend prediction step, the validated model is used to predict future fault development trends. This step uses the final model from the model validation step and the current fault status data as input, and the output is a series of predicted values ​​representing future fault states.

[0187] In the critical point analysis step, based on the results of trend prediction, combined with the safety threshold or performance requirements of the equipment, possible critical time points are identified. This step uses the output of the trend prediction step, combined with predefined threshold standards, to generate one or more time points indicating that the equipment may reach a dangerous or failure state.

[0188] In the uncertainty assessment step, uncertainty analysis is performed on the forecast results, which can include confidence interval calculations or Monte Carlo simulations. This step uses the results of trend forecasting and critical point analysis to output forecast results with uncertainty ranges.

[0189] In the result interpretation and report generation step, the results of the previous steps are combined to generate a detailed fault development trend analysis report. The report includes the predicted fault development trend, possible critical time points, the predicted uncertainty range, and maintenance recommendations based on this information. This step integrates the outputs of all previous steps to provide comprehensive information support for decision makers.

[0190] Example 1: Apply the ARIMA model to the historical data of bearing vibration amplitude to predict the change trend of vibration amplitude in the next month and estimate the time to reach the warning value. Example 2: Use the LSTM network to analyze the historical leakage data of the sealing system to predict the growth trend of future leakage and provide a basis for maintenance planning. Example 3: Based on the frequency spectrum change history of pressure fluctuations in the water guide part 3, the support vector regression (SVR) method is applied to predict the development trend of cavitation intensity and estimate the time point that may cause serious damage.

[0191] Example 4: For the fault mode of increased vibration caused by water-guide bearing wear, in the data collection and preprocessing step, the vibration amplitude data of the water-guide bearing in the past two years are collected, with a total of 17,520 hourly measurement points (2 years × 365 days × 24 hours), ranging from 20μm to 85μm. First, data cleaning is performed to remove outliers and process missing data. Then, the hourly data is downsampled to daily averages, resulting in 730 daily average data points. These data are time series normalized. At the same time, the current fault mode (water-guide bearing wear) and its confidence (85%), as well as the latest fault severity level ("serious") are sorted. The output of this step is a structured, time-aligned daily average data set, which lays a more reliable foundation for subsequent analysis.

[0192] In the time series feature extraction step, the 730 pre-processed daily average vibration amplitude data were analyzed to extract key features. Through observation, it was found that the data showed an obvious long-term upward trend and weak seasonal fluctuations (possibly related to seasonal load changes). Using time series decomposition technology, the data was decomposed into trend, seasonal and residual components. After calculating the first-order difference, the data showed good stability. These features provide an important basis for the subsequent model selection, indicating that the SARIMA (seasonal ARIMA) model is suitable.

[0193] In the model selection step, based on the results of time series feature analysis, the SARIMA model is selected for trend forecasting. Considering the characteristics of the data (obvious trend, weak seasonality, and stabilization after the first-order difference), the model that may be suitable is initially determined to be SARIMA(p,d,q)(P,D,Q)m, where p, d, q, P, D, Q, and m need to be determined through further analysis. Using the grid search method, combined with indicators such as AIC (Akaike Information Criterion) and BIC (Bayesian Information Criterion), it is finally determined that the SARIMA(1,1,1)(1,1,1)7 model is most suitable for this data set, where possible weekly seasonality is taken into account.

[0194] In the model training step, the first 584 data points (about 1 year and 7 months of data) are used to train the SARIMA (1,1,1) (1,1,1) 7 model. During the training process, the maximum likelihood estimation method is used to determine the model parameters. After training, the model is able to capture the long-term upward trend, seasonal fluctuations, and short-term fluctuation characteristics of the data. The output of this step is a trained SARIMA model, ready for subsequent forecasting tasks.

[0195] In the model validation step: the remaining 146 data points (about 5 months of data) are used to validate the trained model. The root mean square error (RMSE), mean absolute percentage error (MAPE), and coefficient of determination (R²) between the predicted and actual values ​​are calculated. The validation results show that the model has good predictive ability, with MAPE below 3% and R² above 0.95, confirming that the model is suitable for medium- and long-term trend forecasting.

[0196] In the trend prediction step, the verified SARIMA (1,1,1) (1,1,1)7 model is used to predict the average daily vibration amplitude for the next 90 days based on all 730 historical data points. The prediction results show that the vibration amplitude will continue to rise, but the rate of increase may slow down slightly. For example, the predicted vibration amplitude after 90 days is 92μm, while taking seasonal fluctuations into account. The output of this step not only shows the future trend of fault development, but also reflects the possible seasonal impact, providing a more comprehensive basis for subsequent decision-making.

[0197] In the critical point analysis step, according to the equipment's safe operation standards, the safety threshold of the vibration amplitude is determined to be 85μm per day and 90μm for the instantaneous maximum value. Comparing this threshold with the predicted trend, it is found that on the 75th day, the predicted daily average (85.5μm) will exceed the daily average safety threshold for the first time. Taking into account intraday fluctuations, the instantaneous maximum value may exceed 90μm around the 60th day. Therefore, the 60th day is determined as the critical time point. This information is crucial for formulating maintenance plans and provides a more accurate time window.

[0198] In the uncertainty assessment step, the uncertainty analysis of the prediction results was performed considering the 85% confidence level of the failure mode and the inherent uncertainty of the model prediction. Using the Monte Carlo simulation method, 10,000 simulations were generated to obtain the 95% confidence interval of the predicted value. For example, the 95% confidence interval of the predicted daily mean value of 88μm on the 60th day is 83-93μm. At the same time, considering the impact of seasonal factors on uncertainty, it was found that the uncertainty of the prediction increased slightly in the season with heavier loads. This step increases the reliability of the prediction results and provides a more comprehensive risk assessment basis for decision making.

[0199] In the result interpretation and report generation step, all analysis results are integrated to generate a detailed fault development trend analysis report. The report includes long-term vibration amplitude rising trend analysis, seasonal fluctuation pattern recognition, warnings that may exceed the safety threshold after 60 days, predicted uncertainty ranges, and maintenance recommendations based on the current "serious" level fault status. It is recommended to complete the bearing overhaul or replacement within 50 days, and increase the daily monitoring frequency to once every 4 hours from now on. At the same time, the report also includes load management recommendations to slow down the wear rate. This comprehensive report provides management with a more accurate, comprehensive and actionable basis for maintenance decisions, which helps to optimize maintenance plans and resource allocation.

[0200] 7. Generate a diagnostic report

[0201] The input of this step is the results of all the previous analysis steps, including fault mode identification results, fault severity assessment, fault development trend analysis, and other relevant operating parameters and historical data. The processing process involves information integration, importance ranking, and report formatting. Information integration summarizes the results of various analyses to ensure the comprehensiveness of the report; importance ranking is based on the severity, urgency, and potential impact of the fault, highlighting key information; report formatting generates easy-to-understand documents according to predefined templates, including text descriptions, data tables, and visual charts. The processing is based on user needs, industry standards, and best practices. The output result is a comprehensive, clear, and targeted fault diagnosis report.

[0202] Example 1: Generate a diagnostic report on a bearing fault. The report first summarizes the detected inner race fault of the bearing, indicating that its current severity is "medium" (vibration amplitude increased by 75%). It then lists the fault characteristics in detail, including key parameters such as characteristic frequency and vibration amplitude. The report also contains a trend graph showing the change in vibration amplitude over the past week and the predicted trend for the next two weeks. Finally, the report provides maintenance recommendations, recommending that the bearing be replaced at the next planned shutdown (estimated in 10 days), and lists the required spare parts list and replacement steps.

[0203] Example 2: Generate a diagnostic report for a cavitation problem in a water distribution system. The report first summarizes the current status of the cavitation fault and rates its severity as "medium". The report then analyzes the pressure fluctuation data in detail, including spectrum analysis graphs and time series graphs. The trend analysis section uses color-coded charts to intuitively display the historical changes and predicted trends in cavitation intensity. The report also includes a comparative analysis showing the similarities and differences between the current cavitation situation and historical cases. Finally, the report proposes short-term mitigation measures (such as adjusting operating conditions) and long-term solutions (such as considering material upgrades or structural optimization at the next overhaul).

[0204] Example 3: Generate a comprehensive unit health report. The report first provides an overall score that reflects the overall health of the unit. The report then details the status of each subsystem (such as the rotating part, bearing part, water guide part, etc.), each of which contains a summary of the current status, a list of key parameters, and a trend chart. For any abnormalities or potential problems detected, the report will provide a detailed description, including possible causes, development trends, and recommended treatment methods. The report also contains a maintenance plan recommendation, which proposes an optimized inspection and maintenance schedule based on the condition and predicted trends of each component. Finally, the report is accompanied by a dashboard-style visual summary, using a red, yellow, and green color system to intuitively display the health status of each component.

[0205] In a further embodiment, the specific processing process of the data integration analysis module may further include the following steps after the above steps 1 to 6: 7. Cross Validation

[0206] The input of this step is the fault diagnosis results of each component. The processing process includes result consistency check and conflict resolution. The consistency check compares the relevant fault information between different components; conflict resolution handles inconsistent diagnosis results by setting priorities or using a voting mechanism. The processing basis is the relationship and influence between the various components of the unit. The output result is a cross-validated and more reliable fault diagnosis conclusion.

[0207] Specifically, the cross-validation process includes the following steps.

[0208] In the result collection step, the preliminary fault diagnosis results of various components (such as rotating parts, bearing parts, water guide parts, sealing parts and embedded parts) are collected. These results include information such as fault type, severity and confidence.

[0209] In the consistency check step, the relevant fault information between different components is compared to check the consistency of the diagnostic results. For example, check whether the vibration abnormality of the bearing part is consistent with the imbalance diagnosis result of the rotating part. The output is a consistency report that identifies consistent and inconsistent diagnostic results.

[0210] In the conflict identification step, conflicting diagnostic results are identified based on the consistency check results. For example, if the water-conducting part diagnoses a water hammer phenomenon, but the rotating part does not detect a corresponding impact response, this constitutes a conflict.

[0211] In the conflict resolution step, the identified conflicts are resolved by setting priorities or using a voting mechanism. Priorities can be based on the importance of components or the reliability of sensors. The voting mechanism can consider the diagnosis results of the majority of components. The output is a unified diagnosis result after resolving the conflicts.

[0212] In the comprehensive evaluation step, the consistent diagnostic results and the conflict resolution results are considered to form the final fault diagnosis conclusion. This conclusion should reflect the overall health status of the unit, taking into account the interrelationships and impacts between the various components.

[0213] In the reliability assessment step, the reliability of the final fault diagnosis conclusion is evaluated, including the quantification of confidence or uncertainty.

[0214] In the output generation step, a cross-validated and more reliable fault diagnosis report is generated. This report should include a comprehensive fault diagnosis conclusion, diagnostic details of each component, conflict resolution instructions (if any), and overall reliability assessment.

[0215] Example 1: When rotor imbalance is detected, the vibration characteristics of the relevant bearings and the leakage of the shaft seal are checked at the same time to confirm whether there are consistent abnormal performances and improve the credibility of the diagnosis. Example 2: If water hammer is detected in the water-guiding part, but the corresponding impact response is not detected in the rotating part, the system will further analyze possible causes, such as sensor failure or signal transmission delay. Example 3: When the results of embedded part stress detection and sealing part leakage detection are inconsistent, the system will assign different weights to the two detection results based on historical reliability data and make a comprehensive judgment.

[0216] 8. Failure Cause Analysis

[0217] The specific steps of fault cause analysis are as follows.

[0218] In the data collation step, the cross-validated fault diagnosis results and related operating parameters are collated to form a structured data set. This includes information such as fault type, severity, and equipment operating status. The output of this step is the basis of the entire analysis process and will be used in all subsequent steps.

[0219] In the causal model construction step, based on the data set collated in the data collation step, a causal relationship model (such as a Bayesian network or decision tree) is constructed in combination with the working principle of the equipment and historical fault data. This model describes the relationship between fault symptoms, operating parameters, and potential causes, and will be used as a core tool in the cause inference step.

[0220] In the expert knowledge integration step, the data set from the data collation step is used as a reference, combined with the experience and knowledge of external experts, to establish a structured expert knowledge base, including a rule set or a case base. This knowledge base will play a key role in assisting inference in the cause inference step, verifying the results in the expert system verification step, and adjusting the probability estimate in the probability assessment step.

[0221] In the cause inference step, the causal model constructed in the causal model construction step is used to infer possible fault causes in combination with the current fault data in the data collation step and the expert knowledge base in the expert knowledge integration step. This step outputs a preliminary list of fault causes and their initial probability estimates, providing a basis for subsequent verification and evaluation.

[0222] In the expert system verification step, the expert knowledge base of the expert knowledge integration step is used to verify and correct the results of the cause inference step. This step ensures that the inference results are consistent with the expert experience and outputs a verified list of fault causes, which is directly used for the next step of probability assessment.

[0223] In the probability assessment step, based on the results of the expert system verification step, combined with the expert knowledge of the expert knowledge integration step, and with reference to the original data of the data collation step, the final probability assessment of each possible failure cause is performed. The output of this step is a list of failure causes with reliability estimates, which provides the core content for the final report.

[0224] In the result output step, the key information of all previous steps, especially the final probability assessment results of the probability assessment step, is summarized to generate a complete fault cause analysis report. This report includes a list of possible fault causes, probability estimates of each cause, and key evidence or reasoning processes supporting these conclusions, providing a direct basis for subsequent decision-making and maintenance actions.

[0225] Example 1: For bearing overheating failure, the system will analyze parameters such as lubricating oil temperature, viscosity, flow rate, etc., and use Bayesian network to infer possible causes in combination with vibration characteristics, such as insufficient lubrication, bearing wear, or excessive load. Example 2: For abnormal guide vane opening, the system will check the hydraulic system pressure, stress distribution of the guide vane mechanism, etc., and analyze possible causes of failure through decision trees, such as hydraulic oil leakage, mechanical jamming, or control system failure. Example 3: For abnormal increase in the temperature of the generator stator winding, the system will use the case reasoning system to match similar historical failures in combination with load conditions, cooling water parameters, etc., and analyze possible causes such as insulation aging and reduced cooling system efficiency.

[0226] 9. Generate a diagnostic report

[0227] The input to this step is the analysis results of all the previous steps. The processing process includes information integration, importance ranking and report formatting. Information integration summarizes the results of various analyses; importance ranking is based on the severity and urgency of the fault; and report formatting generates an easy-to-understand report according to a predefined template. The processing is based on user needs and industry standards. The output is a comprehensive fault diagnosis report.

[0228] Example 1: A fault diagnosis report is generated, indicating that the unit has a minor inner ring bearing fault, which is expected to develop into a moderate fault within the next two weeks. It is recommended to perform maintenance during the next planned shutdown. The report includes a vibration spectrum analysis diagram, a trend prediction curve, and recommended maintenance measures. Example 2: The report shows that there is moderate cavitation in the water guide system, which has affected efficiency but has not yet caused serious damage. It is predicted that it may develop into a serious level within 3 months. The report is accompanied by a pressure fluctuation analysis, material loss estimation, and repair plan recommendations. Example 3: The diagnostic report indicates that there is a potential insulation problem in the stator winding of the generator. Although the current impact is not significant, there is a trend of accelerated deterioration. The report includes a temperature distribution diagram, insulation resistance change trend, and recommended monitoring enhancement measures.

[0229] 10. Warning and suggestion generation

[0230] The generation of warnings and suggestions specifically includes the following steps.

[0231] In the information extraction step, key information is extracted from the diagnostic report, including fault type, severity, development trend, and impact scope. This information is the basic input for all subsequent steps.

[0232] In the risk assessment step, the risk level of each fault is assessed based on the output of the information extraction step. Each fault is assigned a risk score considering the severity of the fault and the predicted development trend. This risk assessment result will be directly used for the next step of prioritization.

[0233] In the prioritization step, the results of the risk assessment step are combined with the importance of the equipment and the difficulty of repair to prioritize each fault. The output of this step will determine the order and urgency of subsequent countermeasures.

[0234] In the countermeasure formulation step, according to the results of the priority classification, combined with the equipment characteristics and the current operation plan, a specific maintenance countermeasure is formulated for each high-priority fault. This step needs to refer to the fault details in the information extraction step to ensure the pertinence and feasibility of the countermeasure.

[0235] In the warning information generation step, based on the results of risk assessment and prioritization, warning information is generated. The warning information should include fault description, risk level and urgency. This step directly depends on the output of the previous steps.

[0236] In the maintenance proposal integration step, the results of the countermeasure formulation step are integrated into specific maintenance proposals. These proposals include short-term countermeasures and long-term solutions, taking into account the results of the prioritization.

[0237] In the report generation step, the outputs of the warning information generation and maintenance suggestion integration steps are summarized to generate the final warning and suggestion report. This report clearly shows the risk, priority, warning information and specific maintenance suggestions for each fault.

[0238] Example 1: For a tiny crack detected on the inner ring of a bearing, the system generates an early warning message suggesting increasing the frequency of vibration monitoring to once every four hours, and recommends replacing the bearing at the next planned shutdown, while providing a list of required spare parts and replacement steps. Example 2: For an abnormality in the turbine guide vane adjustment system, the system issues a yellow early warning, recommends a detailed inspection within 72 hours, and provides a list of professionals and tools that may be required. Example 3: For insulation problems in the generator stator winding, the system generates long-term monitoring recommendations, including increasing the frequency of partial discharge testing, installing online monitoring equipment, and arranging insulation updates at the next overhaul, while providing possible technical improvement plans.

[0239] 11. Self-learning and optimization

[0240] The input of this step is the historical diagnosis results and actual maintenance feedback. The processing process includes performance evaluation, parameter adjustment and knowledge update. Performance evaluation compares the difference between the diagnosis results and the actual situation; parameter adjustment optimizes the parameter settings of various algorithms; knowledge update expands and corrects the fault mode library. The processing is based on statistical analysis and machine learning theory. The output is the optimized diagnosis model and updated knowledge base.

[0241] Example 1: The system compares the bearing fault diagnosis results with the actual maintenance findings and finds that the judgment accuracy for certain low-speed operating conditions is low, and accordingly adjusts the feature extraction method and judgment threshold under the corresponding conditions. Example 2: By analyzing the guide vane system fault diagnosis records for the past year, the system discovered some new fault modes, automatically added these new fault modes to the fault library, and adjusted the corresponding diagnosis rules. Example 3: Based on the accuracy statistics of the generator stator temperature warning, the system adaptively adjusted the parameters of the temperature prediction model to improve the accuracy and timeliness of the warning. Specifically, the system found that the original linear prediction model did not work well when the load changed rapidly, so it introduced nonlinear factors and used the support vector regression (SVR) model to improve temperature prediction, thereby improving the accuracy of warnings under various conditions.

[0242] Specifically, the rotating part detection module is used to detect the balance state of the rotating part 1 in the pumped-storage unit, the bearing part detection module is used to detect the rotation state of the bearing part 2 in the pumped-storage unit and diagnose the fault, the water guiding part detection module is used to determine whether a fault occurs through the water flow characteristics of the water guiding part 3, the sealing part detection module is used to perform real-time pressure detection on the sealing part 4 in the pumped-storage unit, the buried part detection module is used to collect the water flow characteristics in the buried and pre-buried pipes, and the data integration and analysis module is used to obtain the fault diagnosis results corresponding to the rotating part detection module, the bearing part detection module, the water guiding part detection module, the sealing part detection module and the buried part detection module respectively, and perform comprehensive analysis to identify the fault mode and / or fault development trend of the pumped-storage unit.

[0243] Among them, the rotating part detection module is used to monitor the operating status of rotating parts in the pumped storage unit, such as turbine shafts and impellers. The rotating part detection module collects data such as rotation speed, vibration frequency, abnormal sound and temperature through sensors installed at key locations to monitor the operating status of the equipment in real time. If the sensor detects abnormal vibration or abnormal temperature rise, the system will immediately analyze the data to determine possible imbalance and wear problems, and issue an alarm in time, allowing the maintenance team to respond quickly and avoid more serious mechanical failures.

[0244] In addition, by capturing the sound signals generated by rotating parts (such as turbine shafts and impellers) during operation, these sound signals are converted into digital format, and then frequency analysis is performed through corresponding algorithms (such as fast Fourier transform FFT, short-time Fourier transform STFT, wavelet transform), and faults are identified by analyzing the spectrum of the sound signal. Under normal operating conditions, each rotating part has a specific sound spectrum. When a part has problems such as wear, cracks or looseness, it will cause a shift in the sound spectrum. The sound feature extraction unit can detect the offset value of these sound spectra and compare them with the set offset threshold to determine the presence of a fault.

[0245] Spectral shift refers specifically to changes in the frequency components of sound signals. During the operation of a pumped-storage unit, a normally operating rotor will generate a specific sound spectrum that contains sound signals with different frequency components. Spectral shift can manifest itself in the following ways: frequency drift, which refers specifically to the movement of certain frequency components to other frequency positions in the spectrum; amplitude change, which refers specifically to significant changes in the amplitude (intensity) of certain frequency components; the appearance of new frequency components, which refers specifically to the appearance of frequency components that do not exist in normal operation; and the disappearance of original frequency components, which refers specifically to the disappearance of frequency components that should exist in normal operation.

[0246] Determining whether the spectrum deviation reaches the fault level is achieved through the following steps: first, under normal operation, establish a baseline spectrum of the main shaft rotor in the rotating part 1 of the pumped storage unit; monitor the sound signal of the main shaft rotor in the rotating part 1 of the pumped storage unit in real time, and compare it with the baseline spectrum to obtain the spectrum deviation value; set the deviation threshold, including frequency drift, amplitude change, etc. If the monitored spectrum deviation exceeds the set threshold, it is considered that there may be a fault.

[0247] The bearing detection module is responsible for monitoring the operating status of the bearing 2 and evaluating the health status of the bearing 2 through infrared thermal imaging technology, vibration analysis and fiber optic sensing technology. The module can detect the temperature distribution, deformation and stress state, and vibration characteristics of the bearing 2. For example, if the temperature of the bearing 2 rises abnormally or the vibration mode changes, the system will analyze these changes to diagnose potential faults, such as insufficient lubrication or overload, so that maintenance measures can be taken in advance to prevent equipment failure.

[0248] The water guide detection module is used to monitor the dynamics of water flow, including flow rate, pressure and possible leakage. By installing a flow meter 22 and a pressure sensor 19 in the water guide 3, the module can collect key data in real time and analyze the data to detect whether there is any abnormality in the water guide 3, such as pipeline leakage or blockage. For example, if a sudden drop in flow is detected while the pressure remains unchanged, it may indicate a blockage in the pipeline, and the system will prompt inspection and maintenance.

[0249] The seal detection module is used to detect the integrity of the sealing system in the pumped storage unit. By monitoring the pressure changes in the sealing part, the module can detect signs of seal failure in a timely manner. For example, if the system detects that the pressure of the seal 4 continues to drop, it may be due to wear or damage of the sealing material. This timely fault detection allows the maintenance team to take quick action to replace the damaged seal and prevent possible leakage or other related problems.

[0250] The embedded detection module is used to monitor the status of embedded structures such as pipes and valves. The module uses water level and flow sensors to monitor the water dynamics in the pipe, and other sensors to detect possible structural deformation or damage. For example, if the water level rises abnormally and the flow rate decreases, it may indicate a blockage in the downstream pipe.

[0251] The data integration analysis module is used to integrate the data from all the above detection modules. This module uses advanced algorithms to comprehensively analyze the collected data to identify and predict the failure mode and failure development trend of the pumped storage unit. For example, if abnormal signals are detected from the rotating part 1 and the bearing part 2 at the same time, the module will analyze the correlation of these data, determine the root cause of the fault, and provide detailed diagnostic results and maintenance recommendations to the operator.

[0252] Specifically, the data integration and analysis module uses, for example, machine learning or data mining technology to conduct a comprehensive analysis of the data collected by each detection module. These data include vibration data and sound data of the rotating part 1, temperature data and vibration data of the bearing part 2, flow signal and pressure data of the water guide part 3, etc. By analyzing the patterns and trends of these data, the data integration and analysis module can identify the difference between normal operation and abnormal state, thereby predicting the possibility and time of fault occurrence, and diagnosing the specific type and cause of the fault that has occurred.

[0253] The fault diagnosis process of the data integration and analysis module specifically includes: receiving real-time data from each detection module; ensuring data quality and consistency through preprocessing, cleaning and integrating data; extracting key features associated with fault diagnosis from the preprocessed data; applying statistical analysis and machine learning algorithms to identify abnormal patterns and / or trends in the data; comparing the identified abnormal patterns with known fault patterns to determine the type and cause of the fault; and generating a detailed diagnostic report, including the fault type, cause, severity, and maintenance recommendations.

[0254] The processing process of the data integration analysis module includes the following steps.

[0255] Fault diagnosis results are collected from each detection module. These data include fault type, occurrence time, location, and severity. For example, the rotating part detection module reports that the bearing 2 is unbalanced, the water guide part detection module reports that the pressure drops abnormally, and the sealing part detection module reports that a potential leak is found.

[0256] Data preprocessing: During the data preprocessing phase, data cleaning, verification, and formatting are performed to ensure data quality and consistency in preparation for subsequent analysis. For example, abnormal data points generated by sensor failure in the bearing detection module are removed. The timestamp format of all modules is unified for time series analysis. The missing flow signal in the water guide detection module is filled in by using the average of the previous and next time points.

[0257] Fault pattern recognition uses statistical analysis and machine learning techniques to analyze the merged data and identify the fault patterns of the pumped storage unit. For example, cluster analysis found that multiple bearings 2 frequently overheated under similar operating conditions. Vibration analysis identified the frequent vibration patterns of rotating parts at specific speeds. Based on time series analysis, it was found that each time the pressure of the water guide 3 dropped, there was a short increase in flow, indicating a possible valve failure.

[0258] Perform trend analysis, which analyzes the time series data of fault occurrences, identifies fault occurrence trends, and the correlation between different faults. For example, analysis shows that the failure rate of bearing part 2 increases every summer, which may be related to the high temperature environment. It is detected that every time a large equipment is started, the leakage reports of sealing part 4 increase, which may be caused by the start-up vibration. The pressure anomaly of water guide part 3 is highly correlated with the flow anomaly of embedded part 5, indicating the overall fluidity problem of the system.

[0259] Through the above steps, the data integration analysis module can effectively identify and predict faults, provide scientific data support for the maintenance of pumped storage units, and thus improve the reliability and efficiency of the equipment.

[0260] According to some embodiments of the present invention, the pumped-storage unit fault detection device also includes: an operation recording module, which is used to record four operating conditions of the pumped-storage unit, namely, power generation condition, pumping condition, power generation phase adjustment condition, and water pump phase adjustment condition.

[0261] See also Figure 4 As shown, the rotating part detection module includes a rotor balancing unit, a sound feature extraction unit and a vibration energy analysis unit.

[0262] The rotor balancing unit is used to detect the balance state of the main shaft rotor in the rotating part 1 in real time based on the automatic dynamic balancing system. When the rotor is unbalanced, the mass distribution of the rotor is changed by removing weight or adding counterweight on the rotor to reduce the rotor vibration caused by the eccentric centrifugal force of the center of mass.

[0263] The sound feature extraction unit is used to obtain the sound signal generated when the main shaft rotor in the rotating part 1 of the pumped storage unit is running, perform frequency analysis on the sound signal according to the signal processing algorithm, and determine whether there is a fault in the rotor of the rotating part 1 according to the result of the frequency analysis and a fault identification model based on the sound features.

[0264] The vibration energy analysis unit is used to collect the amplitude, frequency, rotor speed and vibration direction of the rotor through a vibrometer or a vibration sensor 12, calculate the vibration energy of the rotor according to the vibration energy formula, obtain vibration energy data, and output the vibration energy data. Among them, the rotation speed of the rotor is expressed in revolutions per minute (RPM) or angular velocity (radians per second), which may include the linear velocity of the rotor (the distance the outer edge of the rotor moves per unit time) or the angular velocity (the angle of rotation of the rotor per unit time). The vibration direction refers to the specific direction of the vibration in space, which may be along different axes (X-axis, Y-axis, Z-axis in the spatial rectangular coordinate system) or in any direction between these axes. By determining the vibration direction, the source and characteristics of the vibration can be analyzed more accurately, so that the fault can be diagnosed and processed.

[0265] The specific processing process of the sound feature extraction unit is as follows.

[0266] The sound signal of the main shaft rotor in the rotating part 1 of the pumped storage unit is collected by the sensor during operation, and the features thereof are extracted, including frequency components, amplitude, time domain and frequency domain information, etc.

[0267] Signal processing algorithms (such as Fast Fourier Transform FFT, Short-time Fourier Transform STFT, and Wavelet Transform) are used to analyze sound signals and extract useful frequency features from them. The above algorithms convert sound signals from the time domain to the frequency domain to reveal their frequency components and energy distribution.

[0268] Perform fault pattern recognition. This includes: establishing a fault recognition model to classify and store the sound spectrum characteristics under normal operation and various fault conditions. The model can be built based on historical data, experimental data, and expert knowledge through machine learning, statistical analysis, or pattern recognition.

[0269] In actual operation, the sound signal of the main shaft rotor in the rotating part 1 of the pumped storage unit is collected and analyzed in real time, and the extracted frequency features are compared with the features in the fault identification model. If the monitored sound signal features match certain fault features in the model, it is determined that the main shaft rotor in the rotating part 1 of the pumped storage unit has a corresponding fault.

[0270] For example, when the main shaft rotor of a pumped storage unit is operating normally, it will produce a specific sound spectrum. This spectrum contains different frequency components and their amplitude distribution. When the rotor has faults such as wear, cracks or imbalance, the spectrum of the sound signal will shift (for example, the intensity of certain frequency components increases or decreases, new frequency components appear, etc.). By analyzing and comparing these changes, the fault identification model can determine whether there is a fault in the rotor and its type.

[0271] According to some embodiments of the present invention, the rotating part detection module further includes an energy recovery unit for utilizing mechanical vibration energy (such as Fig.10 The vibration energy converter in the 6) recovery method converts the vibration energy generated by the pumped storage unit during operation into electrical energy. This part of the electrical energy is converted using piezoelectric materials, electromagnetic induction and other technologies, and energy recovery is achieved by converting mechanical vibration into electrical energy, thereby improving the energy utilization efficiency of the system.

[0272] See also Fig.10As shown, specifically, the rotor balancing unit introduces an automatic dynamic balancing system. The frequency detection sensor of the automatic dynamic balancing system detects the balance state of the main shaft rotor in the rotating part 1 in real time, and determines whether the rotor is balanced by detecting the rotation speed or rotation frequency of the rotor. When the rotation speed or rotation frequency of the rotor deviates from the original frequency (the difference between the two) by more than the corresponding deviation degree or the preset deviation threshold, it is in an unbalanced state. When the unbalance is detected, the automatic dynamic balancing system removes weight or adds counterweight to the rotor to change the mass distribution of the rotor, so as to reduce the rotor vibration caused by the eccentric centrifugal force of the center of mass, so as to achieve the purpose of smooth operation of the main shaft rotor.

[0273] In this embodiment, the system sets a speed threshold and a rotation frequency threshold. When the speed or rotation frequency of the rotor deviates from the preset threshold, it is determined to be an unbalanced state and the corresponding corrective measures are initiated. The speed deviation refers to the difference between the actual speed of the rotor and its original speed. The rotation frequency deviation refers to the difference between the actual rotation frequency of the rotor and its original rotation frequency.

[0274] Specifically, the automatic dynamic balancing system includes a frequency monitoring sensor 7, a signal processing unit and a control unit.

[0275] The frequency monitoring sensor 7 is used to detect the rotation speed or frequency of the main shaft rotor in the rotating part 1 in real time, and obtain a detection signal to determine whether the rotor is in a balanced state.

[0276] The signal processing unit is used to process the speed or frequency signal collected from the frequency monitoring sensor 7, filter and analyze the signal, and identify the unbalanced state. The processing process specifically includes: using a high-speed data acquisition device to convert the analog signal output by the frequency monitoring sensor 7 into a digital signal, and storing it in the memory of the signal processing unit; applying a digital filtering algorithm (such as a low-pass filter, a high-pass filter, a band-pass filter, etc.) to remove high-frequency noise and low-frequency drift in the digital signal; analyzing the digital signal spectrum through fast Fourier transform (FFT), identifying the frequency component, calculating the actual speed or frequency, and comparing the calculation result with the preset balance state reference value to determine whether the rotor is in an unbalanced state, as well as the direction and degree of deviation.

[0277] The control unit is used to decide whether to remove weight or add counterweight on the rotor based on the analysis results of the signal processing unit, and control the corresponding actuator to operate. The processing process specifically includes: the control unit receives the analysis results from the signal processing unit, which include whether the rotor is in an unbalanced state, the direction of deviation and the degree of deviation. When the rotor is in an unbalanced state, the position and weight of the weight to be removed or added are calculated based on the analysis results corresponding to the unbalanced state. A control signal is generated and sent to the actuator, instructing it to add counterweight or remove weight. Among them, the direction and degree of deviation specifically refer to the direction in which the rotor deviates from the balanced state (for example, the speed is too high or too low) and the numerical value of the deviation.

[0278] Among them, when the rotor is in an unbalanced state, based on the analysis results corresponding to the unbalanced state, the position and weight that need to be removed or added with counterweight are calculated, which specifically includes: receiving the unbalanced state information from the signal processing unit, including the direction of deviation (speed is too high or too low) and the degree of deviation (specifically the numerical value). According to the direction of deviation and the degree of deviation, the characteristics of the current unbalanced mass are determined. For the speed is too high, it is necessary to remove the weight; for the speed is too low, it is necessary to add counterweight. The specific weight that needs to be removed or added with counterweight is determined by the algorithm. The algorithm will calculate according to the total mass of the rotor, the degree of deviation and the preset parameters of the system. For example, the larger the deviation, the more weight may need to be adjusted. Determine the specific position that needs to be adjusted. The position will be relative to the position of the unbalanced mass to effectively offset the unbalanced state. For example, if the unbalanced mass is on the left side of the rotor, it needs to be adjusted on the right side. Generate detailed adjustment instructions based on the calculated weight and position. These instructions will contain the specific position and weight information that need to be added or removed with counterweight.

[0279] An example is given below: Assuming the total rotor mass of the pumped storage unit is 10,000 kg, the detected speed deviation is 20 RPM, and the system preset parameter is that 0.1 kg of weight needs to be adjusted for every 1 RPM deviation.

[0280] The received data is: speed deviation 20RPM, total mass of the rotor 10000 kg.

[0281] Apply preset parameters: System rule adjusts weight by 0.1 kg for every 1 RPM deviation.

[0282] Calculate the adjusted weight: Adjusted weight M = 20RPM*0.1kg / RPM = 2kg.

[0283] Determine the adjustment position: Add a 2 kg weight to the designated position on the right side of the rotor to restore balance.

[0284] The actuator is used to adjust the mass distribution of the rotor by adding or removing weights under the command of the control unit, thereby reducing the centrifugal force caused by the center of mass offset and thus reducing the rotor vibration.

[0285] In addition, the frequency monitoring sensor 7 is also used to detect the balance state of the rotor after adjustment, and transmit the feedback signal back to the control unit to ensure that the adjustment achieves the expected effect and to make further adjustments.

[0286] See also Fig.10 As shown, the sound sensor 11 is used to collect and obtain the sound signal generated when the pumped storage unit is running. When the spectrum of the sound signal deviates from the original spectrum, there is a fault.

[0287] See also Figure 5 As shown, the bearing detection module includes an infrared imaging unit, an optical fiber sensing unit, a bearing detection unit and a fault diagnosis unit.

[0288] The infrared imaging unit is used to detect the temperature distribution of the bearing part 2 using infrared thermal imaging technology.

[0289] The optical fiber sensing unit is used to embed the optical fiber sensor 14 into the bearing part 2 or the supporting structure of the bearing part 2 to detect the deformation and stress state of the bearing part 2 in real time.

[0290] The bearing detection unit is used to detect the working state and health state of the bearing 2 in real time using sensing technology and algorithms.

[0291] The fault diagnosis unit is used to perform fault diagnosis on the bearing part 2 according to the temperature distribution state, deformation state, stress state and vibration state of the bearing part 2.

[0292] See also Figure 5 As shown, according to some embodiments of the present invention, the bearing detection module also includes a fault prediction unit, which is used to establish a prediction model for the bearing 2 fault, compare it with historical data, predict the occurrence time of the bearing 2 fault and issue a warning signal.

[0293] Specifically, see Fig.10 As shown, the infrared imaging unit uses infrared thermal imaging technology (thermal imager 13) to detect the temperature distribution of the bearing part 2 in real time. The temperature distribution image of the surface of the bearing part 2 can be obtained non-contactly through infrared thermal imaging technology, and analyzed to detect abnormal temperature areas, thereby identifying potential thermal faults.

[0294] The optical fiber sensing unit embeds the optical fiber sensor 14 into the bearing part 2 or the supporting structure of the bearing part 2 to detect the deformation information and stress state of the bearing part 2 in real time. By detecting the optical signal change of the optical fiber sensor 14, the force condition of the bearing part 2 can be obtained in real time, and the working state of the bearing part 2 can be evaluated according to the deformation information and stress distribution condition.

[0295] The bearing detection unit uses sensing technology and algorithms to detect the working status and health status of the bearing part 2 in real time. The bearing detection unit collects vibration, sound, current and other signals of the bearing part 2, and combines signal processing algorithms to analyze the operation status of the bearing part 2 in real time to identify possible fault characteristics.

[0296] The sensors and algorithms used above include the following steps.

[0297] The vibration signal of the bearing 2 is detected by using the acceleration sensor 8, and the vibration spectrum is analyzed by fast Fourier transform FFT to identify the abnormal vibration mode.

[0298] A microphone or an acoustic sensor is used to capture the sound signal generated when the bearing part 2 is in operation, and a short-time Fourier transform STFT or a wavelet transform is used to perform frequency analysis to detect abnormal sound characteristics.

[0299] The motor current signal is monitored by the current sensor, and the time series analysis and autoregressive model (AR model) are applied to analyze the change trend and abnormal fluctuation of the current signal.

[0300] The data collected by the above sensors are input into a machine learning model (such as a support vector machine (SVM) or a deep neural network (DNN). The trained model can automatically identify and classify different types of bearing part 2 faults.

[0301] The fault prediction unit establishes a fault prediction model for the bearing part 2, and uses historical data for training, predicts the time of occurrence of the fault of the bearing part 2, and issues a warning signal. Based on a machine learning algorithm or a deep learning model, the fault prediction unit can use historical data and real-time detection data to accurately predict the time of occurrence of the fault of the bearing part 2, and issue an alarm in time so that maintenance personnel can take corresponding maintenance measures to avoid production interruption and equipment damage caused by the fault.

[0302] During operation, the support vector machine algorithm can be used to classify and predict the bearing part 2 faults. By training the SVM model, features such as the rotation temperature and rotation speed of the bearing part 2 and the stress distribution of the bearing part 2 during rotation can be extracted from the bearing part 2 detection data and classified to identify the normal working state and different types of fault states of the bearing part 2. At the same time, the SVM model can be used to perform regression analysis on the fault data of the bearing part 2 to predict the remaining life of the bearing part 2 or the time when a fault occurs.

[0303] See also Figure 6As shown, the water guide detection module includes a hydraulic characteristic detection unit, an intelligent water leakage detection unit, a material fatigue detection unit and a water quality detection unit.

[0304] The hydraulic characteristic detection unit is used to detect and analyze the pressure, flow rate, and flow hydraulic characteristics of the water flow, and determine whether there are any abnormal conditions in the water pipe; the intelligent water leakage detection unit is used to combine sensor technology and data analysis algorithms to detect parameters such as soil moisture, surface water level, groundwater level, etc. around the water pipe, identify the location and degree of water leakage in real time, and issue an alarm in advance to avoid the occurrence of water leakage accidents; the material fatigue detection unit is used to use the strain sensor 18 to detect the deformation and vibration of the water pipe, analyze the strain and vibration data of the pipe to evaluate the fatigue damage degree of the pipe, and predict the life of the pipe; the water quality detection unit is used to detect water quality parameters in the water pipe, such as pH value, turbidity, dissolved oxygen, etc., to determine whether there are corrosion, pollution or biological growth problems inside the pipe.

[0305] The operation status data of the water guiding part 3 include: water flow characteristic data of the water guiding part 3 , pressure data of the water guiding part 3 and structural status data of the water guiding part 3 .

[0306] See also Fig.11 As shown, the water flow characteristic data of the water guide 3 can be obtained in the following ways: a flow sensor (flow meter 22) is used to measure the flow rate of the water flow of the water guide 3, that is, the amount of water passing through a certain cross section per unit time. A flow velocity sensor (flow meter 15) is used to measure the velocity of the water flow of the water guide 3. An ultrasonic flow meter is used to measure the flow rate and flow velocity by using the change in the propagation velocity of ultrasonic waves in water. An electromagnetic flow meter is used to measure the flow rate by detecting the induced electromotive force of the conductive liquid in the magnetic field.

[0307] See also Fig.11 As shown, the pressure data of the water guide 3 can be obtained in the following ways: a pressure sensor 19 is installed on the pipe or structure of the water guide 3 to monitor the pressure of the water flow in real time. A differential pressure sensor is used to measure the pressure difference between two points and can provide more accurate fluid dynamics information.

[0308] See also Fig.11 As shown, the structural status data of the water guide 3 can be obtained in the following ways: a strain gauge is installed on the structural surface of the water guide 3 to measure the structural deformation and stress. An acceleration sensor 8 is used to monitor the vibration and dynamic response of the water guide 3 and evaluate the structural health status. A laser displacement sensor is used to detect the tiny displacement and deformation of the structural surface of the water guide 3. A temperature sensor is used to monitor the temperature change of the water guide 3, which helps to analyze the influence of thermal expansion and contraction on the water guide 3.

[0309] Through various types of sensor data, the operating status of the water guide 3 can be fully monitored. By integrating the water flow characteristic data, pressure data and structural status data, the fault and health status of the water guide 3 can be judged more accurately.

[0310] See also Fig.11 As shown, specifically, first, the hydraulic characteristic detection unit determines whether there is any abnormality in the water pipe by detecting and analyzing the hydraulic characteristics of the water flow, such as pressure, flow velocity, and flow rate. By using equipment such as pressure sensor 19, flow meter 15 and flow meter 22, the hydraulic characteristic detection unit can detect the dynamic changes of the water flow in the pipe in real time, and with the help of data analysis algorithms, identify possible pipe blockage, damage or flow abnormality.

[0311] Specifically, the intelligent water leakage detection unit combines sensor technology and data analysis algorithms to detect parameters such as soil moisture, surface water level, and groundwater level around the water pipe to identify the location and extent of the water leakage in real time and issue an alarm in advance to avoid the occurrence of water leakage accidents. The intelligent water leakage detection unit uses soil moisture sensors (hygrometers 17), water level sensors 21 and other equipment to detect the moisture changes in the surrounding environment in real time, combined with data analysis algorithms, to accurately locate the water leakage location, evaluate the degree of the water leakage, and take timely repair measures.

[0312] See also Fig.11 As shown, specifically, the material fatigue detection unit can use the strain sensor 18 to detect the deformation and vibration of the water pipe, and evaluate the fatigue damage of the pipe by analyzing the strain and vibration data of the pipe, and predict the life of the pipe. The material fatigue detection unit uses a high-precision strain sensor 18 to detect the deformation of the pipe in real time, and combines with the vibration sensor 12 to analyze the vibration spectrum of the pipe, and quantitatively evaluates the fatigue damage of the pipe through an algorithm, providing a basis for pipeline maintenance.

[0313] Specifically, see Fig.11 As shown, the water quality detection unit detects water quality parameters in the water pipe, such as pH value (through Fig.11 The water quality detection unit uses a water quality sensor to detect water quality parameters in real time, and uses a data analysis algorithm to analyze the detection data to promptly detect abnormal water quality inside the pipeline to ensure the safe operation of the water pipeline. Logistic regression algorithms are used to detect intelligent water leakage. By collecting data obtained by the sensors of the intelligent water leakage detection unit, such as soil moisture, surface water level, groundwater level and other parameters, a logistic regression model is constructed to classify and predict the leakage status, thereby accurately identifying the location and degree of the leakage.

[0314] See also Figure 7 As shown, the sealing part detection module includes a pressure detection unit and an intelligent lubrication unit.

[0315] The pressure detection unit is used to detect the pressure changes of the sealing part 4 and analyze the pressure data; the intelligent lubrication unit is used to detect the lubrication condition of the sealing part 4 in real time, and adjust the amount and frequency of lubricant use according to the actual working conditions of the main shaft bearing part 2 collected by the bearing part detection module.

[0316] Specifically, the actual working conditions of the main shaft bearing portion 2 include: the temperature distribution state, deformation state, stress state and vibration state of the bearing portion 2 .

[0317] See also Fig.11 As shown, specifically, first, the pressure detection unit (pressure sensor 19) is responsible for detecting the pressure change of the sealing part 4 and analyzing the pressure data. By installing the pressure sensor 19 on the sealing part 4, the pressure sensor 19 can detect the pressure condition of the sealing part 4 in real time. Using the data acquisition system, the pressure detection unit transmits the collected pressure data to the background system for analysis. In the background system, a data processing algorithm, such as a sliding average method or Fourier transform, can be used to process and analyze the pressure data to identify whether there is abnormal pressure or leakage in the sealing part 4, so as to take corresponding maintenance measures in time to ensure the sealing performance of the equipment.

[0318] Secondly, the intelligent lubrication unit introduces an intelligent lubrication system to detect the lubrication condition of the sealing part in real time and adjust the amount and frequency of lubricant use according to the actual working conditions. The intelligent lubrication unit detects the lubrication condition of the sealing part in real time by installing a lubrication state sensor or a lubricating oil flow sensor. At the same time, combined with the working state and environmental conditions of the equipment, the intelligent lubrication system can automatically adjust the amount and frequency of lubricant use through the lubricant distributor 20 to ensure that the sealing part is always in the best lubrication state. The algorithm embodiments that can be used include a PID control algorithm based on feedback control, which detects and adjusts the lubrication state data obtained by the sensor in real time, so that the amount and frequency of lubricant use can be adaptively adjusted according to the working state of the equipment, thereby achieving the best lubrication effect.

[0319] See also Figure 8 As shown, the embedded part detection module includes a pipeline water level detection unit.

[0320] The pipeline water level detection unit is used to detect the water level and flow in the pipeline through the water level sensor 21 and the flow sensor, obtain water level and flow signals, and use the fuzzy logic system and the obtained water level and flow signals to predict the changes in the water level and flow of the buried pipeline and the pre-buried pipeline of the buried part 5.

[0321] Specifically, see Fig.11As shown, the pipeline water level detection unit uses a fuzzy logic system to predict the changes in the water level and flow of the buried pipeline and the pre-buried pipeline. The pipeline water level detection unit detects the changes in the water level and flow in the pipeline system in real time by installing a water level sensor 21 and a flow sensor (flow meter 22). With the help of the fuzzy logic system, the pipeline water level and flow can be predicted and analyzed based on the data obtained by the above-mentioned water level sensor 21 and the flow sensor, combined with pre-set rules and fuzzy reasoning methods. The fuzzy control algorithm can be used to realize automatic control of the water pump according to the changes in the water level and flow, so as to maintain the stable operation state of the pipeline system.

[0322] See also Fig. 9 As shown, the operation recording module includes a power generation condition recording unit, a pumping condition unit, a power generation phase adjustment recording unit and a water pump phase adjustment recording unit.

[0323] The power generation condition recording unit is used to record the power output of the unit in real time, and record the output voltage and current of the unit to detect voltage imbalance or current abnormality; the pumping condition unit is used to record the flow, pressure and speed of the water pump, as well as the water level of the reservoir in real time; the power generation phase modulation recording unit is used to record and detect the phase relationship of the voltage and current of each phase output by the generator; the water pump phase modulation recording unit is used to detect the position of the water pump rotor and record the water pump current and voltage.

[0324] Specifically, see Fig.10 As shown, the power generation condition recording unit is responsible for recording the power output, output voltage and current of the generator set in real time. The ammeter 9 is used to record the current of the generator set in real time, and the voltmeter 10 is used to record the output voltage of the generator set in real time. By recording these parameters, the system can detect voltage imbalance or current abnormality, so as to take corresponding measures in time. For example, the power factor correction algorithm can be used to calculate and adjust the power factor of the generator set based on the recorded power output, voltage and current data to ensure that it operates in the best state.

[0325] The pumping condition unit is responsible for recording the flow, pressure and speed of the pump, as well as the water level of the reservoir in real time. Real-time detection of these parameters helps to evaluate the performance of the pump system and detect any abnormalities in a timely manner. For example, the flow anomaly detection algorithm can be used to analyze the recorded flow signal to detect problems such as pipe blockage or pump wear.

[0326] The power generation phase adjustment recording unit is used to record and detect the phase relationship between the voltage and current of each phase of the generator output, which helps to detect whether the generator set has phase imbalance or imbalance, and judge whether the system is in a stable operating state. By recording the phase relationship data in real time and combining it with the phase correction algorithm, the generator set can be phase adjusted and corrected to ensure the balance and stability between the voltage and current of each phase.

[0327] The pump phase recording unit detects the position of the pump rotor and records the pump current and voltage. These data help determine the operating status and efficiency of the pump and detect any abnormalities in a timely manner. For example, the rotor position detection algorithm can be used to analyze the recorded rotor position data to detect whether the pump has abnormal vibration or unstable operation. The power factor correction algorithm can be used to ensure that the operation recording module can effectively detect and record the operating status of the unit and pump system, and provide timely fault diagnosis and early warning.

[0328] See also Figure 3 As shown, according to some embodiments of the present invention, the pumped storage unit fault detection device further includes an intelligent decision support system, which is used to generate fault handling and maintenance suggestions based on the identified fault mode and / or fault development trend of the pumped storage unit. The specific principle of the intelligent decision support system is as above and will not be repeated here.

[0329] It can be seen from the description of the above embodiments that the pumped storage unit fault detection method and detection device provided by the present invention have at least the following advantages.

[0330] (1) The pumped storage unit fault detection method provided by the present invention can provide a pumped storage unit with a fault diagnosis report under different operating conditions, avoiding the one-sidedness of capturing the unit status under only one operating condition. The data capture of multiple operating conditions makes it easier to expose invisible faults, improves the accuracy of fault detection, and is conducive to the detection and corresponding maintenance of unit faults. The method solves the defects of the existing pumped storage unit fault detection method that data collection is relatively one-sided, faults caused by different operating conditions are not easy to expose, resulting in low fault detection accuracy.

[0331] (2) The pumped storage unit fault detection method provided by the present invention determines whether there is an abnormality in the water pipe by detecting and analyzing the hydraulic characteristics of the water flow, such as pressure, flow velocity, and flow rate, through a hydraulic characteristic detection unit. By using equipment such as a pressure sensor 19, a velocity meter, and a flow meter 22, the hydraulic characteristic detection unit can detect the dynamic changes of the water flow in the water pipe in real time, and identify problems such as pipe blockage, damage, or flow rate abnormality with the help of a data analysis algorithm.

[0332] (3) The pumped storage unit fault detection device provided by the present invention can obtain the fault signals corresponding to the rotating part 1, the bearing part 2, the water guide part 3, the sealing part 4 and the embedded part 5 of the pumped storage unit when the pumped storage unit is operating in the power generation condition, the pumping condition, the power generation phase adjustment condition and the water pump phase adjustment condition through the rotating part detection module, the bearing part detection module, the water guide part detection module, the sealing part detection module and the embedded part detection module, respectively, and comprehensively analyze the above fault signals through the data integration analysis module to identify the fault mode and / or fault development trend of the pumped storage unit. Similarly, the defects of the existing pumped storage unit fault detection method that the data collection is relatively one-sided and the faults caused by different working conditions are not easy to expose, resulting in low fault detection accuracy, are solved.

[0333] (4) The pumped storage unit fault detection device provided by the present invention converts the vibration energy generated by the pumped storage unit during operation into electrical energy by using the mechanical vibration energy recovery technology through the energy recovery unit. The conversion of this part of electrical energy adopts piezoelectric materials, electromagnetic induction and other technologies, and realizes energy recovery by converting mechanical vibration into electrical energy, thereby improving the energy utilization efficiency of the system.

Claims

1. A pumped storage unit fault detection method, characterized in that: include: Obtaining the fault signals of the rotating part, the bearing part, the water guide part, the sealing part and the embedded part when the pumped storage unit is operating in power generation condition, pumping condition, power generation phase adjustment condition and water pump phase adjustment condition; Extracting characteristic parameters of the rotating part fault signal, the bearing part fault signal, the water guide part fault signal, the sealing part fault signal and the embedded part fault signal; fusing characteristic parameters of the rotating part fault signal, the bearing part fault signal, the water guide part fault signal, the sealing part fault signal and the embedded part fault signal to determine the fused characteristic parameters; Based on the fused characteristic parameters, the failure mode of the pumped storage unit is identified, and the failure mode of the pumped storage unit and the confidence level corresponding to the failure mode are determined; Based on the failure mode of the pumped storage unit, the characteristic parameters corresponding to the failure mode, and the degree of deviation from the preset normal operating parameter range, the severity of the identified failure mode is evaluated to determine the severity level of the failure; Based on historical fault signals, the fault mode of the pumped storage unit and its confidence, and the severity level of the pumped storage unit fault, the fault development trend is predicted, and the predicted fault development trend and critical time point are determined; Based on the failure mode of the pumped-storage unit and its confidence, the severity level of the failure, the predicted failure development trend and critical time point, a failure diagnosis report is generated, wherein the failure diagnosis report includes the failure mode of the pumped-storage unit, the severity level of the failure and the failure development trend.

2. The pumped storage unit fault detection method according to claim 1, characterized in that: The step of obtaining a fault signal of a rotating part of the pumped-storage unit when the unit is operating in a power generation condition, a pumping condition, a power generation phase adjustment condition, and a water pump phase adjustment condition comprises: obtaining a sound signal, an amplitude signal, a frequency signal, a rotation speed signal, and a vibration direction signal of a rotor of the rotating part, and determining the fault signal of the rotating part based on the sound signal, the amplitude signal, the frequency signal, the rotation speed signal, and the vibration direction signal of the rotor; And / or, the obtaining of a bearing fault signal when the pumped-storage unit is operating in a power generation condition, a pumping condition, a power generation phase adjustment condition, and a water pump phase adjustment condition comprises: obtaining a temperature distribution signal, a deformation signal, a stress state signal, and a vibration state signal of the bearing, and determining the bearing fault signal based on the temperature distribution signal, the deformation signal, the stress state signal, and the vibration state signal of the bearing; And / or, the obtaining of the water guide fault signal when the pumped-storage unit is operating in a power generation condition, a pumping condition, a power generation phase adjustment condition, and a water pump phase adjustment condition comprises: obtaining operation status data of the water guide, and determining the water guide fault signal based on the operation status data of the water guide, wherein the operation status data of the water guide comprises water flow characteristic data and pressure data, and the water flow characteristic data comprises a water flow pressure signal, a water flow velocity signal, and a water flow hydraulic characteristic signal of the water flow of the water guide; And / or, the obtaining of the seal fault signal when the pumped-storage unit is operating in a power generation condition, a pumping condition, a power generation phase adjustment condition, and a water pump phase adjustment condition comprises: obtaining a pressure change signal of the seal, and determining the seal fault signal based on the pressure change signal; And / or, the method of obtaining the buried part fault signal when the pumped-storage unit is operating in power generation condition, pumping condition, power generation phase adjustment condition and water pump phase adjustment condition includes: obtaining the water level signal and flow signal of the buried part, and determining the buried part fault signal based on the water level signal and the flow signal.

3. The pumped storage unit fault detection method according to claim 2, characterized in that: The step of determining the rotor fault signal based on the sound signal of the rotor of the rotor comprises: Establish a fault identification model to classify and store the sound frequency characteristics of the pumped storage unit in normal operation and various fault states; comparing the sound frequency features extracted from the sound signal of the rotor of the rotating part with the sound spectrum features in the fault identification model; Determining the rotating part fault signal based on a comparison result of the sound frequency feature extracted from the sound signal and the sound spectrum feature in the fault identification model; The step of determining the rotor fault signal based on the amplitude signal, the frequency signal, the rotation speed signal and the vibration direction signal of the rotor of the rotor comprises: Calculating the vibration energy of the rotor of the rotating part based on the amplitude signal, the frequency signal and the speed signal of the rotor of the rotating part to obtain the vibration energy data of the rotor of the rotating part; Based on the vibration energy data of the rotor of the rotating part and the vibration direction signal of the rotor of the rotating part, the source and characteristics of the vibration are determined to determine the fault signal of the rotating part.

4. The pumped storage unit fault detection method according to claim 2, characterized in that: The determining the bearing fault signal based on the temperature distribution signal of the bearing part comprises: determining a temperature distribution image of the bearing portion based on the temperature distribution signal of the bearing portion; Based on the temperature distribution image of the bearing part, determining the abnormal temperature area of ​​the bearing part to determine the bearing part fault signal; The determining the bearing fault signal based on the deformation signal and the stress state signal of the bearing part comprises: Acquiring an optical signal of a fiber optic sensor located in the bearing portion or in a supporting structure of the bearing portion; Based on the change of the optical signal, determining the deformation signal and the stress state signal of the bearing part to determine the fault signal of the bearing part; The determining the bearing fault signal based on the vibration state signal of the bearing comprises: Performing time domain analysis and frequency domain analysis on the vibration state signal of the bearing part to determine the operating frequency, vibration amplitude and vibration mode of the bearing part; Based on the operating frequency, vibration amplitude and vibration mode of the bearing part, the vibration state of the bearing part is determined to determine the bearing part fault signal.

5. The pumped storage unit fault detection method according to claim 2, characterized in that: The step of obtaining a bearing fault signal of the pumped-storage unit when the unit is operating in a power generation condition, a pumping condition, a power generation phase adjustment condition, and a water pump phase adjustment condition also includes: Establishing a fault prediction model for the bearing portion; Inputting the temperature distribution signal, deformation signal, stress state signal and vibration state signal of the bearing part into the fault prediction model of the bearing part; Based on the output result of the fault prediction model of the bearing part, the fault occurrence time of the bearing part is determined and a warning signal is issued. The output result of the prediction model includes a fault occurrence time prediction signal, a fault warning signal and a fault type identification signal.

6. The pumped storage unit fault detection method according to claim 2, characterized in that: The running state data of the water guide also includes structural state data, and the structural state data includes strain data and vibration data of the water guide; And / or, the operating status data of the water guide also includes environmental status data; And / or, the operating status data of the water guiding part also includes water quality status data.

7. A pumped storage unit fault detection device, characterized in that: It includes a rotating part detection module, a bearing part detection module, a water guide part detection module, a sealing part detection module, an embedded part detection module and a data integration and analysis module, wherein the rotating part detection module, the bearing part detection module, the water guide part detection module, the sealing part detection module and the embedded part detection module are all in communication connection with the data integration and analysis module; The rotating part detection module, the bearing part detection module, the water guide part detection module, the sealing part detection module and the embedded part detection module are respectively used to obtain the rotating part fault signal, the bearing part fault signal, the water guide part fault signal, the sealing part fault signal and the embedded part fault signal when the pumped storage unit is operating in the power generation condition, the pumping condition, the power generation phase adjustment condition and the water pump phase adjustment condition; The data integration analysis module is used for: Extract characteristic parameters of rotating part fault signals, bearing part fault signals, water guide part fault signals, sealing part fault signals and embedded part fault signals; fusing characteristic parameters of the rotating part fault signal, the bearing part fault signal, the water guide part fault signal, the sealing part fault signal and the embedded part fault signal to determine the fused characteristic parameters; Based on the fused characteristic parameters, the failure mode of the pumped storage unit is identified, and the failure mode of the pumped storage unit and the confidence level corresponding to the failure mode are determined; Based on the failure mode of the pumped storage unit, the characteristic parameters corresponding to the failure mode, and the degree of deviation from the preset normal operating parameter range, the severity of the identified failure mode is evaluated to determine the severity level of the failure; Based on historical fault signals, the fault mode and confidence level of the pumped storage unit, and the fault severity level of the pumped storage unit, the fault development trend is predicted, and the predicted fault development trend and critical time point are determined; Based on the specific failure mode of the pumped-storage unit and its confidence level, the severity level of the failure, the predicted failure development trend and critical time point, a failure diagnosis report is generated, wherein the failure diagnosis report includes the failure mode of the pumped-storage unit, the severity level of the failure and the failure development trend.

8. The pumped storage unit fault detection device according to claim 7, characterized in that: The rotating part detection module includes a sound feature extraction unit, a vibration energy analysis unit and a comprehensive fault diagnosis unit, and the sound feature extraction unit, the vibration energy analysis unit and the comprehensive fault diagnosis unit are all connected to the data integration analysis module in communication; The sound feature extraction unit is used to obtain the sound signal generated when the main shaft rotor in the rotating part is running, and perform frequency analysis on the sound signal according to the signal processing algorithm; The vibration energy analysis unit is used to obtain the amplitude signal, frequency signal, speed signal and vibration direction signal of the rotor, and calculate the vibration energy data of the rotor based on the amplitude signal, frequency signal and speed signal of the rotor to determine the fault signal of the rotor; The comprehensive fault diagnosis unit is used to obtain and integrate the detection results of the sound feature extraction unit and the vibration energy analysis unit and perform multi-dimensional analysis to determine the fault signal of the rotating part.

9. The pumped storage unit fault detection device according to claim 7, characterized in that: The bearing detection module includes an infrared imaging unit, an optical fiber sensing unit, a vibration detection unit and a fault diagnosis unit, and the infrared imaging unit, the optical fiber sensing unit, the vibration detection unit and the fault diagnosis unit are all communicatively connected with the data integration and analysis module; The infrared imaging unit is used to detect the temperature distribution state signal of the bearing part; The optical fiber sensing unit is used to detect the deformation signal and stress state signal of the bearing part; The vibration detection unit is used to use the vibration signal of the bearing part, and identify the operating frequency, vibration amplitude and vibration mode of the bearing part by performing time domain analysis and frequency domain analysis on the vibration signal of the bearing part, so as to determine the vibration state signal of the bearing part; The fault diagnosis unit is used to determine the bearing fault signal according to the temperature distribution state signal, deformation signal, stress state signal and vibration state signal of the bearing.

10. The pumped storage unit fault detection device according to claim 7, characterized in that: The water guide detection module includes a hydraulic characteristic detection unit, a pressure detection unit and a structural state detection unit, and the hydraulic characteristic detection unit, the pressure detection unit and the structural state detection unit are all connected to the data integration and analysis module for communication; The hydraulic characteristic detection unit is used to obtain water flow characteristic data of the water guide part, and judge the fault signal of the water guide pipe in the water guide part according to the water flow characteristic data, wherein the water flow characteristic data includes the water flow pressure signal, flow velocity signal and flow hydraulic characteristic signal of the water guide part; The pressure detection unit is used to obtain pressure data of the water guide part, and determine a pressure fault signal of the water guide part according to the pressure data; The structural state detection unit is used to obtain structural state data of the water guide part, and determine the structural fault signal of the water guide part according to the structural state data, wherein the structural state data includes strain data and vibration data of the water guide part.

Citation Information

Cited By

  • Proximity sensor self-learning identification system based on multi-dimensional signal processing

    CN119397377A

  • Proximity sensor self-learning recognition system based on multi-dimensional signal processing

    CN119397377B

  • Leakage fault diagnosis method and system for oil seal of mining speed reducer

    CN119688168A

  • Diamond polishing and grinding online detection system

    CN120195178A

  • An online detection system for diamond polishing and grinding

    CN120195178B