Speed regulation system fault diagnosis model construction method, fault diagnosis method and device

Abstract

The invention provides a speed regulation system fault diagnosis model construction method and a fault diagnosis method and device, and the construction method comprises the steps: constructing a physical simulation model of a hydraulic system of a water turbine, and carrying out the simulation of typical working conditions and fault behaviors based on the physical simulation model, and obtaining simulation data; based on real data and the simulation data, generating synthetic data through a deep learning algorithm; and training an initial model by taking the simulation data, the real data and the synthetic data as a training set to obtain the fault diagnosis model. According to the speed regulation system fault diagnosis model construction method and the fault diagnosis method and device provided by the invention, the training set is greatly expanded, the problem of scarcity of real fault samples is solved, the expansion samples are ensured to be consistent with physical rules in distribution, frequency spectrum and energy levels, and the training effect in a small sample / unbalanced scene is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of turbine speed control system technology, and in particular to a method for constructing a fault diagnosis model for a speed control system, a fault diagnosis method, and a device. Background Technology

[0002] Fault diagnosis technology for hydro turbine speed control systems has evolved through three generations: from "mechanism model" to "data-driven" and then to "digital twin fusion". Early research was mainly based on qualitative mechanism methods such as fault tree analysis (FTA) and failure mode and effects analysis (FMEA), relying on expert experience to construct fault trees. These methods could identify weak links in the system during the design phase, but it was difficult to achieve quantitative diagnosis of uncertainties in the online state.

[0003] With the gradual improvement of power plant sensor networks, pure data-driven methods, represented by neural networks and support vector machines, are becoming more and more popular. These methods utilize a large number of normal and fault samples for offline training, enabling end-to-end fault diagnosis with an accuracy rate of over 95% in a laboratory environment.

[0004] In the past three years, digital twin technology has been introduced into this field, forming a dual-engine architecture centered on "mechanism simulation + data-driven approach": First, high-fidelity mechanism models are established using platforms such as AMESim and Simscape, and virtual faults are injected to generate expanded samples. Then, deep learning models such as convolutional neural networks (CNN) and long short-term memory networks (LSTM) are used to train diagnostic models, and finally, expert systems output handling suggestions. This technical approach has been demonstrated in engineering applications in the speed regulation subsystems of giant hydroelectric units such as the Three Gorges Dam and Xiangjiaba Dam—especially in the hydraulic and regulating systems. It has reduced the diagnostic response time from minutes to seconds and the false alarm rate by about 30%, and has become the mainstream technology for current engineering applications.

[0005] However, existing technologies still face several bottlenecks that urgently need to be addressed: First, the reliability of virtual fault data is insufficient. Digital twin systems often rely on behavioral-level simulation to generate fault samples, but cross-physical-domain degradation mechanisms such as hydraulic oil deterioration, valve core jamming, and seal aging are difficult to fully parameterize and model in platforms like AMESim. This results in significant physical drift between virtual samples and real fault states, which exacerbates model bias when used directly for training. Second, the problems of small sample sizes and class imbalance remain prominent. The mean time between failures (MTBF) of large-scale unit hydraulic systems exceeds [a certain threshold]. In real-world fault samples, samples generated by traditional methods such as Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs) can increase the amount of data, but they often violate physical constraints such as the conservation of mass, momentum, and energy, leading to the model learning "false features" and resulting in the contradictory phenomenon of "high accuracy in the laboratory and high false negative rate in reality" in the field.

[0006] Therefore, there is an urgent need to develop a new fault diagnosis architecture that can integrate physical information enhancement generation and dynamic decision fusion, so as to systematically improve the accuracy, interpretability and engineering applicability of fault diagnosis of turbine speed control systems. Summary of the Invention

[0007] This invention provides a method for constructing a fault diagnosis model for a speed control system, a fault diagnosis method, and an apparatus to address the shortcomings of existing technologies, such as insufficient reliability of virtual fault data and poor accuracy and interpretability of fault diagnosis in turbine speed control systems due to small sample sizes and class imbalance. This invention provides a method for constructing a fault diagnosis model for a speed control system, comprising: A physical simulation model of the hydraulic system of the water turbine is constructed, and simulations of typical working conditions and fault behaviors are performed based on the physical simulation model to obtain simulation data; Based on real data and the simulation data, synthetic data is generated using deep learning algorithms; The simulation data, real data, and synthetic data are used as the training set to train the initial model, thus obtaining the fault diagnosis model.

[0008] According to the method for constructing a fault diagnosis model for a speed control system provided by the present invention, the construction of a physical simulation model of a hydraulic turbine system includes: Mathematical modeling is performed on the working mechanism of oil pumps, combination valves, and relays in the hydraulic system of a water turbine. The typical operating conditions include oil pump start-up and shutdown, oil pump rotation, and air supply valve start-up and shutdown. The fault behaviors include faults and alarms. The faults include analog signal acquisition faults, oil pump motor faults, timeout faults, solenoid valve faults, communication faults, and power supply faults. The alarms include oil pressure alarms, oil level alarms, temperature alarms, and oil-water mixing alarms.

[0009] According to the fault diagnosis model construction method for speed control systems provided by this invention, the working mechanism of the oil pump in the hydraulic system of a water turbine is mathematically modeled, including: The pitch is calculated based on the lead and number of helical blades. The flow rate of the oil pump is calculated based on the pitch, the diameter of the helical blades, the diameter of the pump shaft, the water lifting cross-sectional area ratio, and the rotational speed. The shaft power is calculated based on the oil pump's flow rate, liquid lift height, liquid density, and oil pump efficiency.

[0010] According to the method for constructing a fault diagnosis model for a speed control system provided by the present invention, a mathematical model is performed on the working mechanism of the servo in the hydraulic system of a water turbine, including: When the relay is activated, the net piston force is calculated based on the piston rod load, the pressure exerted on the piston by the opening chamber, the piston area of ​​the opening chamber, the pressure exerted on the piston by the closing chamber, the piston area of ​​the closing chamber, and the angle between the relay and the horizontal plane. Based on the net piston force and the total stroke of the relay, the operating work of the relay is calculated. Calculate the starting oil volume based on the piston area of ​​the starting chamber, piston displacement, and dead capacity of the starting chamber of the relay; calculate the oil volume of the stopping chamber based on the piston area of ​​the stopping chamber, total stroke of the relay, piston displacement, and dead capacity of the stopping chamber of the relay. Based on the pressure exerted on the piston by the start-up chamber, the pressure exerted on the piston by the stop-up chamber, and the preset leakage coefficient, the leakage between the start-up chamber and the stop-up chamber is calculated. The liquid flow rate in the start-up chamber is calculated based on the flow rate of the start-up chamber pipeline, the leakage amount, the piston rod moving speed, and the piston area of ​​the start-up chamber. The liquid flow rate in the shutdown chamber is calculated based on the flow rate in the shutdown chamber pipe, the leakage amount, the piston rod moving speed, and the piston area of ​​the shutdown chamber.

[0011] According to the speed control system fault diagnosis model construction method provided by the present invention, after obtaining the fault diagnosis model, the method further includes: When the system structure, oil type, or control strategy changes, the mechanism parameters of the physical simulation model are updated, and simulations of typical operating conditions and fault behaviors are performed based on the updated physical simulation model to obtain updated simulation data. Based on the updated simulation data, the fault diagnosis model is iterated to obtain an updated fault diagnosis model.

[0012] The present invention also provides a fault diagnosis method, comprising: Real-time monitoring of key parameters in the hydraulic system of the water turbine, including pressure, flow rate, temperature, and vibration; Based on the trained fault diagnosis model, fault feature signals are extracted from the key parameters. Based on the extracted fault feature signals, the fault modes of the turbine hydraulic system are identified. The fault diagnosis model is obtained based on the speed regulation system fault diagnosis model construction method.

[0013] The present invention also provides a device for constructing a fault diagnosis model for a speed control system, comprising: The simulation unit is used to construct a physical simulation model of the hydraulic system of the water turbine, and to simulate typical working conditions and fault behaviors based on the physical simulation model to obtain simulation data. A data generation unit is used to generate synthetic data based on real data and the simulation data using a deep learning algorithm. The model training unit is used to train the initial model using the simulation data, real data, and synthetic data as the training set to obtain the fault diagnosis model.

[0014] The present invention also provides a fault diagnosis device, characterized in that it comprises: The parameter monitoring unit is used to monitor key parameters in the hydraulic system of the water turbine in real time, including pressure, flow rate, temperature and vibration. The fault identification unit is used to extract fault feature signals from the key parameters based on the trained fault diagnosis model, and to identify the fault mode of the turbine hydraulic system based on the extracted fault feature signals. The fault diagnosis model is obtained based on the speed regulation system fault diagnosis model construction method according to any one of claims 1 to 5.

[0015] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the speed control system fault diagnosis model construction method or fault diagnosis method as described above.

[0016] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the speed control system fault diagnosis model construction method or fault diagnosis method as described above.

[0017] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the speed control system fault diagnosis model construction method or fault diagnosis method as described above.

[0018] The present invention provides a method for constructing a fault diagnosis model for a speed control system, a fault diagnosis method, and a device. It generates physically labeled simulation data through a high-fidelity physical simulation model, and then generates synthetic data through physical simulation and deep learning. This ensures that the generated data is physically plausible, while significantly expanding the training set. This addresses the scarcity of real fault samples and ensures that the expanded samples are consistent with physical laws in terms of distribution, spectrum, and energy, significantly improving training performance in small-sample / imbalanced scenarios. Furthermore, the combined training of these three methods enables the neural network to automatically satisfy physical laws during training and inference phases, fundamentally suppressing "spurious features" and "outlier decisions," and enhancing the model's extrapolation ability and reliability under real-world conditions. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is one of the flowcharts illustrating the method for constructing a fault diagnosis model for a speed control system provided by the present invention.

[0021] Figure 2 This is a schematic diagram of the combined valve principle provided by the present invention.

[0022] Figure 3 This is a schematic diagram of the relay device structure provided by the present invention.

[0023] Figure 4 This is a schematic diagram of the relay device provided by the present invention.

[0024] Figure 5 This is a schematic diagram of the oil pump start-stop logic provided by the present invention.

[0025] Figure 6 This is a schematic diagram of the start / stop logic of the air replenishment valve provided by the present invention.

[0026] Figure 7 This is a flowchart illustrating the fault diagnosis method provided by the present invention.

[0027] Figure 8 This is a schematic diagram of the structure of the speed regulation system fault diagnosis model construction device provided by the present invention.

[0028] Figure 9 This is a schematic diagram of the fault diagnosis device provided by the present invention.

[0029] Figure 10 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0031] To address the issues of insufficient reliability of virtual fault data in existing technologies, and the poor accuracy and interpretability of fault diagnosis in turbine speed control systems due to small sample sizes and class imbalance, this invention provides a method for constructing a fault diagnosis model for speed control systems. This method first constructs a physical simulation model of the turbine hydraulic system, then simulates typical operating conditions and fault behaviors based on the physical simulation model to obtain simulation data. Next, synthetic data is generated using a deep learning algorithm based on real and simulation data. Finally, the simulation data, real data, and synthetic data are used as a training set to train the initial model, resulting in a fault diagnosis model.

[0032] The method provided in this invention utilizes a high-fidelity physical simulation model to generate simulation data with physical labels, exhibiting excellent interpretability and physical consistency. This enables the neural network to automatically satisfy physical laws during the training and inference phases, fundamentally suppressing "spurious features" and "outlier decisions," and improving the model's extrapolation ability and reliability under real-world conditions.

[0033] Then, synthetic data is generated through physical simulation and deep learning, ensuring that the generated data is reasonable in terms of physical laws. At the same time, it greatly expands the training set, especially the number of samples of rare fault categories, so that the deep neural network, which is originally difficult to train, can obtain sufficient learning materials. This solves the problem of scarce real fault samples and ensures that the expanded samples are consistent with physical laws in terms of distribution, spectrum and energy, significantly improving the training effect in small sample / imbalanced scenarios.

[0034] Mechanism simulation data provides standard answers under ideal and typical fault conditions, while AI-generated synthetic data can introduce more diverse data variants that approximate the real distribution. Training with these three methods enables the model to not only memorize real data but also understand the deeper physical behavior of the system, thus exhibiting stronger generalization ability and stability when faced with unseen operating conditions or noise interference.

[0035] In addition, physical constraints provide physically interpretable anchor points for network output. Each fault criterion can be traced back to explicit equations such as pressure, flow, and heat, enabling maintenance personnel to intuitively understand "why it was judged as valve core jamming" rather than just getting a black box probability.

[0036] This invention can be applied to scenarios requiring fault diagnosis of turbine speed control systems. The execution entity of this method can be an electronic device such as a terminal device, computer, server, server cluster, or a specially designed speed control system fault diagnosis model building device, or a fault diagnosis model building device installed within such an electronic device. This device can be implemented through software, hardware, or a combination of both.

[0037] Figure 1This is one of the flowcharts illustrating the method for constructing a fault diagnosis model for a speed control system provided by the present invention, such as... Figure 1 As shown, the method includes the following steps 110-130.

[0038] Step 110: Construct a physical simulation model of the hydraulic system of the water turbine, and simulate typical working conditions and fault behaviors based on the physical simulation model to obtain simulation data.

[0039] Specifically, the physical simulation model is a mathematical model based on the working principles of a hydraulic turbine system, such as fluid mechanics, mechanical dynamics, and electromagnetism. It simulates the static and dynamic behavior of the system through a series of differential equations, algebraic equations, and logical rules. This model is a virtual digital mapping of the actual physical system. The physical simulation model is implemented using multiphysics simulation software such as AMESim, MATLAB / Simulink, and Dymola. These software programs provide rich libraries of hydraulic, mechanical, and control components, support graphical modeling, and are suitable for system-level behavioral simulation.

[0040] Typical operating conditions refer to the various states of a hydroelectric turbine during normal operation, such as: start-up, shutdown, no-load, power regulation, and primary frequency regulation. Fault behavior simulation refers to simulating the occurrence of specific faults in a physical simulation model by modifying model parameters or structure. For example, limiting the valve core movement speed can simulate valve core jamming, increasing leakage gaps can simulate internal leakage in the main pressure distribution valve, and superimposing deviations on the output signal can simulate sensor drift.

[0041] After running the simulation model, the output data sequence that changes over time is the simulation data. The simulation data may include, for example, various sensor readings such as pressure, displacement, flow rate, voltage, as well as system state variables and fault labels.

[0042] Step 120: Generate synthetic data based on real and simulated data using deep learning algorithms.

[0043] Specifically, real data is historical data collected from the condition monitoring system of the actual operating turbine units on site. It typically includes normal condition data and a small amount of recorded fault data. The data format is similar to that of simulation data, but may contain more noise and uncertainty.

[0044] The deep learning algorithms mentioned here can be algorithms used for generative models, such as generative adversarial networks (GANs), diffusion models (DMs), and recurrent neural network (RNNs).

[0045] Taking the diffusion model as an example, simulated and real data are normalized to the same scale and cut into sample fragments of equal length. The preprocessed mixed data is then input into the diffusion model for training. The model learns the subtle patterns of "normal data" and various types of "fault data." After training, a random noise vector and a specified fault label are input into the generator model, and the model outputs a labeled synthetic fault data that closely approximates the real fault distribution.

[0046] Step 130: Use simulation data, real data, and synthetic data as training sets to train the initial model and obtain the fault diagnosis model.

[0047] Specifically, the training set is a dataset used to train machine learning models. Here, the training set consists of three parts: simulated data (mechanistic knowledge), real data (field knowledge), and synthetic data (AI-enhanced knowledge), forming a diverse, large-scale dataset that covers all fault modes.

[0048] The initial model is the initial state of the deep learning model to be trained, and its parameters are randomly initialized. The initial model can be a traditional machine learning model (such as SVM, decision tree, random forest) or a deep learning model (such as GAN, RNN), and this embodiment of the invention does not specifically limit it.

[0049] The fault diagnosis model is the final model with fault identification capabilities obtained after training. The training process includes data preprocessing, feature selection, model training and validation, ultimately achieving intelligent diagnosis and identification of fault states in hydraulic systems. Given real-time sensor data, the fault diagnosis model can output a diagnostic result, such as "System normal" or "The probability of XX fault occurring is XX%", etc.

[0050] Understandably, the fault diagnosis model learns the system behavior guaranteed by the physical mechanism (simulation data), fits the actual situation on site (real data), and overcomes the bottleneck of scarce samples (synthetic data), thus achieving high accuracy, high generalization ability and strong robustness.

[0051] In some possible implementations, a physical simulation model of the hydraulic system of the turbine is constructed, including: mathematical modeling of the working mechanism of the oil pump, combination valve and servo in the hydraulic system of the turbine.

[0052] Specifically, the oil pump is the power source of the hydraulic system, and its simulation model focuses on simulating the process of converting mechanical energy into hydraulic energy, as well as the output flow rate and pressure. Key parameters include: displacement, speed, volumetric efficiency, mechanical efficiency, and outlet pressure pulsation.

[0053] For oil pump modeling, the pitch can be calculated based on the lead and number of helical blades. The flow rate of the oil pump can be calculated based on the pitch, helical blade diameter, pump shaft diameter, water lifting cross-sectional area ratio, and rotational speed. Then, the shaft power can be calculated based on the oil pump flow rate, liquid lifting height, liquid density, and oil pump efficiency.

[0054] A screw pump is a positive displacement pump with a unique construction, mainly composed of a drive motor and reducer, a connecting rod and connecting rod housing, a stator, and a rotor. During operation, the rotor, driven by the motor, rotates planetarily within the stator. The mating rotor and the elastic bushings of the stator form several non-communicating sealed cavities. Due to the rotation of the rotor, the sealed cavities move axially from the suction end to the discharge end, and the medium is continuously transported from the suction end to the discharge end within the cavities.

[0055] The simulation modeling algorithm for oil pumps is as follows: Calculate pitch :

[0056] in It is expressed as lead, which is the axial distance traveled by the helical blades as they spiral up 360° around the pump shaft. This is expressed as the number of spiral blades.

[0057] Calculate flow :

[0058] in Indicates the diameter of the helical blade; Indicates the pump shaft diameter; Indicates the water lifting section ratio; Indicates rotational speed.

[0059] Calculate shaft power :

[0060] in This indicates the increased liquid density; Indicates the height to which the liquid has been lifted; This indicates the pump's efficiency.

[0061] For speed control systems, the purpose of calculating the oil pump capacity is to determine the rated oil delivery rate and rated pressure of the hydraulic pumps in the hydraulic system. According to the technical specifications for turbine control systems, "combined and separate hydraulic systems should be equipped with two oil pumps, each with an oil delivery rate sufficient to compensate for oil leakage, and a safety factor of at least 2. Typically, the oil delivery rate per minute of each pump should not exceed 0.65 times the volume of the servo drive." Given: Total volume of the guide vane servo drive. Total volume of the blade relay If two identical oil pumps are installed, then the oil delivery capacity of each pump will not exceed:

[0062] Total oil delivery capacity of the two oil pumps:

[0063] In other embodiments, simulation modeling is performed on the combination valve. The combination valve is a hydraulic device with functions such as unloading, safety, and non-return. It is widely used at the oil pump outlet of the oil supply device. Its performance directly affects the performance of the entire hydraulic system and is an important hydraulic component in the oil supply device.

[0064] Figure 2 This is a schematic diagram of the combined valve principle provided by the present invention, as shown below. Figure 2 As shown, the combination valve consists of two cartridge valves and two pilot valves, and it includes the functions of a check valve, an unloading valve, and a safety valve.

[0065] When the oil pump starts, the check valve is closed due to the action of P1. Initially, the solenoid of the unloading valve (a two-position four-way solenoid directional valve) is de-energized, and the pressurized oil in the main valve control chamber returns through the unloading valve, flowing back to the oil tank through the main valve. 5-6 seconds after the pump starts, the unloading valve is energized and reverses direction, allowing pressurized oil to enter the main valve control chamber, which then quickly closes. The pressure at port P rises to a certain level, overcoming the back pressure of the check valve, opening the check valve to supply oil to the pressure tank. After the pump stops, the unloading valve returns to its initial de-energized state.

[0066] When the oil pump supplies oil to the pressure tank, the system's working pressure can be adjusted by the spring preload of the safety valve; when the pressure at port P1 is greater than the rated pressure of the safety valve, the safety valve is unloaded, and the pressurized oil delivered by the oil pump is discharged into the return oil tank.

[0067] In other embodiments, mathematical modeling of the working mechanism of the servo in the hydraulic system of a water turbine can be achieved in the following ways: When the relay is activated, the net piston force is calculated based on the piston rod load, the pressure exerted on the piston by the opening chamber, the piston area of ​​the opening chamber, the pressure exerted on the piston by the closing chamber, the piston area of ​​the closing chamber, and the angle between the relay and the horizontal plane. Based on the net piston force and the total stroke of the relay, the operating work of the relay is calculated. Calculate the starting oil volume based on the piston area of ​​the starting chamber, piston displacement, and dead capacity of the starting chamber of the relay; calculate the oil volume of the stopping chamber based on the piston area of ​​the stopping chamber, total stroke of the relay, piston displacement, and dead capacity of the stopping chamber of the relay. Based on the pressure exerted on the piston by the start-up chamber, the pressure exerted on the piston by the stop-up chamber, and the preset leakage coefficient, the leakage between the start-up chamber and the stop-up chamber is calculated. Calculate the liquid flow rate in the start-up chamber based on the flow rate of the start-up chamber pipeline, the leakage rate, the piston rod moving speed, and the piston area of ​​the start-up chamber. Calculate the liquid flow rate in the shutdown chamber based on the flow rate of the shutdown chamber pipeline, leakage, piston rod moving speed, and piston area of ​​the shutdown chamber.

[0068] Specifically, a hydraulic cylinder is a hydraulic actuator that converts the pressure energy of a liquid into mechanical energy to drive a working mechanism to perform linear or oscillating movements. Structurally, hydraulic cylinders are divided into three main categories: piston type, plunger type, and rotary type. The servo motor controlled by the speed control system is structurally classified as a single-piston rod hydraulic cylinder.

[0069] A single-piston rod hydraulic cylinder, also known as a differential hydraulic cylinder, extends the piston rod when high-pressure oil enters the rodless chamber and causes the rod chamber to be under low pressure; when high-pressure oil enters the rod chamber and causes the rodless chamber to be under low pressure, the piston rod retracts; when a differential connection is used, i.e., the two chambers are connected, it becomes a differential cylinder.

[0070] Figure 3 This is a schematic diagram of the relay device structure provided by the present invention, as shown below. Figure 3 As shown, where Indicates the dead capacity of the relay's start-up chamber; Indicates the dead capacity of the relay shutdown chamber; Indicates the diameter of the relay piston; Indicates the diameter of the piston rod; Indicates the diameter of the start-up chamber of the relay; Indicates the diameter of the relay shutdown chamber; Indicates the angle between the relay device and the horizontal plane; Indicates the load mass of the relay; This indicates the total stroke of the relay.

[0071] Figure 4 This is a schematic diagram of the relay device provided by the present invention, as shown below. Figure 4 As shown, where Indicates the pressure in the start-up chamber pipeline; Indicates the pressure in the shutdown chamber pipeline; Indicates the piston area of ​​the starting chamber; Indicates the piston area of ​​the shutdown chamber; Indicates piston displacement; This refers to the pressure exerted on the piston by the starting chamber; This refers to the pressure exerted on the piston by the shut-off chamber. This indicates the piston rod load.

[0072] Calculate the piston area of ​​the start-up chamber :

[0073] Calculate the piston area of ​​the shutdown chamber :

[0074] When the relay is activated, after eliminating friction, the net force acting on the piston is:

[0075] Calculate the operating power of the relay :

[0076] Calculate the oil volume in the start-up chamber :

[0077] Calculate the oil volume in the shutdown chamber :

[0078] Calculate the leakage between the start-up chamber and the stop-down chamber. :

[0079] in It is expressed as the leakage coefficient between the two oil chambers.

[0080] Calculate the liquid flow rate in the start-up chamber :

[0081] in Expressed as flow rate in the start-up chamber pipe; This is expressed as the piston rod's moving speed.

[0082] Calculate the liquid flow rate in the shutdown chamber :

[0083] in This is expressed as the flow rate in the shutdown chamber pipe.

[0084] Based on the above embodiments, typical operating conditions include oil pump start-up and shutdown, oil pump rotation, and air replenishment valve start-up and shutdown.

[0085] Figure 5 This is a schematic diagram of the oil pump start / stop logic provided by the present invention, as shown below. Figure 5 As shown, when the system is in automatic operation, the oil pump starts when the oil pressure sensor detects a signal lower than the pump start pressure or when the pressure switch pump start node sends feedback to the PLC controller. At this time, the oil flows through the screw pump and the combination valve back to the return oil tank. Five seconds after the oil pump starts (this setting is adjustable), the combination valve loading solenoid valve is energized, and the oil flows through the oil pump and the combination valve to the pressure oil tank.

[0086] When the system is in automatic operation, the combined valve will unload when the oil pump pressure sensor detects a signal higher than the pump stop pressure or when the pressure switch stops the pump and sends feedback to the PLC controller. At this time, oil flows back to the return tank through the combined valve, stopping the supply of oil to the pressure tank. After a 5-second delay (adjustable), the oil pump stops operating.

[0087] There are two main types of oil pump rotation: 1. Without considering the oil pump running time, the two oil pumps take turns operating as the main oil pump.

[0088] 2. Considering the oil pump running time, the oil pump with the shortest total running time should be used as the main oil pump.

[0089] Figure 6 This is a schematic diagram of the start / stop logic of the air replenishment valve provided by the present invention, as shown below. Figure 6 As shown, when the system is running automatically, the automatic air replenishment device will activate when the tank pressure is lower than the set air replenishment start oil pressure and the tank oil level is higher than the set air replenishment start liquid level. Conversely, when the system is running automatically, the automatic air replenishment device will deactivate when the tank pressure is higher than the set air replenishment stop oil pressure or the tank oil level is lower than the set air replenishment stop liquid level.

[0090] Here, fault behavior includes faults and alarms. Faults include analog signal acquisition faults, oil pump motor faults, timeout faults, solenoid valve faults, communication faults, and power supply faults. Alarms include oil pressure alarms, oil level alarms, temperature alarms, and oil-water mixing alarms.

[0091] Analog signal acquisition faults include limit-over faults and jump faults. A limit-over fault occurs when the sampled value received by the analog module exceeds the set upper or lower limit, triggering a channel limit-over fault report. A jump fault occurs when the sampled value received by the analog module changes by a greater than the set value within a specified time, triggering a channel jump fault report. This fault applies to sensor-acquired signals, such as those from liquid level and oil pressure sensors.

[0092] Oil pump motor failure: When the motor soft starter alarms or the motor overheats, the oil pump will automatically switch to manual mode.

[0093] Timeout fault: When the oil pump or automatic air replenishment valve runs for longer than the set time, an oil pump or automatic air replenishment valve timeout fault will be reported.

[0094] Solenoid valve malfunction: When the PLC sends a control signal to activate a solenoid valve according to the program flow, and does not receive a feedback signal from the solenoid valve within a specified time, a solenoid valve malfunction will be reported.

[0095] Communication failure: No communication messages are received at the receiving end; parse the received messages and determine whether there is a communication failure based on the communication protocol.

[0096] Power failure: Add a relay to both ends of the power supply and connect the relay output signal to the PLC. When the output signal changes, it is judged as a power failure.

[0097] Oil pressure alarm: A high pressure alarm will be triggered when the pressure in the oil tank exceeds the set value. The alarm level is determined by the set value. A low pressure alarm will be triggered when the pressure in the oil tank falls below the set value. The alarm level is also determined by the set value.

[0098] Oil level alarm: A high oil level alarm will be triggered when the oil level in the pressure tank or return tank is higher than the set value. The alarm level is determined by the set value. A low oil level alarm will be triggered when the oil level in the pressure tank or return tank is lower than the set value. The alarm level is determined by the set value.

[0099] Temperature alarm: When the temperature of the return oil tank is higher than the set value, a high temperature alarm for the return oil tank will be generated.

[0100] Oil-water mixing alarm: When the oil-water mixing transmitter in the return oil tank alarms, an oil-water mixing alarm will be generated.

[0101] In other embodiments, when the system structure, oil type, or control strategy changes, the mechanism parameters of the physical simulation model are updated, and simulations of typical operating conditions and fault behaviors are performed based on the updated physical simulation model to obtain updated simulation data; the parameters of the fault diagnosis model are iterated based on the updated simulation data to obtain an updated fault diagnosis model.

[0102] Specifically, considering that the existing model update mechanism is highly dependent on manual intervention, when a new model, new oil or new operating strategy is introduced, the mechanism model parameters need to be recalibrated and the deep learning network needs to be retrained offline. This process requires domain experts and data scientists to work together and takes 2-3 weeks to complete, which cannot meet the development requirements of predictive maintenance for system self-evolution and rapid response.

[0103] To achieve automated online model updates, this invention introduces a digital twin-driven adaptive evolution mechanism: by collecting system operation data in real time, it automatically inverts and optimizes key mechanism parameters in the physical simulation model, such as oil elastic modulus, valve core friction coefficient, and servo valve flow gain, reducing reliance on manual calibration. Simultaneously, it constructs an incremental learning and online learning framework, enabling the fault diagnosis model to continuously fine-tune based on online-generated updated simulation data and real-time collected new samples, without requiring complete retraining. This significantly reduces computational overhead and update time, ultimately forming a fault diagnosis system with self-sensing and self-evolving capabilities, meeting the needs for agile response and predictive maintenance in practical engineering. When the system structure, oil type, or control strategy changes, only the mechanism parameters need to be adjusted at the simulation end to generate corresponding new fault data in batches. The network can quickly self-update through incremental learning, without the need to recalibrate a large number of physical experiments or expert rules as in traditional methods. This shortens the model iteration cycle from the "week" level to the "hour" level, truly realizing the online evolution of predictive maintenance.

[0104] In other embodiments, the fault diagnosis model trained by the above method can be used for real-time fault diagnosis of the hydraulic system of a water turbine. Figure 7 This is a flowchart illustrating the fault diagnosis method provided by the present invention, as shown below. Figure 7 As shown, a fault diagnosis method is provided, including: Step 710: Monitor key parameters in the hydraulic system of the turbine in real time, including pressure, flow rate, temperature and vibration. Step 720: Based on the trained fault diagnosis model, fault feature signals are extracted from the key parameters. Based on the extracted fault feature signals, the fault mode of the turbine hydraulic system is identified. The fault diagnosis model is obtained based on the speed regulation system fault diagnosis model construction method.

[0105] Specifically, the root causes of failures in the hydraulic systems of water turbines often involve the interaction of multiple factors, including mechanical, hydraulic, control, and environmental factors. Deterioration of hydraulic oil and excessive levels of contaminant particles accelerate component wear and cause jamming; aging of system seals or improper material selection can easily lead to leakage problems; prolonged high-load operation causes fatigue failure of valve components, thus affecting the stability of system flow and pressure; and control strategies failing to adapt to changes in workload in a timely manner may also lead to regulation failure and slow execution. Harsh environmental conditions such as high temperature and humidity, and severe dust pollution can also exacerbate the risk of system failure and shorten its service life. Therefore, this embodiment uses real-time monitoring of key parameters in the hydraulic system of the water turbine for fault diagnosis.

[0106] Let the set of key parameters of the hydraulic system be P={p,q,T,v}, where p represents pressure; q represents flow rate; T represents temperature; and v represents vibration.

[0107] These key parameters are monitored in real time, that is, P(t) is observed continuously over time t, where P(t)∈R 4 . Let P(t) represent a four-dimensional real space, where each dimension corresponds to a real-time monitored parameter. Since the parameter is a function of continuous time P(t), each parameter value comes from sensor readings, which are typically real numbers. A set of signal processing methods, M = {spectral analysis, wavelet transform, empirical mode decomposition}, is used to extract fault feature signals from the monitored parameter signal P(t), resulting in a fault feature signal set F(t). F(t) is expressed by the formula:

[0108] in Let |M| represent the i-th method in the set of signal processing methods M, and |M| represent the number of elements in set M (here |M|=3).

[0109] The extracted fault feature signals are fed into the fully connected layer and classifier of the model. The classifier calculates a probability vector, where each element represents the confidence level that the input data belongs to a certain fault mode or normal state. Fault modes can include mechanical faults, hydraulic faults, control faults, and environmental faults.

[0110] The fault diagnosis method provided in this embodiment can directly reuse the sensors (pressure, flow, temperature, vibration) and AMESim twin platform already deployed in existing units without additional hardware investment; the algorithm adopts a modular design (the physical constraint layer is pluggable), which can be seamlessly coupled with existing CNN, LSTM, Transformer and other networks, reducing the cost of on-site modification and personnel training, and has the conditions for rapid replication in the hydraulic subsystem of large hydropower stations.

[0111] The following describes the speed control system fault diagnosis model construction device provided by the present invention. The speed control system fault diagnosis model construction device described below and the speed control system fault diagnosis model construction method described above can be referred to in correspondence.

[0112] Figure 8 This is a schematic diagram of the structure of the speed control system fault diagnosis model construction device provided by the present invention, as shown below. Figure 8 As shown, the device includes: Simulation unit 810 is used to construct a physical simulation model of the hydraulic system of the water turbine, and to simulate typical working conditions and fault behaviors based on the physical simulation model to obtain simulation data. The data generation unit 820 is used to generate synthetic data based on real data and the simulation data using a deep learning algorithm. The model training unit 830 is used to train the initial model using the simulation data, real data and synthetic data as the training set to obtain the fault diagnosis model.

[0113] Based on the above embodiments, the simulation unit is specifically used for: Mathematical modeling is performed on the working mechanism of oil pumps, combination valves, and relays in the hydraulic system of a water turbine. The typical operating conditions include oil pump start-up and shutdown, oil pump rotation, and air supply valve start-up and shutdown. The fault behaviors include faults and alarms. The faults include analog signal acquisition faults, oil pump motor faults, timeout faults, solenoid valve faults, communication faults, and power supply faults. The alarms include oil pressure alarms, oil level alarms, temperature alarms, and oil-water mixing alarms.

[0114] Based on the above embodiments, the simulation unit is specifically used for: The pitch is calculated based on the lead and number of helical blades. The flow rate of the oil pump is calculated based on the pitch, the diameter of the helical blades, the diameter of the pump shaft, the water lifting cross-sectional area ratio, and the rotational speed. The shaft power is calculated based on the oil pump's flow rate, liquid lift height, liquid density, and oil pump efficiency.

[0115] Based on the above embodiments, the simulation unit is specifically used for: When the relay is activated, the net piston force is calculated based on the piston rod load, the pressure exerted on the piston by the opening chamber, the piston area of ​​the opening chamber, the pressure exerted on the piston by the closing chamber, the piston area of ​​the closing chamber, and the angle between the relay and the horizontal plane. Based on the net piston force and the total stroke of the relay, the operating work of the relay is calculated. Calculate the starting oil volume based on the piston area of ​​the starting chamber, piston displacement, and dead capacity of the starting chamber of the relay; calculate the oil volume of the stopping chamber based on the piston area of ​​the stopping chamber, total stroke of the relay, piston displacement, and dead capacity of the stopping chamber of the relay. Based on the pressure exerted on the piston by the start-up chamber, the pressure exerted on the piston by the stop-up chamber, and the preset leakage coefficient, the leakage between the start-up chamber and the stop-up chamber is calculated. The liquid flow rate in the start-up chamber is calculated based on the flow rate of the start-up chamber pipeline, the leakage amount, the piston rod moving speed, and the piston area of ​​the start-up chamber. The liquid flow rate in the shutdown chamber is calculated based on the flow rate in the shutdown chamber pipe, the leakage amount, the piston rod moving speed, and the piston area of ​​the shutdown chamber.

[0116] Based on the above embodiments, the device further includes an updating unit, used for: When the system structure, oil type, or control strategy changes, the mechanism parameters of the physical simulation model are updated, and simulations of typical operating conditions and fault behaviors are performed based on the updated physical simulation model to obtain updated simulation data. Based on the updated simulation data, the fault diagnosis model is iterated to obtain an updated fault diagnosis model.

[0117] Figure 9 This is a schematic diagram of the fault diagnosis device provided by the present invention, as shown below. Figure 9As shown, the device includes: The parameter monitoring unit 910 is used to monitor key parameters in the hydraulic system of the water turbine in real time, including pressure, flow rate, temperature and vibration. The fault identification unit 920 is used to extract fault feature signals from the key parameters based on the trained fault diagnosis model, and to identify the fault mode of the turbine hydraulic system based on the extracted fault feature signals. The fault diagnosis model is obtained based on the speed regulation system fault diagnosis model construction method.

[0118] Figure 10 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 10 As shown, the electronic device may include: a processor 1010, a communication interface 1020, a memory 1030, and a communication bus 1040. The processor 1010, communication interface 1020, and memory 1030 communicate with each other via the communication bus 1040. The processor 1010 can call logical instructions in the memory 1030 to execute a speed regulation system fault diagnosis model construction method or a fault diagnosis method. The speed regulation system fault diagnosis model construction method includes: constructing a physical simulation model of the turbine hydraulic system; simulating typical operating conditions and fault behaviors based on the physical simulation model to obtain simulation data; generating synthetic data using a deep learning algorithm based on real data and the simulation data; and training an initial model using the simulation data, real data, and synthetic data as a training set to obtain the fault diagnosis model.

[0119] The fault diagnosis method includes: real-time monitoring of key parameters in the hydraulic system of the turbine, including pressure, flow rate, temperature and vibration; based on a trained fault diagnosis model, extracting fault feature signals from the key parameters; and based on the extracted fault feature signals, identifying the fault mode of the hydraulic system of the turbine. The fault diagnosis model is obtained based on the fault diagnosis model construction method for speed regulation systems.

[0120] Furthermore, the logical instructions in the aforementioned memory 1030 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0121] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the speed regulation system fault diagnosis model construction method or fault diagnosis method provided by the above methods. The speed regulation system fault diagnosis model construction method includes: constructing a physical simulation model of a hydraulic turbine system; simulating typical working conditions and fault behaviors based on the physical simulation model to obtain simulation data; generating synthetic data based on real data and the simulation data using a deep learning algorithm; and training an initial model using the simulation data, real data, and synthetic data as a training set to obtain the fault diagnosis model.

[0122] The fault diagnosis method includes: real-time monitoring of key parameters in the hydraulic system of the turbine, including pressure, flow rate, temperature and vibration; based on a trained fault diagnosis model, extracting fault feature signals from the key parameters; and based on the extracted fault feature signals, identifying the fault mode of the hydraulic system of the turbine. The fault diagnosis model is obtained based on the fault diagnosis model construction method for speed regulation systems.

[0123] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the speed regulation system fault diagnosis model construction method or fault diagnosis method provided by the above methods. The speed regulation system fault diagnosis model construction method includes: constructing a physical simulation model of a hydraulic turbine system; simulating typical working conditions and fault behaviors based on the physical simulation model to obtain simulation data; generating synthetic data based on real data and the simulation data using a deep learning algorithm; and training an initial model using the simulation data, real data, and synthetic data as a training set to obtain the fault diagnosis model.

[0124] The fault diagnosis method includes: real-time monitoring of key parameters in the hydraulic system of the turbine, including pressure, flow rate, temperature and vibration; based on a trained fault diagnosis model, extracting fault feature signals from the key parameters; and based on the extracted fault feature signals, identifying the fault mode of the hydraulic system of the turbine. The fault diagnosis model is obtained based on the fault diagnosis model construction method for speed regulation systems.

[0125] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0126] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0127] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a fault diagnosis model for a speed control system, characterized in that, include: A physical simulation model of the hydraulic system of the water turbine is constructed, and simulations of typical working conditions and fault behaviors are performed based on the physical simulation model to obtain simulation data; Based on real data and the simulation data, synthetic data is generated using deep learning algorithms; The simulation data, real data, and synthetic data are used as the training set to train the initial model, thus obtaining the fault diagnosis model.

2. The method for constructing a fault diagnosis model for a speed control system according to claim 1, characterized in that, The physical simulation model of the hydraulic turbine system includes: Mathematical modeling is performed on the working mechanism of oil pumps, combination valves, and relays in the hydraulic system of a water turbine. The typical operating conditions include oil pump start-up and shutdown, oil pump rotation, and air supply valve start-up and shutdown. The fault behaviors include faults and alarms. The faults include analog signal acquisition faults, oil pump motor faults, timeout faults, solenoid valve faults, communication faults, and power supply faults. The alarms include oil pressure alarms, oil level alarms, temperature alarms, and oil-water mixing alarms.

3. The method for constructing a fault diagnosis model for a speed control system according to claim 2, characterized in that, A mathematical model is performed on the working mechanism of the oil pump in the hydraulic system of the water turbine, including: The pitch is calculated based on the lead and number of helical blades. The flow rate of the oil pump is calculated based on the pitch, the diameter of the helical blades, the diameter of the pump shaft, the water lifting cross-sectional area ratio, and the rotational speed. The shaft power is calculated based on the oil pump's flow rate, liquid lift height, liquid density, and oil pump efficiency.

4. The method for constructing a fault diagnosis model for a speed control system according to claim 2, characterized in that, Mathematical modeling of the working mechanism of the servo in the hydraulic system of a water turbine is performed, including: When the relay is activated, the net piston force is calculated based on the piston rod load, the pressure exerted on the piston by the opening chamber, the piston area of ​​the opening chamber, the pressure exerted on the piston by the closing chamber, the piston area of ​​the closing chamber, and the angle between the relay and the horizontal plane. Based on the net piston force and the total stroke of the relay, the operating work of the relay is calculated. Calculate the starting oil volume based on the piston area of ​​the starting chamber, piston displacement, and dead capacity of the starting chamber of the relay; calculate the oil volume of the stopping chamber based on the piston area of ​​the stopping chamber, total stroke of the relay, piston displacement, and dead capacity of the stopping chamber of the relay. Based on the pressure exerted on the piston by the start-up chamber, the pressure exerted on the piston by the stop-up chamber, and the preset leakage coefficient, the leakage between the start-up chamber and the stop-up chamber is calculated. The liquid flow rate in the start-up chamber is calculated based on the flow rate of the start-up chamber pipeline, the leakage amount, the piston rod moving speed, and the piston area of ​​the start-up chamber. The liquid flow rate in the shutdown chamber is calculated based on the flow rate in the shutdown chamber pipe, the leakage amount, the piston rod moving speed, and the piston area of ​​the shutdown chamber.

5. The method for constructing a fault diagnosis model for a speed control system according to claim 1, characterized in that, After obtaining the fault diagnosis model, the method further includes: When the system structure, oil type, or control strategy changes, the mechanism parameters of the physical simulation model are updated, and simulations of typical operating conditions and fault behaviors are performed based on the updated physical simulation model to obtain updated simulation data. Based on the updated simulation data, the fault diagnosis model is iterated to obtain an updated fault diagnosis model.

6. A fault diagnosis method, characterized in that, include: Real-time monitoring of key parameters in the hydraulic system of the water turbine, including pressure, flow rate, temperature, and vibration; Based on the trained fault diagnosis model, fault feature signals are extracted from the key parameters. Based on the extracted fault feature signals, the fault mode of the turbine hydraulic system is identified. The fault diagnosis model is obtained based on the speed regulation system fault diagnosis model construction method according to any one of claims 1 to 5.

7. A device for constructing a fault diagnosis model for a speed control system, characterized in that, include: The simulation unit is used to construct a physical simulation model of the hydraulic system of the water turbine, and to simulate typical working conditions and fault behaviors based on the physical simulation model to obtain simulation data. A data generation unit is used to generate synthetic data based on real data and the simulation data using a deep learning algorithm. The model training unit is used to train the initial model using the simulation data, real data, and synthetic data as the training set to obtain the fault diagnosis model.

8. A fault diagnosis device, characterized in that, include: The parameter monitoring unit is used to monitor key parameters in the hydraulic system of the water turbine in real time, including pressure, flow rate, temperature and vibration. The fault identification unit is used to extract fault feature signals from the key parameters based on the trained fault diagnosis model, and to identify the fault mode of the turbine hydraulic system based on the extracted fault feature signals. The fault diagnosis model is obtained based on the speed regulation system fault diagnosis model construction method according to any one of claims 1 to 5.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the speed control system fault diagnosis model construction method as described in any one of claims 1 to 5, or the fault diagnosis method as described in claim 6.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the speed control system fault diagnosis model construction method as described in any one of claims 1 to 5, or the fault diagnosis method as described in claim 6.

Images