Distributed magnetic coupling water turbine guide vane opening adjusting mechanism and adjusting method

Abstract

This invention relates to a distributed magnetically coupled turbine guide vane opening adjustment mechanism and method, comprising an adjustment workstation, a PLC main controller, a communication network, and several guide vane drive units. The adjustment workstation includes a data acquisition module, a data preprocessing module, a working condition identification module, and an opening optimization decision module. Each guide vane drive unit includes a servo motor, a magnetic coupling, and an absolute encoder. The absolute encoder is mounted on the guide vane's shaft and is used to detect the guide vane's rotation angle and feed it back to the PLC main controller and the adjustment workstation. The distributed layout of this invention allows each drive unit to operate independently, resulting in a high degree of modularity. The ring network based on bus communication provides redundant paths. Combined with a machine learning intelligent control system possessing self-sensing and self-diagnostic capabilities, when a single drive unit fails, the system can quickly identify the problem and dynamically compensate by adjusting adjacent guide vanes, ensuring the unit's continuous, safe, and stable operation.

Description

Technical Field

[0001] This invention relates to the field of turbine speed control technology, specifically to a distributed magnetic coupling turbine guide vane opening adjustment mechanism and adjustment method. Background Technology

[0002] The turbine is the core equipment of a hydropower station, and the guide vane mechanism is an important control system of the turbine. The guide vane mechanism drives the movable guide vanes to rotate, thereby creating gaps and opening angles of different degrees between the guide vanes. This creates and changes the circulation of water entering the runner, allowing the water to enter the runner chamber evenly along the circumference with minimal energy loss, driving the runner to rotate and generate electricity, thus ensuring that the turbine has good hydraulic characteristics.

[0003] Traditional mechanically controlled guide vanes suffer from clearances and elastic deformation in the transmission chain, potentially leading to asynchronous movements during adjustment. This can affect the uniformity of circumferential water intake, causing unit vibration and energy loss. Furthermore, the complex structure of the hydraulic servo device poses a risk of seal failure, significantly increasing equipment maintenance requirements and operating costs. In addition, the rigid connection structure limits the independent adjustment capability of the guide vanes, hindering differentiated control of each vane's opening and restricting further improvements in the turbine's hydraulic performance. Meanwhile, existing control methods largely rely on preset models and fixed parameters, making them ill-suited to varying operating conditions such as head, flow rate, and sediment content. They lack adaptive optimization capabilities based on real-time data and historical operating experience, resulting in significant limitations in improving overall unit efficiency, suppressing hydraulic vibration, and achieving intelligent operation and maintenance.

[0004] Eliminating the risk of dynamic seal leakage while ensuring high synchronization accuracy, and providing each guide vane with independent adjustable flexibility, has become a pressing technical challenge for improving the overall performance and reliability of hydro turbines. Furthermore, with the development of smart power plants and digital twin technology, integrating intelligent methods such as big data analysis and machine learning into the guide vane control closed loop to achieve intelligent regulation through self-sensing of operating conditions, self-optimization of strategies, and self-diagnosis of anomalies has also become an important development direction in this field. Therefore, it is urgent to propose a novel guide vane opening adjustment scheme to improve the accuracy, reliability, and intelligence level of hydro turbine control. Summary of the Invention

[0005] The purpose of this invention is to solve the above-mentioned technical problems existing in the prior art by providing a distributed magnetic coupling turbine guide vane opening adjustment mechanism and adjustment method.

[0006] To address the shortcomings of the aforementioned technical problems, the present invention adopts the following technical solution: a distributed magnetic coupling turbine guide vane opening adjustment mechanism, comprising an adjustment workstation, a PLC main controller, a communication network, and several guide vane drive units. The adjustment workstation includes a data acquisition module, a data preprocessing module, a working condition identification module, and an opening degree optimization decision module; The PLC main controller is connected to the guide vane drive unit via a communication network; The guide vane drive unit includes a servo motor, a magnetic coupling coupling, and an absolute encoder. The magnetic coupling coupling includes an external permanent magnet, a sealing partition, and an internal permanent magnet. The external permanent magnet is connected to the output shaft of the servo motor, and the internal permanent magnet is connected to the rotating shaft of the guide vane. The sealing partition is fixed to the valve seat or the housing of the flow regulating device, separating the external permanent magnet from the internal permanent magnet to form a contactless torque transmission. The absolute encoder is set on the rotating shaft of the guide vane and is used to detect the rotation angle of the guide vane and feed it back to the PLC main controller and the regulating workstation.

[0007] As a further optimization of the distributed magnetic coupling turbine guide vane opening adjustment mechanism of the present invention: the servo motor is a hollow shaft brushless motor, and the output shaft of the servo motor is connected to an external permanent magnet through a flange.

[0008] As a further optimization of the distributed magnetic coupling turbine guide vane opening adjustment mechanism of the present invention: the sealing partition of the magnetic coupling is made of non-magnetic corrosion-resistant material, and the external permanent magnet and the internal permanent magnet of the magnetic coupling are made of neodymium iron boron or samarium cobalt permanent magnet materials.

[0009] As a further optimization of the distributed magnetic coupling turbine guide vane opening adjustment mechanism of the present invention: the communication network is an EtherCAT or CAN ring network.

[0010] A method for adjusting the guide vane opening of a distributed magnetically coupled hydro turbine includes: The data acquisition module obtains the head H, flow rate Q, unit active power P, rotational speed n, and actual guide vane opening α. i Motor current / torque T of each drive unit i And real-time data of vibration / pressure pulsation characteristic value V; The data preprocessing module performs filtering, anomaly removal, and normalization on the real-time data and constructs a feature vector X(t); The working condition identification module uses machine learning to output the working condition category k or working condition parameter θ based on X(t); The turbine opening optimization decision module aims to maximize turbine hydraulic efficiency and minimize vibration / pulsation. Under constraints of synchronization deviation, torque upper limit, and opening change rate upper limit, it outputs control variables, including the global target opening α. ref and the differential compensation amount Δα for each guide vane i ; Output the target opening α of each guide vane i cmd αi cmd = α ref + Δα i The PLC main controller distributes the output to each drive unit for execution. When any guide vane is detected to have encoder abnormality, execution error exceeding the threshold, drive unit current / torque abnormality, or vibration / pressure pulsation exceeding the limit, the differential compensation amount Δαi is stopped for the abnormal guide vane and the control is switched to opening given by characteristic curve and PID closed-loop control. The remaining guide vanes are kept under synchronous control or the differential compensation amount is redistributed. The adjustment workstation updates the model or its parameters based on the feedback data after execution, writes the feature vector, target command, execution error, efficiency estimate and vibration / pressure pulsation feature value of each control cycle into the historical database, and incrementally updates the machine learning model or its parameters according to a preset cycle or triggering condition to form an adaptive closed-loop optimization.

[0011] The method for adjusting the guide vane opening of a distributed magnetically coupled turbine as described in claim 5 is characterized in that: the feature vector X(t) includes, but is not limited to, the concatenation of the following three sub-vectors: The basic operating state sub-vector contains the unit's macroscopic operating parameters and is used to determine the unit's approximate position on the characteristic curve; Guide vane distribution and feedback sub-vectors, which include the microstate of the guide vane system and the system's response feedback, are used to capture the effects of asynchronous regulation; Dynamic trend subvectors, containing the first derivative of key parameters with respect to time, are used to identify transient processes in operating conditions.

[0012] As a further optimization of the distributed magnetic coupling turbine guide vane opening adjustment method of the present invention, the anomaly determination includes, but is not limited to, the following logical rules: Encoder anomaly detection: The data packets fed back by the encoder fail the CRC check N times consecutively; Execution error judgment: If the guide vane angle execution error is greater than the dynamic threshold and the duration is greater than the tolerance time, the actuator is judged to be stuck or the precision is failed. Drive unit abnormality judgment rules: Motor current > 1.2 times rated current for 5 seconds, motor current suddenly drops to no-load current level; Vibration / pressure pulsation exceeding limit judgment rule: V > V warn This triggers a switch in the control strategy.

[0013] The characteristic curve is a preset safe operating envelope. The characteristic curve defines the recommended guide vane opening under different heads to ensure that the unit does not experience cavitation and severe vibration.

[0014] The redistribution specifically involves the following: when a guide vane fails, redistribution is performed using an adjacent compensation method. The flow field adjustment tasks originally assigned to the failed guide vane R are redistributed to its adjacent guide vanes R according to an inverse weight. +1 and R -1 .

[0015] The incremental update specifically refers to: The system maintains an experience playback buffer, which combines the input vector X(t) and control command u generated in each control cycle with the above feedback data into a new training sample and stores it in the buffer. Set a fixed period to update the model using the new sample set generated that day, or update it immediately when an abnormal event is triggered; The gradient descent method is used to calculate the gradient of the loss function with new samples in the buffer. Only the weights and biases of the last few layers of the network are updated by backpropagation with small steps, so as to keep the feature extraction capability of the bottom layer unchanged. The updated model is validated on the validation set.

[0016] The present invention has the following beneficial effects: 1. The distributed layout of this invention allows each drive unit to operate independently, resulting in a high degree of modularity. Maintenance and replacement of a single guide vane drive unit does not require disassembling the entire synchronization mechanism, significantly reducing maintenance costs and downtime. The use of a magnetic coupling for non-contact torque transmission eliminates the need for a dynamic seal structure, fundamentally eliminating the risk of shaft end leakage and significantly improving system reliability and service life.

[0017] 2. This invention equips each guide vane with an independent servo drive and a high-resolution absolute encoder. Combined with electronic synchronization algorithms and intelligent adjustment strategies based on big data and machine learning, it achieves precise closed-loop control and dynamic optimization of the guide vane opening. The synchronization accuracy and adaptability to working conditions are far superior to traditional mechanical methods.

[0018] 3. The ring network based on bus communication in this invention provides redundant paths. Combined with a machine learning intelligent control system with self-sensing and self-diagnosis capabilities, when a single drive unit fails, the system can quickly identify the fault and dynamically compensate by adjusting adjacent guide vanes, ensuring the continuous, safe and stable operation of the unit. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall structure of the turbine guide vane opening adjustment mechanism of the present invention; Figure 2 This is a block diagram of the guide vane drive unit structure of the present invention; Figure 3 This is a structural block diagram of the magnetic coupling of the present invention; Figure 4This is a schematic diagram of the guide vane drive unit of the present invention; Figure 5 This is a schematic flowchart of the turbine guide vane opening adjustment method of the present invention; Marked in the image: 1. Servo motor; 2. Drive shaft; 3. Bevel gears; 4. External permanent magnet; 5. Sealing partition; 6. Internal permanent magnet; 7. Absolute encoder; 8. Guide vanes. Detailed Implementation

[0020] To better understand the present invention, the following embodiments further illustrate the content of the present invention, but the content of the present invention is not limited to the following embodiments.

[0021] like Figure 1-4 As shown: A distributed magnetic coupling turbine guide vane opening adjustment mechanism is arranged around the turbine volute and includes a guide vane opening adjustment workstation, a PLC main controller, an EtherCAT ring communication network, and several drive units with identical structures, each drive unit independently driving one guide vane.

[0022] The PLC main controller is equipped with an EtherCAT master station interface, supporting the connection of multiple slave nodes to meet the requirements of the number of guide vanes in medium-sized hydro turbines. The PLC has a built-in real-time control kernel with an instruction execution cycle of ≤1ms, which can quickly receive optimization instructions from the adjustment workstation and send control signals to each drive unit, while simultaneously acquiring feedback data from the absolute encoder in real time.

[0023] The EtherCAT ring communication network employs a dual-redundancy design, using shielded twisted-pair cables. The ring topology ensures that the failure of a single node does not affect overall communication stability, meeting the synchronous control requirements of multiple drive units. Each drive unit is equipped with an EtherCAT slave module, forming a closed-loop communication with the PLC master controller to achieve real-time interaction between command issuance and status feedback.

[0024] The guide vane opening adjustment workstation includes a data acquisition interface, a data preprocessing module, a working condition identification module, and an opening optimization decision module. The drive unit includes a hollow shaft brushless servo motor 1, a drive shaft 2, a bevel gear 3, a magnetic coupling coupling, and an absolute encoder 7. The magnetic coupling coupling is as follows: Figure 3As shown, the system includes an external permanent magnet 4, a sealing partition 5, and an internal permanent magnet 6. The external permanent magnet 4 is connected to the output shaft of the servo motor, and the internal permanent magnet 6 is connected to the guide vane 8 via a rotating shaft. Both the external permanent magnet 4 and the internal permanent magnet 6 are made of neodymium iron boron or samarium cobalt permanent magnet materials. The sealing partition 5 is made of non-magnetic, corrosion-resistant stainless steel.

[0025] A sealing partition 5 is fixed to the valve seat or the housing of the flow regulating device, separating the external permanent magnet 4 from the internal permanent magnet 6 to form a contactless torque transmission. An absolute encoder 7 is mounted on the guide vane shaft to detect the rotation angle of the guide vane 8. The PLC main controller is connected to the servo motor 1 and the absolute encoder 7 of each drive unit through the communication network. The guide vane opening adjustment workstation is connected to the PLC main controller, has a built-in adaptive big data model trained based on historical and real-time data, and uses machine learning algorithms to dynamically generate optimized opening adjustment commands.

[0026] <Example 2> like Figure 5 As shown: A method for adjusting the guide vane opening of a distributed magnetically coupled hydro turbine, comprising the following: S1. Data Acquisition and Preprocessing: Acquire head H, flow rate Q, unit active power P, rotational speed n, and actual guide vane opening α. i Motor current / torque T of each drive unit i Vibration / pressure pulsation characteristic value V; filtering, outlier removal, and normalization of the data; forming the feature vector X(t).

[0027] The vibration / pressure pulsation characteristic value V is obtained by performing analog-to-digital conversion on the raw time-domain waveform data collected by the sensor, extracting the amplitude of a specific frequency band using fast Fourier transform, and calculating it in combination with the root mean square value or peak value of the time-domain signal. This is used to quantitatively characterize the hydraulic stability of the unit.

[0028] Specifically, vibration signals xvib(t) are collected using an accelerometer installed on the turbine top cover / guide bearing, and pressure pulsation signals xpres(t) are collected using a dynamic pressure sensor installed on the tailrace / volute. The collected discrete-time series data are processed for detrending and low-pass filtered to eliminate zero drift and high-frequency electromagnetic interference noise.

[0029] The eigenvector X(t) is a multidimensional vector describing the complete dynamic state of the turbine unit at time t. It can be, but is not limited to, a concatenation of the following three sub-vectors: (1) Basic operating state sub-vectors, which contain the macroscopic operating parameters of the unit, are used to determine the approximate position of the unit on the characteristic curve, such as normalized head, active power, speed, flow rate, etc.

[0030] (2) Guide vane distribution and feedback sub-vectors, which include the micro-state of the guide vane system and the system response feedback, are used to capture the effects of asynchronous adjustment, such as the average value and standard deviation of the guide vane opening, the average value of the drive unit torque, and the vibration / pressure pulsation characteristic value.

[0031] (3) Dynamic trend sub-vectors, which contain the first derivative of key parameters with respect to time, are used to identify transient processes in the working conditions, such as power change rate, head change rate, etc.

[0032] S2. Operating condition identification: Based on X(t), output the operating condition category k or operating condition parameter θ, such as head range, load range, cavitation risk level, etc.

[0033] Support vector machine and Gaussian mixture model are used as classification algorithms for the division of discrete working condition categories k; multilayer perceptron and memory network are selected for regression estimation of continuous working condition parameters θ.

[0034] The model receives the feature vector X(t) as input and maps the current turbine operating state to discrete operating condition categories k. For example: k=1, optimal operating zone; k=2, partial load zone; k=3, high load / overload zone; k=4, cavitation / vibration hazard zone, etc. Operating condition parameters θ can be used to quantify current potential risks or performance boundaries, such as cavitation safety factor, hydraulic stability safety factor, etc.

[0035] S3. Optimization solution for opening degree: Using the objective function J = w1·(-η) + w2·V + w3·Σ|α i -α ref For multi-objective / weighted objectives, α is obtained under constraints such as synchronization deviation, torque upper limit, and rate of change upper limit. ref With Δα i (The goal is to maximize the hydraulic efficiency of the turbine and minimize vibration / pulsation, and to output control quantities under constraints such as synchronization deviation, upper limit of torque, and upper limit of opening change rate. The control quantities include the global target opening α.) ref and the differential compensation amount Δα for each guide vane i ).

[0036] Maximum efficiency refers to the hydraulic efficiency of the turbine unit. The two indicators of maximum efficiency and minimum vibration / pulsation constitute a multi-objective optimization relationship that combines competition and cooperation. Priority dynamically switches according to operating conditions: efficiency takes precedence under normal operating conditions; safety takes precedence under hazardous operating conditions, with minimum vibration being the next priority.

[0037] Synchronization deviation refers to the actual opening α of any single guide vane. ref The absolute value of the difference between the average opening of all guide vanes and the average opening of all guide vanes is used to control the synchronization deviation and prevent the excessive difference in opening of adjacent guide vanes from causing serious water leakage between stages or generating lateral water thrust, so that the guide vanes maintain macroscopic synchronization and microscopic differences.

[0038] The torque limit refers to the maximum torque limit of a single drive unit's motor. It limits the individual motor to prevent burnout or to prevent exceeding the maximum transmission torque of the magnetic coupler, which could lead to demagnetization and slippage.

[0039] The upper limit of the opening change rate refers to the global opening change rate, which limits the rate at which all guide vanes close or open as a whole. If they close too quickly, it will cause destructive water hammer pressure in the water intake pipe.

[0040] According to the formula J = w1·(-η) + w2·V + w3·Σ|α i -α ref The optimization module dynamically adjusts the weighting coefficients w1, w2, w3, and constraints in J based on the output operating condition category k and parameter θ, achieving adaptive switching of the control strategy. For example, when the operating condition is k=1, the system determines that it is safe at this time, and thus adjusts the strategy to increase the efficiency weight w1 and decrease the vibration weight w2, so that the guide vane opening reaches the theoretical maximum efficiency point as much as possible. When the operating condition is k=4, or when parameter θ indicates a high risk of cavitation, the system strategy adjusts to increase the vibration weight w2, sacrificing some efficiency to find the guide vane opening combination with the least vibration, suppressing vortex bands or resonance.

[0041] The search space of the optimization algorithm is defined as an N+1 dimensional vector u, u=[α ref， Δα 1， Δα 2，...... Δα n ], α ref As the reference opening, Δα i This is the fine-tuning compensation amount for each guide vane.

[0042] During the solution process, any candidate solution u must simultaneously satisfy the following physical constraints; otherwise, it will be deemed an invalid solution. For example: physical limit constraint α. ref + Δα i ≤α max Synchronization deviation constraint Δα i ≤δ limit (e.g., ≤±3°); Actuator capability constraint T i (u) ≤T rated (Motor rated torque), etc.

[0043] A hybrid architecture combining XGBoost and multilayer perceptron is employed. M candidate control vector particle swarms are randomly generated. Each candidate vector u is input into a surrogate model, which predicts the unit's efficiency η and eigenvalue V under given conditions. The prediction results are substituted into the aforementioned dynamic objective function J to calculate the fitness of the current particle. According to the particle swarm algorithm rules, the position and velocity of each particle are updated, converging towards the direction of minimum J (i.e., maximum efficiency and minimum vibration / pulsation, while satisfying constraints on synchronization deviation, torque upper limit, and opening change rate upper limit). When the number of iterations reaches the upper limit or J no longer changes significantly, the optimal particle position at this point is output, which is the final α. ref With Δα i .

[0044] S4. Command Issuance and Execution: Forming α i cmd = α ref + Δα i Limiting parameters such as speed, acceleration, and torque are distributed by the PLC to each drive unit, which then drives the corresponding guide vanes to rotate via magnetic couplings.

[0045] S5. Closed-loop evaluation and feedback: Calculate the execution error e i =α i -α i cmd Efficiency estimation η, motor load T i Feedback quantities such as vibration characteristics V are used to trigger protection or degradation strategies if they exceed the limits.

[0046] The target opening of each guide vane output is α. i cmd = α ref + Δα i The PLC main controller distributes the data to each drive unit for execution. When any guide vane is detected to have an encoder malfunction, execution error exceeding a threshold, abnormal current / torque in the drive unit, or excessive vibration / pressure pulsation, the differential compensation amount Δα is stopped from being sent to the abnormal guide vane. i It then switches to characteristic curve-based opening setting and PID closed-loop control, and maintains synchronous control for the remaining guide vanes or redistributes the difference compensation amount.

[0047] Anomaly detection can include, but is not limited to, the following specific logical rules and dynamic thresholds for real-time monitoring.

[0048] Encoder anomaly detection: The data packets fed back by the encoder fail the CRC check N times consecutively.

[0049] Execution error determination: If the guide vane angle execution error is greater than the dynamic threshold and the duration is greater than the tolerance time, the actuator is determined to be stuck or its accuracy is not working.

[0050] Drive unit abnormality judgment rules: Motor current > 1.2 times rated current for 5 seconds (overload), motor current suddenly drops to no-load current level (decoupling).

[0051] Vibration / pressure pulsation exceeding limit judgment rule: V > V warn (e.g., 200um) triggers a switch in the control strategy.

[0052] The characteristic curve is a preset safe operating envelope, which defines the recommended guide vane opening at different water heads to ensure that the unit does not experience cavitation and severe vibration.

[0053] When a guide vane fails, the flow field adjustment task originally assigned to the failed guide vane R is redistributed using an adjacent compensation method. The task is then redistributed to its adjacent guide vane R with an inverse weight. +1 and R -1 .

[0054] S6. Online Update: (X(t), α) i cmd The parameters (η, V, e) are written into the historical database, and the model parameters are updated according to a set cycle. (The adjustment workstation updates the model or its parameters based on the feedback data after execution, writes the feature vector, target command, execution error, efficiency estimate and vibration / pressure pulsation feature value of each control cycle into the historical database, and incrementally updates the machine learning model or its parameters according to a preset cycle or trigger condition to form an adaptive closed-loop optimization).

[0055] Feedback data includes hydraulic efficiency η, active power P, vibration / pulsation characteristic vector V, actual guide vane opening, motor current and torque, etc.

[0056] Incremental updates are performed as follows: The system maintains an experience replay buffer. The input vector X(t) generated in each control cycle, the control command u, and the aforementioned feedback data are combined to form a new training sample, which is then stored in the buffer. A fixed period is set, and the model is updated using the new sample set generated that day, or immediately upon triggering an abnormal event. Gradient descent is used to calculate the gradient of the loss function using the new samples in the buffer. Only the weights and biases of the last few layers of the network are updated via small-step backpropagation, maintaining the underlying feature extraction capability. The updated model is then validated on a validation set.

[0057] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various modifications or variations within the scope of the claims, which do not affect the essence of the present invention.

Claims

1. A distributed magnetic coupling turbine guide vane opening adjustment mechanism, characterized in that, It includes a regulating workstation, a PLC main controller, a communication network, and several guide vane drive units; The adjustment workstation includes a data acquisition module, a data preprocessing module, a working condition identification module, and an opening degree optimization decision module; The PLC main controller is connected to the guide vane drive unit via a communication network; The guide vane drive unit includes a servo motor, a magnetic coupling coupling, and an absolute encoder. The magnetic coupling coupling includes an external permanent magnet, a sealing partition, and an internal permanent magnet. The external permanent magnet is connected to the output shaft of the servo motor, and the internal permanent magnet is connected to the rotating shaft of the guide vane. The sealing partition is fixed to the valve seat or the housing of the flow regulating device, separating the external permanent magnet from the internal permanent magnet to form a contactless torque transmission. The absolute encoder is set on the rotating shaft of the guide vane and is used to detect the rotation angle of the guide vane and feed it back to the PLC main controller and the regulating workstation.

2. The distributed magnetic coupling turbine guide vane opening adjustment mechanism as described in claim 1, characterized in that: The servo motor is a hollow shaft brushless motor, and the output shaft of the servo motor is connected to an external permanent magnet through a flange.

3. The distributed magnetic coupling turbine guide vane opening adjustment mechanism as described in claim 1, characterized in that: The sealing diaphragm of the magnetic coupling is made of non-magnetic and corrosion-resistant material, and the external and internal permanent magnets of the magnetic coupling are made of neodymium iron boron or samarium cobalt permanent magnet materials.

4. The distributed magnetic coupling turbine guide vane opening adjustment mechanism as described in claim 1, characterized in that: The communication network is an EtherCAT or CAN ring network.

5. A method for adjusting the guide vane opening of a distributed magnetically coupled turbine, based on the guide vane opening adjustment mechanism described in claim 1, characterized in that, include: The data acquisition module obtains the head H, flow rate Q, unit active power P, rotational speed n, and actual guide vane opening α. i Motor current / torque T of each drive unit i And real-time data of vibration / pressure pulsation characteristic value V; The data preprocessing module performs filtering, anomaly removal, and normalization on the real-time data and constructs a feature vector X(t); The working condition identification module uses machine learning to output the working condition category k or working condition parameter θ based on X(t); The turbine opening optimization decision module aims to maximize turbine hydraulic efficiency and minimize vibration / pulsation. Under constraints of synchronization deviation, torque upper limit, and opening change rate upper limit, it outputs control variables, including the global target opening α. ref and the differential compensation amount Δα for each guide vane i ; Output the target opening α of each guide vane i cmd α i cmd = α ref + Δα i The PLC main controller distributes the commands to each drive unit for execution. When any guide vane is detected to have an encoder malfunction, execution error exceeding the threshold, abnormal current / torque of the drive unit, or excessive vibration / pressure pulsation, the differential compensation amount Δα is stopped from being sent to the abnormal guide vane. i It switches to characteristic curve-based opening setting and PID closed-loop control, and maintains synchronous control for the remaining guide vanes or redistributes the difference compensation amount. The adjustment workstation updates the model or its parameters based on the feedback data after execution, writes the feature vector, target command, execution error, efficiency estimate and vibration / pressure pulsation feature value of each control cycle into the historical database, and incrementally updates the machine learning model or its parameters according to a preset cycle or triggering condition to form an adaptive closed-loop optimization.

6. The method for adjusting the guide vane opening of a distributed magnetically coupled turbine as described in claim 5, characterized in that: The feature vector X(t) includes, but is not limited to, the concatenation of the following three sub-vectors: The basic operating state sub-vector contains the unit's macroscopic operating parameters and is used to determine the unit's approximate position on the characteristic curve; Guide vane distribution and feedback sub-vectors, which include the microstate of the guide vane system and the system's response feedback, are used to capture the effects of asynchronous regulation; Dynamic trend subvectors, containing the first derivative of key parameters with respect to time, are used to identify transient processes in operating conditions.

7. The method for adjusting the guide vane opening of a distributed magnetically coupled turbine as described in claim 5, characterized in that: Anomaly detection includes, but is not limited to, the following logical rules: Encoder anomaly detection: The data packets fed back by the encoder fail the CRC check N times consecutively; Execution error judgment: If the guide vane angle execution error is greater than the dynamic threshold and the duration is greater than the tolerance time, the actuator is judged to be stuck or the precision is failed. Drive unit abnormality judgment rules: Motor current > 1.2 times rated current for 5 seconds, motor current suddenly drops to no-load current level; Vibration / pressure pulsation exceeding limit judgment rule: V > V warn This triggers a switch in the control strategy.

8. The method for adjusting the guide vane opening of a distributed magnetically coupled turbine as described in claim 5, characterized in that: The characteristic curve is a preset safe operating envelope. The characteristic curve defines the recommended guide vane opening under different heads to ensure that the unit does not experience cavitation and severe vibration.

9. The method for adjusting the guide vane opening of a distributed magnetically coupled turbine as described in claim 5, characterized in that: The redistribution specifically involves the following: when a guide vane fails, redistribution is performed using an adjacent compensation method. The flow field adjustment tasks originally assigned to the failed guide vane R are redistributed to its adjacent guide vanes R according to an inverse weight. +1 and R -1 .

10. The method for adjusting the guide vane opening of a distributed magnetically coupled turbine as described in claim 5, characterized in that: The incremental update specifically refers to: The system maintains an experience playback buffer, which combines the input vector X(t) and control command u generated in each control cycle with the above feedback data into a new training sample and stores it in the buffer. Set a fixed period to update the model using the new sample set generated that day, or update it immediately when an abnormal event is triggered; The gradient descent method is used to calculate the gradient of the loss function with new samples in the buffer. Only the weights and biases of the last few layers of the network are updated by backpropagation with small steps, so as to keep the feature extraction capability of the bottom layer unchanged. The updated model is validated on the validation set.

Images