Intelligent control system and method for deepwater hydraulic automatic hook beam

Abstract

The invention discloses an intelligent control system and method for a deepwater hydraulic automatic hook beam, and belongs to the technical field of deepwater hydraulic control. The system comprises a multi-sensor fusion module, an AI control unit, a redundant hydraulic loop and a remote monitoring module. The multi-sensor fusion module collects bolt displacement, hydraulic system state and watertight cable health data in real time; the AI control unit comprises a fuzzy PID (Proportion Integration Differentiation) controller and a reinforcement learning optimization module, and dynamically adjusts control parameters according to environmental disturbance; the redundant hydraulic loop is automatically switched when the main loop breaks down; the watertight cable health detection adopts an impedance-high frequency signal dual-mode detection method. The control method comprises the steps of system initialization, dynamic environment modeling, fuzzy PID and reinforcement learning cooperative control, fault intelligent diagnosis, emergency response and the like. The problems that the hydraulic bolt is low in positioning precision and poor in reliability in the deepwater environment are solved, and the operation reliability and the intelligent level of a hydropower station gate system are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of deep-water hydraulic automatic hook beam control technology, and specifically relates to an intelligent control system and method for deep-water hydraulic automatic hook beams. Background Technology

[0002] In hydropower station gate control systems, the hydraulic automatic hook beam, as a key connecting device, undertakes the core functions of gate opening, closing, and locking. Under deep-water operating conditions (50-300 meters), traditional hydraulic automatic hook beams face severe challenges in control accuracy: due to the combined effects of dynamic environmental factors such as water flow disturbance, water pressure fluctuations, and mechanical vibration, the positioning accuracy of the hydraulic pins significantly decreases, leading to misalignment between the pins and the gate slot, which in turn causes equipment jamming, sealing failure, and even structural damage. Existing solutions mostly employ static control strategies and single sensor feedback, which cannot adapt to the dynamic changes in deep-water environments. The pin positioning error typically exceeds 2mm, far exceeding the ±0.5mm accuracy requirement allowed for deep-water operations. This technical bottleneck severely restricts the reliable operation and intelligent upgrading of deep-water gate systems in hydropower stations. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide an intelligent control system and method for a deep-water hydraulic automatic hook beam. The present invention solves the problems of low positioning accuracy and poor reliability of hydraulic pins in deep-water environments, and significantly improves the operational reliability and intelligence level of the gate system of hydropower stations.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: An intelligent control system for a deep-water hydraulic automatic hook beam includes: A multi-sensor fusion module is used to collect real-time data on pin displacement, hydraulic system status, and watertight cable health. The AI ​​control unit, connected to the multi-sensor fusion module, includes a fuzzy PID controller and a reinforcement learning optimization module, used to dynamically adjust control parameters according to environmental disturbances. A redundant hydraulic circuit, including a main hydraulic circuit and a backup hydraulic circuit, is connected to the AI ​​control unit and is used to automatically switch when the main hydraulic circuit fails. The remote monitoring module is communicatively connected to the AI ​​control unit and is used for data transmission and remote command issuance. The multi-sensor fusion module includes a high-precision displacement sensor, a watertight cable health detection unit, and a hydraulic system monitoring sensor.

[0005] Preferably, the watertight cable health detection unit is used to perform an impedance-high frequency signal dual-mode detection method, including: Impedance measurement module, used to monitor cable impedance values ​​in real time and compare them with reference values; A high-frequency signal injection module is used to inject a 1MHz high-frequency signal into the cable and detect the reflection coefficient; The cable condition assessment module is used to comprehensively judge the insulation condition of cables based on impedance change rate and reflection coefficient.

[0006] Preferably, the AI ​​control unit includes: The dynamic environment modeling module constructs a water flow disturbance prediction model based on an LSTM network. Fuzzy PID controller dynamically adjusts PID parameters based on displacement error and the rate of change of error; The reinforcement learning optimization module constructs a reward function based on pin positioning accuracy, action time, and hydraulic energy consumption, and iteratively optimizes the fuzzy rule base.

[0007] Preferably, it further includes a fault diagnosis and emergency response module, the fault diagnosis and emergency response module comprising: LSTM fault diagnosis model is used to analyze multi-source time-series sensor data to identify the types and probabilities of hydraulic system faults. The fault level classification unit divides faults into Level I, Level II, and Level III. The emergency strategy execution unit automatically executes the corresponding emergency strategy according to the fault level, including pressure compensation, redundant loop switching or emergency system depressurization.

[0008] The intelligent control method of the intelligent control system for the deep-water hydraulic automatic hook beam includes the following steps: System initialization and parameter calibration, establishing sensor reference data and control parameters; Dynamic environment modeling and real-time data acquisition are used to obtain data on water flow velocity, water pressure, and equipment status. A fuzzy PID and reinforcement learning collaborative control algorithm is adopted to dynamically compensate for the impact of environmental disturbances on the pin positioning. Intelligent fault diagnosis is performed through LSTM networks, and automated emergency response is executed according to the fault level. Data storage and remote collaborative optimization provide a basis for performance optimization for long-term system operation.

[0009] Preferably, the fuzzy PID and reinforcement learning collaborative control algorithm specifically includes: The pin displacement error and the rate of change of error are fuzzified into multiple fuzzy subsets; Calculate the PID parameter correction value based on the preset fuzzy rule base; Define the state space, action space, and reward function for reinforcement learning; The long-term control performance is optimized by iteratively updating the weights of fuzzy rules using the Q-learning algorithm.

[0010] Preferably, the watertight cable condition monitoring employs an impedance-high-frequency signal dual-mode detection method, including: Calculate the cable insulation attenuation rate: ΔK_ins=(Z_reference-Z(t)) / Z_reference; A 1MHz high-frequency signal is injected, and the signal reflection coefficient γ is detected. When ΔK_ins>0.2 or γ>0.1, the cable is determined to have a fault risk and an early warning is triggered.

[0011] Preferably, the intelligent fault diagnosis and emergency response includes: Normalize and apply sliding window processing to the time series data of pressure, flow, and impedance; Time series features are extracted using a 3-layer LSTM network, and the fault category and probability are output. When a Level I fault is identified, pressure compensation and enhanced monitoring are implemented. When a Level II fault is identified, switch to the backup hydraulic circuit and reduce the operating speed. When a Level III fault is detected, the system will perform forced pin retraction, system depressurization, and emergency alarm.

[0012] Preferably, the LSTM fault diagnosis model uses a sliding time window of 30 sampling points. The input includes hydraulic pressure, flow rate, and cable impedance time series data of the past 30 seconds. When the output fault category confidence is ≥85%, it is determined to be a valid fault.

[0013] Preferably, the dynamic environment modeling includes: Establish an autoregressive moving average model (ARMA(p,q)) for water flow disturbance, where p=3 and q=2; The total water pressure P_total is calculated based on the water depth h and the real-time velocity v: P_total = ρgh + 0.5·ρ·v². The actual position deviation of the door slot is obtained through the visual recognition module, and a spatial compensation matrix T is constructed. The water flow disturbance value, water pressure change and gate slot position deviation are integrated into a comprehensive environmental disturbance vector, which is used to compensate control parameters in real time.

[0014] An intelligent control device employs the intelligent control method of the aforementioned intelligent control system for deep-water hydraulic automatic hook beams, comprising: Embedded processing unit, configured with a quad-core ARM processor with a main frequency of ≥1.2GHz, is used to run AI control algorithms; The watertight junction box features a double-sealed structure and contains an internal humidity sensor. The emergency power module has a built-in supercapacitor bank and can provide backup power for ≥30 minutes when the main power is interrupted; The underwater communication unit supports both underwater acoustic and fiber optic communication modes. The anti-interference shielding structure adopts an electromagnetic shielding cover and shock-absorbing bracket to meet the requirements of working environment at a water depth of 300 meters.

[0015] Preferably, the software architecture of the embedded processing unit includes: Real-time data acquisition layer, sampling frequency ≥200Hz, data buffer depth ≥10 seconds; The intermediate layer of environmental modeling enables real-time calculation of water flow disturbance model, water pressure change model and gate slot offset compensation model; The core layer of intelligent control includes a fuzzy PID controller, a reinforcement learning optimizer, and an LSTM fault diagnosis engine. The decision-making and execution layer generates hydraulic valve control signals and alarm information based on control commands and fault diagnosis results; The layers communicate using a publish-subscribe messaging mechanism, with a critical control command processing latency of <10ms.

[0016] Preferably, the redundant hydraulic circuit adopts a dual-pump dual-valve architecture, with the main circuit and the backup circuit sharing a hydraulic oil tank but having an independent filtration and cooling system, and the pin displacement fluctuation during the switching process is <0.5mm.

[0017] The present invention can achieve the following beneficial effects: 1. This invention integrates fuzzy PID control and reinforcement learning algorithms, combined with dynamic environment modeling based on LSTM networks. The system can perceive and compensate for the impact of deep-water environmental factors such as water flow disturbance and water pressure fluctuation on the positioning of the pin in real time, significantly improving the docking accuracy and reliability of the pin and the slot, and effectively solving the technical problem of inaccurate positioning of traditional systems in complex underwater environments.

[0018] 2. This invention employs a dual-circuit redundant hydraulic design with primary and backup circuits and a multi-level fault diagnosis and response strategy. The system possesses the capabilities for autonomous fault identification, graded assessment, and emergency response. When an abnormal state is detected, it can automatically execute measures such as pressure compensation, circuit switching, or emergency protection based on the severity of the fault, significantly improving the safety and continuity of deep-water operations and preventing equipment damage and operational interruptions. Attached Figure Description

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a system architecture diagram of the present invention; Figure 2 This is a flowchart of the control algorithm of the present invention (fuzzy PID parameter adjustment and reinforcement learning optimization process). Figure 3 This is a structural diagram of the fault diagnosis model of the present invention. Detailed Implementation

[0020] like Figure 1-3 As shown, an intelligent control system and method for a deep-water hydraulic automatic hook beam belongs to the field of deep-water hydraulic automatic hook beam control technology. It addresses the technical challenges faced by hydropower station gate grab beams in deep-water environments (50-300 meters), including insufficient control accuracy, low reliability, lack of intelligent monitoring, and weak emergency response capabilities. This invention achieves high-precision positioning, real-time status monitoring, intelligent fault diagnosis, and automated emergency response of the hydraulic pin in deep-water environments through multi-sensor fusion, artificial intelligence control algorithms, redundant design, and intelligent fault diagnosis technology. The invention will be further described in detail below with reference to the accompanying drawings.

[0021] I. System Composition This intelligent control system comprises four core modules: 1. Multi-sensor fusion module: Used for real-time acquisition of pin displacement, hydraulic system status, and watertight cable health data. This module includes: High-precision displacement sensor: Real-time monitoring of pin extension / retraction displacement, with an accuracy of ±0.1mm; Watertight Cable Health Detection Unit: Performs impedance-high frequency signal dual-mode detection to assess cable insulation status in real time; Hydraulic system monitoring sensors: including pressure sensors, flow meters, and temperature sensors, covering all parameters of the oil circuit status.

[0022] 2. AI Control Unit: Connected to the multi-sensor fusion module, including: Fuzzy PID controller: dynamically adjusts PID parameters based on displacement error and the rate of change of error; Reinforcement learning optimization module: Construct a reward function based on pin positioning accuracy, action time, and hydraulic energy consumption, and iteratively optimize the fuzzy rule base; Dynamic environment modeling module: Constructs a water flow disturbance prediction model based on LSTM network.

[0023] 3. Redundant hydraulic circuits: These include a main hydraulic circuit and a backup hydraulic circuit, connected to the AI ​​control unit for automatic switching in case of a main hydraulic circuit failure. A dual-pump, dual-valve architecture is adopted. The main and backup circuits share a hydraulic oil tank but have independent filtration and cooling systems. During switching, the pin displacement fluctuation is <0.5mm.

[0024] 4. Remote monitoring module: Communicates with the AI ​​control unit for data transmission and remote command issuance, supporting both fiber optic and underwater acoustic communication modes.

[0025] II. Watertight Cable Health Detection Unit: The watertight cable health detection unit implements an impedance-high frequency signal dual-mode detection method, including: 1. Impedance Measurement Module: Real-time monitoring of cable impedance value Z(t) and comparison with reference value Z_reference to calculate cable insulation attenuation rate. ΔK_ins=(Z_reference-Z(t)) / Z_reference 2. High-frequency signal injection module: Injects a 1MHz high-frequency signal into the cable and detects the signal reflection coefficient γ.

[0026] 3. Cable Condition Assessment Module: This module comprehensively assesses the cable insulation condition based on the impedance change rate and reflection coefficient. When ΔK_ins > 0.2 or γ > 0.1, it determines that the cable has a fault risk and triggers an early warning.

[0027] This dual-mode detection method effectively solves the problem that traditional single detection mode cannot fully assess the health status of cables, and is especially suitable for early warning of watertight cable faults caused by long-term immersion or mechanical damage.

[0028] III. AI Control Unit The AI ​​control unit consists of three core components: 1. Dynamic Environment Modeling Module: Based on LSTM network, a water flow disturbance prediction model is constructed. Input the water flow velocity and direction time series data of the past 10 seconds, and output the water flow disturbance prediction value of the next 2 seconds. The prediction error is ≤ ±0.1m / s.

[0029] 2. Fuzzy PID controller: The pin displacement error e(t) is divided into 7 fuzzy subsets: NB, NM, NS, ZO, PS, PM, PB; The error rate of change ec(t) is divided into 7 fuzzy subsets, and the membership function adopts a triangular function; Forty-nine fuzzy rules were established, and the centroid method was used to calculate the PID parameter correction values ​​for defuzzification. The final control parameters are Kp=Kp0+ΔKp, Ki=Ki0+ΔKi, Kd=Kd0+ΔKd.

[0030] 3. Reinforcement Learning Optimization Module: The state space is S=[|e(t)|,t_action,E], where |e(t)| is the positioning accuracy, t_action is the action time, and E is the hydraulic energy consumption; Action space A = [Adjust fuzzy rule weights, correct PID parameter range, optimize pin action trajectory]; Reward function R=α(1 / |e(t)|)+β(1 / t_action)+γ(1 / E), where α=0.5, β=0.3, γ=0.2; The Q-learning algorithm is used to update the action value function, and the fuzzy rule base is iterated once for each completed pin-locking action.

[0031] IV. Fault Diagnosis and Emergency Response Module The fault diagnosis and emergency response module includes: 1. LSTM fault diagnosis model: The input includes hydraulic pressure, flow rate, and cable impedance timing data for the past 30 seconds (a sliding time window of 30 sampling points). Time series features were extracted using a 3-layer LSTM network (with 64, 32, and 16 neurons in the hidden layers, respectively). The dropout layer (dropout rate = 0.2) suppresses overfitting; Output the fault category (oil leakage, cable damage, plug jamming, normal) and the fault probability. The fault category is considered valid when the confidence level is ≥85%.

[0032] 2. Fault Level Classification Unit: Faults are classified into three levels: Level I: Minor faults that do not affect core functions (e.g., minor oil leakage, dP / dt≤0.1MPa / s). Level II: Moderate fault, requiring reduced load operation (e.g., cable insulation attenuation, K_ins(t)≤0.8); Level III: Serious malfunction requiring emergency shutdown (e.g., pin jamming t_block≥3s, oil circuit rupture dP / dt≥1MPa / s).

[0033] 3. Emergency Response Strategy Execution Unit: Level I fault: Perform pressure compensation (increase P_set=0.5MPa) and enhance monitoring; Level II fault: Switch to backup hydraulic circuit and reduce speed (v_set drops to 5mm / s); Level III fault: Forced retraction of the latch, system depressurization (P_relief=0MPa), and emergency alarm.

[0034] V. Intelligent Control Methods The intelligent control method of the present invention includes the following steps: 1. System initialization and parameter calibration: Input hardware parameters: rated pressure of hydraulic pump (25MPa), maximum stroke of pin (150mm), displacement sensor range (0-200mm), and reference value of watertight cable impedance (50Ω±5Ω). Input environmental baseline data: operating water depth H, initial water temperature T0, hydrostatic pressure baseline value P0=ρgH; The initial sensor data is processed using a mean filtering algorithm to remove random noise (the filter window size is set to 5 sampling points). Calibrate the displacement sensor: using the fully retracted position of the pin as the reference zero point and the fully extended position as the calibration endpoint, generate a displacement-voltage curve (linearity error ≤ ±0.05%). Establish a cable health benchmark model: Calculate the initial insulation state coefficient K_ins = Z_measured / Z_reference (normal range 0.9-1.1); Verify the hydraulic circuit sealing performance: continuously monitor the initial pressure for 30 seconds. If the pressure drop is ≤0.5MPa, the seal is considered normal.

[0035] 2. Dynamic environment modeling and real-time data acquisition: Real-time data acquisition: water flow velocity V, water flow direction θ, real-time water pressure P_real, water temperature change ΔT, real-time pin displacement D(t), real-time hydraulic circuit pressure P(t), hydraulic oil flow rate Q(t), and real-time cable impedance Z(t). A timestamp alignment algorithm (synchronization error ≤ 1ms) is used to integrate multi-source data; A water flow disturbance model is constructed based on an LSTM network, and the predicted value of water flow disturbance for the next 2 seconds is output. Calculate the water pressure compensation coefficient: K_p=1+ΔP / 25, where ΔP=P_real-P0; The cable insulation attenuation rate ΔK_ins is calculated by impedance-high frequency signal dual-mode detection.

[0036] 3. Fuzzy PID and Reinforcement Learning Cooperative Control Algorithm: The fuzzy subsets of e(t) and ec(t) are fuzzified. Calculate the PID parameter correction value based on the preset fuzzy rule base; The long-term control performance is optimized by iteratively updating the weights of fuzzy rules using the Q-learning algorithm. Dynamically compensate for the impact of environmental disturbances such as water flow and water pressure on the action of the pin, and achieve high-precision positioning of the pin (positioning error ≤ ±0.5mm).

[0037] 4. Dynamic Environment Modeling: Establish an autoregressive moving average model (ARMA(p,q)) for water flow disturbance, where p=3 and q=2; The total water pressure P_total is calculated based on the water depth h and the real-time velocity v: P_total = ρgh + 0.5·ρ·v². The actual position deviation of the door slot is obtained through the visual recognition module, and a spatial compensation matrix T is constructed. The water flow disturbance value, water pressure change and gate slot position deviation are integrated into a comprehensive environmental disturbance vector, which is used to compensate control parameters in real time.

[0038] 5. Intelligent fault diagnosis and emergency response: Normalize and apply sliding window processing to the time series data of pressure, flow, and impedance; Time series features are extracted using a 3-layer LSTM network, and the fault category and probability are output. The corresponding emergency strategy will be automatically executed based on the fault level.

[0039] 6. Data storage and remote collaboration optimization: Operational data is stored using the AES-256 encryption algorithm, and archived in categories of "date-operating condition-equipment number" with a retention period of ≥1 year; Real-time data is synchronized to the surface control console via fiber optic communication (transmission rate ≥100Mbps), with a latency ≤50ms; Based on historical operational data, a genetic algorithm is used to optimize the weights of the reinforcement learning reward function and the fuzzy rule base to generate the optimal control parameter template.

[0040] VI. Intelligent Control Equipment The intelligent control device of the present invention includes: 1. Embedded processing unit: Configured with a quad-core ARM processor with a clock speed of ≥1.2GHz; The software architecture is divided into four layers: Real-time data acquisition layer: sampling frequency ≥200Hz, data buffer depth ≥10 seconds; Intermediate layer for environmental modeling: Enables real-time calculation of water flow disturbance model, water pressure change model and gate slot offset compensation model; The core layer of intelligent control includes a fuzzy PID controller, a reinforcement learning optimizer, and an LSTM fault diagnosis engine. Decision execution layer: Generates hydraulic valve control signals and alarm information; The layers communicate using a publish-subscribe messaging mechanism, with a critical control command processing latency of <10ms.

[0041] 2. Watertight junction box: It adopts a double-layer sealing structure and is equipped with a humidity sensor to ensure sealing performance in a water depth of 300 meters.

[0042] 3. Emergency power module: Built-in supercapacitor bank, which can provide ≥30 minutes of backup power when the main power is interrupted.

[0043] 4. Underwater communication unit: Supports dual modes of underwater acoustic communication and fiber optic communication to ensure communication reliability.

[0044] 5. Anti-interference shielding structure: It adopts an electromagnetic shielding cover and shock-absorbing bracket to meet the requirements of working environment at a water depth of 300 meters.

[0045] Example 1: Taking the operation of a deep-water gate at the Three Gorges Hydropower Station as an example: 1. Scenario parameters: operating water depth 120 meters, water flow velocity 1.8 m / s, gate weight 500 tons, hydraulic automatic hook beam model ZHL-120, maximum pin stroke 150 mm, target pin positioning error ≤ ±0.5 mm.

[0046] 2. Implementation process: Initialization and calibration: Input hydraulic pump rated pressure 25MPa, pin initial displacement 0mm, cable impedance reference value 50Ω, initial water pressure P0=1.176MPa; after calibration, the displacement sensor sensitivity is 0.01mm / mV, the initial cable insulation coefficient K_ins=1.02, and the initial PID parameters Kp=5.0, Ki=0.1, Kd=0.5.

[0047] Dynamic environment modeling: real-time water flow velocity is 1.8 m / s, LSTM predicts water flow velocity of 1.9 m / s in the next 2 seconds, water pressure compensation coefficient K_p=1.00096; output displacement error e(t) initial value is 120 mm, error change rate ec(t)=10 mm / s.

[0048] Fuzzy PID and reinforcement learning control: Fuzzy rule matching outputs ΔKp=2.0, ΔKi=-0.02, ΔKd=0.3, and the final PID parameters are Kp=7.0, Ki=0.08, Kd=0.8; the final displacement of the pin is 120.3mm, and the displacement error is 0.3mm.

[0049] Fault diagnosis and emergency response: Simulates a sudden pressure change in the hydraulic circuit (dP / dt=0.8MPa / s, Level II fault). The LSTM model outputs the fault category "oil leakage" with a fault probability of 92%. Switching to the backup hydraulic circuit within 0.5 seconds, the pressure is restored to stability.

[0050] Data storage and optimization: Based on the updated fuzzy rule base, the optimal PID parameter template (Kp=7.2, Ki=0.07, Kd=0.9) for a water depth of 120 meters and a flow rate of 1.8-2.0 m / s is generated.

[0051] This embodiment verifies the high-precision control capability, rapid fault response capability, and autonomous parameter optimization capability of the present invention in actual deep water environment, which fully meets the technical requirements of hydropower station gate beam grabbing operation.

[0052] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

[0053] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. An intelligent control system for a deep-water hydraulic automatic hook beam, characterized in that, include: A multi-sensor fusion module is used to collect real-time data on pin displacement, hydraulic system status, and watertight cable health. The AI ​​control unit, connected to the multi-sensor fusion module, includes a fuzzy PID controller and a reinforcement learning optimization module, used to dynamically adjust control parameters according to environmental disturbances. A redundant hydraulic circuit, including a main hydraulic circuit and a backup hydraulic circuit, is connected to the AI ​​control unit and is used to automatically switch when the main hydraulic circuit fails. The remote monitoring module is communicatively connected to the AI ​​control unit and is used for data transmission and remote command issuance. The multi-sensor fusion module includes a high-precision displacement sensor, a watertight cable health detection unit, and a hydraulic system monitoring sensor.

2. The intelligent control system for the deep-water hydraulic automatic hook beam according to claim 1, characterized in that, The watertight cable health detection unit is used to perform an impedance-high frequency signal dual-mode detection method, including: Impedance measurement module, used to monitor cable impedance values ​​in real time and compare them with reference values; A high-frequency signal injection module is used to inject a 1MHz high-frequency signal into the cable and detect the reflection coefficient. The cable condition assessment module is used to comprehensively judge the insulation condition of cables based on impedance change rate and reflection coefficient.

3. The intelligent control system for the deep-water hydraulic automatic hook beam according to claim 1 or 2, characterized in that, The AI ​​control unit includes: The dynamic environment modeling module constructs a water flow disturbance prediction model based on an LSTM network. Fuzzy PID controller dynamically adjusts PID parameters based on displacement error and the rate of change of error; The reinforcement learning optimization module constructs a reward function based on pin positioning accuracy, action time, and hydraulic energy consumption, and iteratively optimizes the fuzzy rule base.

4. The intelligent control system for the deep-water hydraulic automatic hook beam according to claim 3, characterized in that, It also includes a fault diagnosis and emergency response module, which includes: LSTM fault diagnosis model is used to analyze multi-source time-series sensor data to identify the types and probabilities of hydraulic system faults. The fault level classification unit divides faults into Level I, Level II, and Level III. The emergency strategy execution unit automatically executes the corresponding emergency strategy according to the fault level, including pressure compensation, redundant loop switching, or emergency system depressurization.

5. The intelligent control method of the intelligent control system for the deep-water hydraulic automatic hook beam according to any one of claims 1-4, characterized in that, The method includes the following steps: System initialization and parameter calibration, establishing sensor reference data and control parameters; Dynamic environment modeling and real-time data acquisition are used to obtain data on water flow velocity, water pressure, and equipment status. A fuzzy PID and reinforcement learning collaborative control algorithm is adopted to dynamically compensate for the impact of environmental disturbances on the pin positioning. Intelligent fault diagnosis is performed through LSTM networks, and automated emergency response is executed according to the fault level. Data storage and remote collaborative optimization provide a basis for performance optimization for long-term system operation.

6. The intelligent control method of the intelligent control system for the deep-water hydraulic automatic hook beam according to claim 5, characterized in that, The fuzzy PID and reinforcement learning collaborative control algorithm specifically includes: The pin displacement error and the rate of change of error are fuzzified into multiple fuzzy subsets; Calculate the PID parameter correction value based on the preset fuzzy rule base; Define the state space, action space, and reward function for reinforcement learning; The long-term control performance is optimized by iteratively updating the weights of fuzzy rules using the Q-learning algorithm.

7. The intelligent control method for the intelligent control system of the deep-water hydraulic automatic hook beam according to claim 5 or 6, characterized in that, The watertight cable condition monitoring employs an impedance-high-frequency signal dual-mode detection method, including: Calculate the cable insulation attenuation rate: ΔK_ins=(Z_reference-Z(t)) / Z_reference; A 1MHz high-frequency signal is injected, and the signal reflection coefficient γ is detected. When ΔK_ins>0.2 or γ>0.1, the cable is determined to have a fault risk and an early warning is triggered.

8. The intelligent control method of the intelligent control system for the deep-water hydraulic automatic hook beam according to claim 7, characterized in that, The intelligent fault diagnosis and emergency response include: Normalize and apply sliding window processing to the time series data of pressure, flow, and impedance; Time series features are extracted using a 3-layer LSTM network, and the fault category and probability are output. When a Level I fault is identified, pressure compensation and enhanced monitoring are implemented. When a Level II fault is identified, switch to the backup hydraulic circuit and reduce the operating speed. When a Level III fault is detected, the system will perform forced pin retraction, system depressurization, and emergency alarm.

9. The intelligent control method for the intelligent control system of the deep-water hydraulic automatic hook beam according to claim 8, characterized in that, The LSTM fault diagnosis model uses a sliding time window with 30 sampling points. The input includes hydraulic pressure, flow rate, and cable impedance time series data from the past 30 seconds. When the output fault category confidence level is ≥85%, it is determined to be a valid fault.

10. The intelligent control method of the intelligent control system for the deep-water hydraulic automatic hook beam according to claim 9, characterized in that, The dynamic environment modeling includes: Establish an autoregressive moving average model (ARMA(p,q)) for water flow disturbance, where p=3 and q=2; The total water pressure P_total is calculated based on the water depth h and the real-time velocity v: P_total = ρgh + 0.5·ρ·v². The actual position deviation of the door slot is obtained through the visual recognition module, and a spatial compensation matrix T is constructed. The water flow disturbance value, water pressure change and gate slot position deviation are integrated into a comprehensive environmental disturbance vector, which is used to compensate control parameters in real time.

11. An intelligent control device, characterized in that, The intelligent control method of the intelligent control system for the deep-water hydraulic automatic hook beam according to any one of claims 5-9 is characterized by comprising: Embedded processing unit, configured with a quad-core ARM processor with a main frequency of ≥1.2GHz, is used to run AI control algorithms; The watertight junction box features a double-sealed structure and contains an internal humidity sensor. The emergency power module has a built-in supercapacitor bank and can provide backup power for ≥30 minutes when the main power is interrupted; The underwater communication unit supports both underwater acoustic and fiber optic communication modes. The anti-interference shielding structure adopts an electromagnetic shielding cover and shock-absorbing bracket to meet the requirements of working environment at a water depth of 300 meters.

12. The intelligent control device according to claim 11, characterized in that, The software architecture of the embedded processing unit includes: Real-time data acquisition layer, sampling frequency ≥200Hz, data buffer depth ≥10 seconds; The intermediate layer of environmental modeling enables real-time calculation of water flow disturbance model, water pressure change model and gate slot offset compensation model; The core layer of intelligent control includes a fuzzy PID controller, a reinforcement learning optimizer, and an LSTM fault diagnosis engine. The decision-making and execution layer generates hydraulic valve control signals and alarm information based on control commands and fault diagnosis results; The layers communicate using a publish-subscribe messaging mechanism, with a critical control command processing latency of <10ms.

13. The intelligent control device according to claim 11, characterized in that, The redundant hydraulic circuit adopts a dual-pump dual-valve architecture. The main circuit and the backup circuit share the hydraulic oil tank but have independent filtration and cooling systems. During the switching process, the pin displacement fluctuation is <0.5mm.

Images