A Method and System for Ensuring Navigation Safety of Multi-Level Water-Saving Ship Locks Based on Digital Twins

Abstract

This invention discloses a method and system for ensuring navigation safety in a multi-level water-saving ship lock based on digital twins. The method includes: collecting multi-source sensor data, generating an aligned data stream and quality assessment report, and updating the digital twin state; using the aligned data stream and quality assessment report, obtaining the mooring force probability distribution through non-contact observation and probabilistic fusion; identifying the hydraulic shock response based on the digital twin state and the aligned data stream, and generating a vibration-suppressing opening template using hydraulic phase-locked tracking; within the digital twin system, integrating the mooring force probability distribution and the vibration-suppressing opening template, performing sub-Bruker stochastic control barrier optimization, and outputting an executable valve trajectory. This invention can actively suppress water hammer oscillations and provide robust safety guarantees for uncertain mooring forces, effectively improving the navigation safety of large water-saving ship locks.

Description

Technical Field

[0001] This invention belongs to the field of intelligent shipping and hydraulic control technology, and in particular, it is a method and system for ensuring navigation safety of multi-level water-saving ship locks based on digital twins. Background Technology

[0002] Multi-stage water-saving locks, as key nodes in modern large-scale shipping channels, directly impact the efficiency of the entire water transport system through their operational safety and throughput efficiency. These locks utilize a complex multi-stage water-saving pool system for water recycling; however, valve switching operations during filling and emptying can induce drastic hydraulic transients in the pipeline system. Digital twin technology, a cutting-edge technology deeply integrating the physical world and cyberspace, provides a new technical approach for achieving precise state mapping, dynamic risk simulation, and intelligent optimization control of such complex hydraulic systems, holding significant research importance for ensuring the safe passage of large locks.

[0003] Currently, research on the hydraulics and safety control of ship locks has made some progress. In hydraulic modeling, numerical simulation methods based on one-dimensional or three-dimensional computational fluid dynamics (CFD) are commonly used to simulate and analyze the flow characteristics under predetermined operating conditions. In lock control, programmed control based on programmable logic controllers (PLCs) is widely adopted, executing operations according to preset gate valve opening and closing curves and fixed switching logic. In safety monitoring, the main methods rely on video surveillance, Automatic Identification System (AIS), and water level or pressure sensors at key points to routinely monitor the ship's mooring status and hydraulic parameters, and to issue threshold alarms.

[0004] However, existing technologies still face several deeply interconnected technical bottlenecks when addressing the complex dynamic safety issues of large, multi-stage water-saving locks. Existing control strategies for suppressing water hammer largely rely on passive protection or empirical tuning, such as using S-shaped opening and closing curves to extend valve operation time or timing switching based on fixed phases. Passive strategies cannot track and adapt to the instantaneous phase and frequency drift of the dominant water hammer mode in real time, making it difficult to achieve destructive interference with hydraulic oscillations. This results in significant pressure pulsations and lock chamber water level fluctuations still occurring during valve switching. Unsuppressed hydraulic oscillations directly exacerbate the uncertainty of mooring forces on vessels underway. Existing safety assurance methods lack effective quantification tools and robust control strategies for such uncertainties, typically employing deterministic safety thresholds or fixed safety factors, which cannot address the worst-case probability distribution caused by sensor noise, data quality degradation, or even hydraulic model mismatch. When passive water hammer suppression strategies fail, deterministic safety control cannot provide provable safety guarantees for highly uncertain mooring forces, posing significant safety risks. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for ensuring navigation safety of multi-level water-saving ship locks based on digital twins, so as to solve the above-mentioned problems existing in the prior art.

[0006] According to one aspect of this application, a method for ensuring navigation safety in a multi-level water-saving ship lock based on digital twins includes:

[0007] Collect multi-source sensor data, generate aligned data streams and quality assessment reports, and update the digital twin status;

[0008] By utilizing aligned data streams and quality assessment reports, and through non-contact observation and probabilistic fusion, the probability distribution of mooring force was obtained;

[0009] Based on the digital twin state and aligned data stream, the hydraulic shock response is identified and hydraulic phase-locked tracking is used to generate a vibration suppression opening template;

[0010] In the digital twin system, the probability distribution of the integrated mooring force and the vibration damping opening template are combined to perform random control barrier optimization using a multi-bar bar, and output an executable valve trajectory.

[0011] According to another aspect of this application, a multi-level water-saving lock navigation safety assurance system based on digital twins is provided, comprising:

[0012] The data aggregation unit is used to aggregate multi-source sensor data, generate aligned data streams and quality assessment reports, and update the digital twin status.

[0013] The probabilistic observation unit is used to obtain the mooring force probability distribution through non-contact observation and probabilistic fusion using aligned data streams and quality assessment reports;

[0014] The vibration damping template generation unit is used to identify the hydraulic shock response based on the digital twin state and aligned data stream, and to generate a vibration damping opening template by using hydraulic phase-locked tracking.

[0015] The sub-brute rod control unit is used in a digital twin system to integrate the probability distribution of mooring force and the vibration damping opening template, perform sub-brute rod random control barrier optimization, and output an executable valve trajectory.

[0016] Beneficial effects: Through the above technical solutions, the present invention can actively suppress water hammer oscillations and provide robust safety assurance against uncertain mooring forces, effectively improving the navigation safety of large water-saving locks. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall process of a multi-level water-saving ship lock navigation safety assurance method based on digital twins.

[0018] Figure 2This is a schematic diagram of the process of using a hydraulic phase-locked loop to dynamically track the frequency and instantaneous phase of the dominant water hammer mode to form a phase-locked reference.

[0019] Figure 3 This is a schematic diagram of the process for generating the vibration suppression opening template.

[0020] Figure 4 It performs random control barrier optimization using a distributed blob and outputs a schematic diagram of the executable valve trajectory process. Detailed Implementation

[0021] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] The terms "first," "second," etc., used in the specification and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, or product comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or units.

[0023] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0024] Example 1: A method and system for ensuring navigation safety of multi-level water-saving ship locks based on digital twins.

[0025] In an exemplary system framework, the system may include several functional modules. For example:

[0026] The navigation water level spatiotemporal forecast module is used to obtain dynamic information on the navigation water level of the lock.

[0027] Dynamic positioning and early warning module for vessels passing through locks.

[0028] Used to collect ship safety situational awareness data;

[0029] The lock safety operation calculation module, as the intelligent hub of the system, is used to perform fusion calculation and modeling analysis based on the above multi-source information using digital twin technology to generate collaborative auxiliary decision-making.

[0030] It also includes a ship lock passage scheduling and water-saving pool control module, which, based on the obtained decisions, enables the precise delivery of ship scheduling instructions and the automatic control of water-saving pool operation strategies in a digital twin environment.

[0031] Through the coordinated operation of the aforementioned modules, the system achieves dynamic and coordinated operation of the three elements: the lock, the vessel, and the water-saving pool. This solves the problem in existing technologies of ensuring the normal operation of large water-saving locks and the safety of vessel navigation. Subsequent embodiments will elaborate on the main methods for implementing this system and method, particularly the specific processes for water hammer suppression and uncertainty safety control.

[0032] Example 2 describes the overall process of a multi-level water-saving ship lock navigation safety assurance method based on digital twins provided by the present invention, such as... Figure 1 As shown, specifically, it includes the following steps:

[0033] Step one involves acquiring multi-source sensor data, generating an aligned data stream and a quality assessment report, and updating the digital twin status. Multi-source sensor data may include, but is not limited to: gate chamber water level sequences, pressures before and after valves, valve positions, visual and inertial measurement unit (IMU) data on ship position and attitude, and historical operational data. Generating an aligned data stream ensures strict temporal synchronization of data from all different sources and at different sampling rates. The quality assessment report provides a basis for probabilistic fusion and robust control of parameters such as confidence levels, noise levels, and delays for each sensor channel.

[0034] Step 2: Using the aligned data stream and quality assessment report, the mooring force probability distribution is obtained through non-contact observation and probability fusion.

[0035] To address the challenges of directly measuring mooring force and its inherent high degree of uncertainty, non-contact methods such as vision and inertia are used to observe the cable's geometry. This is combined with hydrodynamic corrections and probability fusion (e.g., Kalman filtering) utilizing uncertainty information from quality assessment reports. The final output is not a single, definitive value, but a probability distribution containing the mean and standard deviation (or variance), which more accurately reflects the actual state of the mooring force and its range of uncertainty.

[0036] Step 3: Based on the digital twin state and aligned data stream, identify the hydraulic shock response and use hydraulic phase-locked tracking to generate a vibration suppression opening template.

[0037] To proactively suppress water hammer during the switching of the water-saving tank, a dynamic response model (i.e., impulse response) from valve actuation to critical section pressure is obtained through online identification (e.g., applying micro-excitation during a safety window). Simultaneously, hydraulic phase-locked loop (H-PLL) technology is employed to track the frequency and instantaneous phase of the dominant water hammer mode in real time. Based on the response model and the phase-locked reference, the system optimizes and solves for an interference micro-pulse sequence, which is then superimposed onto a conventional slow-opening template to form the final vibration-suppressing opening template. The active interference effect of the template generates a pressure wave in the critical section that is out of phase with the dominant water hammer wave, achieving destructive interference and suppressing the water hammer peak value.

[0038] Step four: In the digital twin system, the probability distribution of the mooring force and the vibration damping opening template are combined to perform random control barrier optimization of the sub-Bruker bar, and output an executable valve trajectory.

[0039] To ensure relative safety under uncertainties in mooring force, a Model Predictive Control (MPC) framework is adopted, using the probability distribution of mooring force and the vibration damping opening template as inputs. In this step, the sub-Bruker bar and control barrier construct an uncertainty distribution set centered on the empirical distribution, encompassing all possible worst-case probability distributions (its size is determined by the quality assessment report), and define a safety barrier function (e.g., the difference between the safety threshold and the mooring force). Through sub-Bruker bar optimization, it is ensured that the safety barrier still meets safety constraints under the worst-case probability distribution (e.g., the probability of mooring force exceeding the limit is below a certain minimum value δ). The optimization objective also considers tracking the vibration damping opening template, flow smoothness, and water conservation costs.

[0040] The final output executable valve trajectory is the optimal control sequence that satisfies water hammer suppression, multi-objective optimization, and robust safety assurance.

[0041] Example 3 details a preferred implementation method for collecting multi-source sensor data, generating aligned data streams and quality assessment reports, and updating the digital twin status. This can be broken down into the following steps:

[0042] The sensor access and time alignment are performed, specifically acquiring raw multi-source data channels such as gate chamber water level sequence, pressure before and after valves, valve position, video frame sequence, ship inertial measurement data (IMU), AIS trajectory, historical operation data, and camera intrinsic and extrinsic parameters.

[0043] To generate aligned data stream D _sync This requires performing timestamp unification, sampling rate renormalization, and missing measurement correction. Specifically, in time alignment, the system can estimate the clock skew and end-to-end delay distribution of each channel (e.g., using round-trip timing) and calculate a unified reference time axis t. _ref .

[0044] In resampling and interpolation strategies, a preferred approach for continuous quantities (such as gate water level, pressure, and gate position) is to use causal spline interpolation or piecewise cubic Hermite interpolation algorithms to uniformly interpolate them to the reference time axis t. _ref superior.

[0045] For discrete events (such as video frame timestamps), nearest neighbor alignment can be used. In missing and anomaly detection and repair, preferably, a two-sided Hampel filter can be used to remove spikes or outliers from the data.

[0046] The system performs coordinate and scale registration, completing the fusion of camera-lock chamber-hull coordinates and pixel-physical scale calibration. Specifically, the system acquires the camera's intrinsic and extrinsic parameters. _calib Three-dimensional geometric model of lock chamber and mooring point _geom .

[0047] In a preferred implementation, when estimating the external parameter matrix of the camera-lock chamber-ship body, the EPnP algorithm (efficient PnP algorithm) combined with the Random Sample Consensus (RANSAC) algorithm can be used to unify it to the lock chamber reference coordinate system. The final output is the registration matrix and parameter Reg. _param This is for use in subsequent non-contact observations.

[0048] Perform a data quality assessment, specifically, obtain aligned data stream D. _sync The system calculates health scores for each channel, including noise variance, delay, drift, availability, and jitter. Preferably, the power spectral density of signals such as water level and pressure can be estimated to generate noise parameters relevant to the operating conditions. Based on the above analysis, the system generates fusion weights w for each channel. _sens Combined with confidence labels, and compiled into a quality assessment report (QC). _report It is used for probabilistic fusion of mooring force observations (as input to the noise model) and for setting the radius of the uncertainty set (i.e., the Wasserstein radius) for the sub-Bruker optimization.

[0049] Perform digital twin state updates and agent model initialization; specifically, the system acquires the aligned data stream D. _sync With Quality Control (QC) report _report Updating the digital twin state Twin under existing geometric and hydraulic boundary conditions _state (t) (e.g., updating the water level field, valve status, ship relative attitude, etc.). Simultaneously, historical operational data (Hist) can be acquired to initialize a low-order hydraulic-inventory proxy model (Model) for model predictive control. _base .

[0050] Through the above steps, a high-quality, highly consistent aligned data stream and a computable digital twin state were constructed, providing a data foundation and model input for observation, vibration suppression, and optimized control.

[0051] Example 4: A detailed description of the specific implementation method for obtaining the mooring force probability distribution using aligned data streams and quality assessment reports through non-contact observation and probabilistic fusion, including the following steps:

[0052] Step 4.1: Extract visual and inertial data from the alignment data stream to obtain the cable geometry sequence. Specifically, the system acquires the alignment data stream D. _sync and registration parameters Reg _param By analyzing video frame sequences... _t Distortion correction and pose estimation were performed, and the results were compared with inertial measurement data (IMU). _t AIS trajectory AIS _t Perform spatiotemporal alignment and output visual-inertial alignment data (VI). _sync Based on this, the system utilizes VI _sync Three-dimensional geometric model of the gate chamber _geom The system uses algorithms such as semantic segmentation and line structure tracking to identify the anchor points and sag curves of each cable in the video image. It also utilizes the pixel-to-meter conversion relationship in the registration parameters to convert the geometric information in the image into a physical scale, outputting the cable geometric sequence Line. _geom (t), the sequence can contain information such as the real-time length, angle of incidence, angle of elevation, and sag depth of each cable.

[0053] Step 4.2: Based on the cable geometry sequence and preset cable material properties, calculate the prior tension. The system acquires the Line... _geom (t) and the preset cable material prior. _prior Material property priors can be stored in a database, containing stress-strain curves, equivalent cross-sectional areas, and nonlinear relaxation-tension thresholds for different types of cables (such as nylon and polypropylene). Based on the real-time geometric parameters of the cable (such as length and sag), the system uses, for example, the catenary equation or a simplified finite element model to back-calculate the equivalent elongation of the cable. By querying the stress-strain curves in the material property priors, the tension prior sequence F is calculated. _line_prior (t).

[0054] Step 4.3, based on the pressure, water level, and ship dynamics data in the aligned data stream, estimates the relative flow velocity and pulsating pressure near the guide hole to obtain the hydrodynamically corrected tension. Tension calculated solely based on prior material properties is insufficient in complex flow environments; this step further estimates the additional impact of hydrodynamic forces on cable tension.

[0055] Specifically, the system utilizes pressure in the aligned data stream (such as p) _up (t), p _down Using η(t) and water level data (such as η(t)), combined with a lock chamber hydraulic model (which can be a computational fluid dynamics (CFD) model or a reduced-order model), the pressure components of the water flow pulsations caused by lock chamber filling and emptying (especially valve operation) at the ship's guide hole location are estimated. Simultaneously, ship dynamic data (such as IMU data) from the aligned data stream are utilized. _t AIS _t This estimates the relative velocity component of the vessel within the lock chamber due to water flow propulsion or its own slight drift. Combining the pulsating pressure component (additional impact on the cable) and the relative velocity component (additional drag force on the cable), the hydrodynamic correction term F is calculated. _hydro_corr (t). The obtained hydrodynamic correction tension F _line_est (t) can be represented as F _line_est (t)=F _line_prior (t)+F _hydro_corr (t). up and down indicate upstream and downstream. For example, p _up This refers to the pressure upstream of the valve. _down This refers to the downstream pressure of the valve.

[0056] Step 4.4: Within the probabilistic fusion framework, the prior tension, hydrodynamically corrected tension, and sensor uncertainties from the quality assessment report are fused to output a mooring force probability distribution, which includes the mean and standard deviation of the mooring force. To obtain a quantitative description of the mooring force uncertainty, the system performs probabilistic fusion. The quality assessment report (QC) is then obtained. _report And based on the sensor weights w _sens The system sets corresponding noise models (e.g., expressed as the measurement noise covariance matrix) for visual measurements, IMU measurements, pressure measurements, and noise parameters. Preferably, this step can be performed within a nonlinear filtering framework (such as the Unscented Kalman Filter (UKF)) or a Bayesian framework. The system uses the prior tension (or corrected tension) as the predicted value of the process model and the observation data (or their derived tension estimates) from the visual, IMU, and pressure channels as the observation vector. Through the prediction-update iteration of the UKF, the system fuses these uncertainties in real time and outputs the mooring force probability distribution F. _line_dist (t). F _line_dist (t) is a probabilistic description of the cable tension for each cable, which can be represented by a Gaussian distribution including its posterior mean μ. _line (t) and posterior standard deviation σ _line (t) (or variance).

[0057] Furthermore, the probability distribution F of the mooring force is obtained. _line_dist(t) After that, the system can also perform dangerous cable identification. For example, the system can identify dangerous cables based on the allowable breaking force F of the cable. _break Given a preset safety factor β, calculate the probability of the mooring force exceeding the limit, i.e., P. _exceed (t)=Pr{F _line >β*F _break When probability P _exceed (t) When the risk exceeds a certain preset threshold, the system will add the cable to the list of dangerous cables. _risk_line (t), and can trigger an early warning.

[0058] Example 5: A detailed description of the specific implementation process for identifying the hydraulic shock response (HIS) and using hydraulic phase-locked loop (H-PLL) tracking to generate a vibration suppression opening template, including the following steps:

[0059] Step 5.1: Apply micro-excitation within the safe injection window and identify the discrete impulse response matrix from the valve to the critical section based on the pressure and gate position response in the aligned data stream. Before performing the identification, the system preferably selects the critical section and the injection window. The system can identify, for example, several locations with the highest historical water hammer pressure peaks or those most sensitive to gate position changes, based on the gate chamber and pipeline topology, through historical data analysis or simulation calculations, as the set K of critical sections. Simultaneously, the system determines the safe injection window (Window) based on the current operating conditions (e.g., whether the vessel is securely moored). _inj Applying minor disturbances inside the window will not affect navigation safety. The system can also be configured with injection interlock conditions, such as prohibiting injection stimuli during the ship's berthing or unberthing phases.

[0060] Within the safe injection window, pseudo-random sequences (PRBS) or micro-steps are superimposed onto the valves of each provincial water tank as micro-excitations. These micro-excitations are small-amplitude, individually designed signals whose amplitude (e.g., 1-5% of the total stroke) and frequency characteristics (such as the broadband characteristics of PRBS) are designed to fully excite the hydraulic dynamics of the system without interfering with normal operation. The aligned data stream D is extracted. _sync The pressure response induced by micro-excitation (i.e., the pressure sequence p at the critical section) _up (t), p _down (t) and the corresponding valve position sequence s _gate_i (t). Preferably, to improve the identification signal-to-noise ratio, the acquired pressure response signal can be first subjected to bandpass filtering or wavelet denoising. The recursive least squares (RLS) algorithm is used to identify the discrete impulse response matrix H[n] online based on the valve position sequence (as input) and the pressure response (as output). Specifically, H[n] is a set containing the discrete impulse response sequence h from the i-th valve (i is the valve index) to the k-th critical section (k is the section index). _k,i[n], where n is the discrete time step. The RLS algorithm continuously corrects h with newly acquired data by setting a forgetting factor λ (for example, λ can be between 0.95 and 0.995). _k , i The estimated value of [n]. In a preferred embodiment, to prevent overfitting or numerical instability in the identification process due to noise interference, the recursive least squares algorithm may introduce a Tikhonov regularization term.

[0061] Step 5.2: Extract the pressure sequence from the aligned data stream, and apply a hydraulic phase-locked loop to dynamically track the frequency and instantaneous phase of the dominant water hammer mode to form a phase-locked reference, such as... Figure 2 As shown. Simultaneously or subsequently, the system executes a hydraulic phase-locked loop (H-PLL) upon identifying H[n]. For the pressure sequence (which can be p... _up (t) or p _down A weighted combination of z(t) is subjected to a bandpass filter (the bandpass range covers the estimated main frequency band of the water hammer, e.g., 0.1Hz to 1.0Hz), and a Hilbert transform is applied to it to obtain the analytic signal z(t). The analytic signal z(t) is a complex signal, z(t) = p _bp (t)+j*H{p _bp (t)}, where p _bp z(t) is the bandpass filtered pressure signal, j is the imaginary unit, and H{} denotes the Hilbert transform. The system solves the analytic signal z(t) to obtain the continuous instantaneous phase trajectory φ. _inst (t). Specifically, the instantaneous phase φ _inst z(t) can be obtained by calculating the argument of z(t), i.e., φ. _inst (t)=atan2(Im(z(t)),Re(z(t))). A phase unwrapping operation needs to be performed on the phase trajectory to eliminate the 2π jump caused by the arctangent function and obtain a continuously accumulated phase. The system is configured with a second-order phase-locked loop filter containing proportional and integral gains; the instantaneous phase trajectory φ is calculated in real time. _inst (t) and the reference phase φ inside the filter _hat The phase error e of (t) _φ (t); The second-order phase-locked loop filter is driven by the phase error to dynamically update the tracking frequency f. _dom (t) and phase-locked phase φ _dom (t), together forming the phase-locked reference PLL _ref (t). Specifically, the phase error e _φ (t)=wrap(φ _inst (t)-φ _hat (t)), where wrap() is a function that limits the error to the interval (-π, π]. The update of the second-order loop filter can be performed in discrete time steps k:

[0062] ω _hat [k]=ω _hat [k-1]+K _i *e _φ [k]*Δt;φ _hat [k]=φ _hat [k-1]+(ω _hat [k]+K _p *e _φ [k])*Δt.

[0063] Among them, K _p For proportional gain, K _i The integral gain is denoted by Δt, and the sampling period is Δt. The final output is a phase-locked reference PLL. _ref (t)={f _dom (t),φ _dom (t)}, where the tracking frequency f _dom [k]=ω _hat [k] / (2π), phase-locked phase is φ _dom [k]=φ _hat [k]. ω _hat [k] represents the tracking angular frequency.

[0064] Step 5.3: Based on the discrete impulse response matrix and phase-locked reference, optimize the solution of the interferometric micropulse sequence with the objective of minimizing the pressure energy at the critical section. The system constructs an optimization problem. The objective function is constructed, represented by the interferometric micropulse sequence v. _i [n] (i is the valve index, n is the time step) is processed by the discrete impulse response matrix H[n] (i.e., h). _k , i [n]) After convolution, the weighted sum of the synthetic pressure energy generated at the critical section k.

[0065] Objective function J _v =min _v Σ _k Σ _n w _k •∣p _k_pred [n]∣ 2 , where p _k_pred [n]=Σ _i (h _k,i *v _i [n], where v is the micro-pulse sequence of all valves. _i The vector to be optimized is composed of [n], where * denotes convolution operation, and w _k It is the weight of the k-th cross section, |...| 2 Represents energy. To achieve active destructive interference, the weight w in the weighted sum... _k (or a more refined time weight w)_k [n]), based on the phase-locked reference PLL _ref The instantaneous phase φ of (t) _dom (t), applying higher weights to the phase interval corresponding to the expected pressure peak (i.e., the wave peak). In other words, by optimizing v _i [n], making the synthesized wave p _k_pred [n] near the peak of the dominant mode (by φ) _dom (t)Prediction) generates a trough with an out-of-phase (approximately π) phase difference, naturally converging to phase cancellation under the objective of minimizing energy. Under device constraints of valve stroke, velocity, and jerk (Jerk), the objective function is solved to obtain the interference micropulse sequence v. _i [n].

[0066] Equipment constraints (e.g., |v) _i [n]-v _i [n-1]|≤Maximum speed V _max ,∣v _i [n]-2*v _i [n-1]+v _i [n-2]|≤ Maximum acceleration J _max This ensures that the solved micropulse is physically executable. Optimization problems can often be transformed into a quadratic programming (QP) or second-order cone programming (SOCP) problem for efficient solution.

[0067] Step 5.4: Determine the overlapping time window based on the phase-locked reference. Within the overlapping time window, generate a slow opening template and fuse it with the interference micropulse sequence. Apply trajectory shaping constraints to generate a vibration-suppressing opening template.

[0068] like Figure 3 As shown, in optimizing v _i At the same time or after [n], the system relies on the phase-locked loop (PLL) reference. _ref The instantaneous phase φ in (t) _dom (t), applying the principle of opening first at the trough and closing last at the peak, determines the overlapping time window W for valve switching. _overlap Utilizing hydraulic principles, at the trough of the system pressure (due to φ) _dom (t) Determining) opening the next valve and closing the previous valve at the peak of the wave is an empirical passive vibration suppression method. This invention combines it with active interference. The system operates within the overlapping time window W. _overlap Within, a slowly changing, gradual opening template θ is generated. _i_base (t) (e.g., a smooth ramp or sinusoidal transition curve), and the interference micropulse sequence v obtained in step 5.3. _i [n] is superimposed onto the slow opening template to form the candidate trajectory θ. _i_candidate (t)=θ _i_base(t)+v _i [n] (after interpolation and time alignment). To ensure the smoothness and executability of the trajectory, the system performs interpolation and time alignment on the superimposed trajectory θ. _i_candidate (t) Apply S-shaped shaping to strictly meet the preset jerk (Jerk) and velocity constraints to form the vibration damping opening template θ. _i_prime (t). S-shaped shaping is a trajectory planning algorithm that can ensure that the velocity and acceleration of the trajectory are zero at the start and end points, and that the jerk (Jerk) is bounded throughout the process, making the actuator (valve) move more smoothly and avoiding secondary excitation.

[0069] Example 6: A detailed description of the specific implementation method for performing random control barrier optimization using a split-bar, and outputting an executable valve trajectory, such as... Figure 4 As shown, it includes the following steps:

[0070] Step 6.1: Assemble the MPC state-output model based on the digital twin state. The system constructs a discrete-time dynamic model for Model Predictive Control (MPC). The model can be based on the initialized surrogate model. _base And the equivalent flow area A of the valve can be utilized by online identification. _i (u) and local loss coefficient K _i Update the model (denoted as Model). _MPC The equation can be expressed as: x(t+1)=f(x(t),u(t)); z(t)=g(x(t),u(t)). Here, x(t) is the system state (e.g., gate chamber water level, water level in the water-saving pool), u(t) is the control input (i.e., the valve opening trajectory to be optimized), and z(t) is the system output (e.g., the predicted mooring force-related variables). f is the state transition function, and g is the system output function.

[0071] Step 6.2: Define the safety barrier function based on the mooring force probability distribution. The system obtains the mooring force probability distribution F. _line_dist (t). To ensure mooring safety, the system defines a safety barrier function B(t). For example, the barrier function can be defined as: B(t) = α*F _break -F _line (t). Wherein, F _line (t) is a random variable representing the mooring force (its distribution is given by F). _line_dist (t) description), F _break α is the allowable breaking force of the cable, and α is the safety factor (e.g., α = 0.5). Under this definition, the system is in a safe state when B(t) is defined.

[0072] Step 6.3: Using the empirical distribution of the mooring force probability distribution as the center and setting the Wasserstein radius based on the quality assessment report, construct a set of uncertainty distributions. This is to address the mooring force distribution F... _line_dist Due to the inherent uncertainty of (t) (i.e., the estimation of the distribution may be inaccurate), the split-bulk method is employed. The system uses sample statistics within a rolling window or F... _line_dist (t) itself is the empirical distribution P _emp According to the quality assessment report (QC) _report Define a Wasserstein radius ρ. Specifically, the Wasserstein radius ρ is a measure of the magnitude of uncertainty. A preferred adaptive strategy is to adaptively increase the Wasserstein radius ρ when quality assessment reports indicate a decrease in confidence, increase in residuals, increase in noise, or increase in latency in sensor channels (such as vision or IMU). Conversely, when data quality is good, ρ is decreased. Based on this, the system constructs a set of uncertainty distributions U(ρ). U(ρ) is based on the empirical distribution P. _emp The set of all possible true probability distributions P centered at a radius ρ and bounded by a radius ρ (e.g., a 1-Wasserstein sphere or a 2-Wasserstein sphere).

[0073] Step 6.4: Assemble a set of hard constraints including vibration damping opening template tracking constraints, and construct an objective function including water hammer index, flow smoothness, and water saving cost. The system constructs a multi-objective optimization function J, specifically including the following steps:

[0074] Align data stream D _sync The time-domain fluctuation variance or frequency-domain dominant mode energy of the medium pressure sequence is defined as the HammerIndex.

[0075] MPC state-output model _MPC The predicted change in total flow rate between adjacent time points (e.g., ΔQ(k) = Q) _total (k)-Q _total The square of (k-1) is defined as the flow smoothness;

[0076] The water cost mapped from operating parameters (such as the water-saving tank level and target water exchange volume) in the digital twin state is defined as the water-saving cost (WaterCost). The objective function J is constructed as a pre-defined weighted sum of the water hammer index, flow smoothness, and water-saving cost: J = Σ _k [λ1·HammerIndex(k)+λ2·(ΔQ(k)) 2 +λ3•WaterCost(k)], where λ1, λ2, λ3 are weight coefficients. Simultaneously, the system assembly hard constraint set C _hardThis includes physical constraints (such as valve stroke, speed, and jerk limit) and operating condition constraints (such as gate chamber water level difference and scour limit). Preferably, the set of hard constraints C _hard Includes vibration damping opening template tracking constraints; specifically, it limits the executable valve trajectory u for solving the optimization problem. _i (t), in the rolling time domain, is related to the vibration damping opening template θ. _i_prime The deviation between (t) and (t) shall not exceed the preset tracking tolerance ε. _track That is: |u _i (t)-θ _i_prime (t)∣≤ε _track This ensures that while the system seeks the goals of smooth flow (λ2,λ3) and water saving, it will not sacrifice the water hammer suppression effect (λ1) and force the final trajectory to closely follow the vibration suppression template.

[0077] Step 6.5 transforms the probabilistic constraints of the safety barrier function under the set of uncertain distributions into deterministic robust barrier constraints. The original probabilistic constraints are: inf P∈U(ρ) Pr _P The constraint B(t) is: bar-δ, which requires that for the worst-case (inf) probability distribution P in the uncertainty set U(ρ), the probability (Pr) of the system remaining safe (B(t) remains safe) is given. _P The value must still be higher than 1-δ (where δ is a very small acceptable default threshold, such as 1e-6). Since probabilistic constraints are difficult to solve directly, the system transforms them into deterministic constraints. Preferably, using the equivalent form of second-order cone programming (SOCP), moment bounds, or semidefinite relaxation methods, the worst-case probabilistic constraints under the set of uncertain distributions are reconstructed into deterministic constraints independent of specific probability distributions. For example, if a 2-Wasserstein radius is used and based on the mean and variance, the constraint can be transformed into a second-order cone (SOCP) constraint, in the form of: μ _B -τ•σ _B ≥B. Where μ _B and σ _B These are the mean and standard deviation (F) of the barrier function B(t), respectively. _line_dist (t)), τ is a robust safety parameter calculated based on the radius ρ and the default threshold δ. Deterministic constraints (denoted as C) _barrier It was subsequently incorporated into the MPC optimization problem.

[0078] Step 6.6: In the rolling time domain, solve the optimization problem that minimizes the objective function under the constraints of the MPC state-output model, the set of hard constraints, and the deterministic robust barrier constraint to obtain the executable valve trajectory.

[0079] The system solves a rolling time-domain optimization problem within each control cycle (e.g., 100-200ms): min_u Js.t.Model _MPC C _hard C _barrier .

[0080] In a preferred implementation, since the optimization problem may be nonlinear, a sequential convex approximation algorithm is employed, using the solution from the previous cycle as the initial value for hot-start, to ensure efficient solution within the real-time budget. After solving, the optimal valve opening trajectory u is obtained. _i *(t). Furthermore, the worst-case distribution safety margin can also be calculated. _worst (t), for example, margin _worst (t)=μ _B -τ•σ _B A margin value (if positive) represents the safety margin the system retains in the worst-case scenario, and can be used to monitor and record the system's robustness and safety. (The last part, "u," appears to be a typo and can be omitted.) _i *(t) Inject the digital twin for final consistency verification. Once passed, issue the executable valve trajectory u. _i_exec (t) to the valve PLC controller.

[0081] Example 7 provides a preferred implementation method, detailing the process of online calibration and adaptive closed-loop circuitry to continuously improve the matching degree and accuracy of vibration suppression and safety control. The closed-loop calibration circuitry is mainly used for adaptive tracking of dynamic changes (such as valve wear, changes in water flow conditions, and sensor drift).

[0082] Preferably, the closed-loop calibration circuit includes the following steps:

[0083] Execute control commands and collect incremental data; specifically, the executable valve trajectory u _i_exec (t) The command is sent to the valve controller for execution. During execution, the system continuously collects newly generated data, such as newly added pressure p before and after the valve. _up_new (t), p _down_new (t), Gate chamber water level η _new (t) and the new video frame sequence Video _new Inertial IMU _new AIS _new Data such as these are used to form the execution-period incremental dataset D. _exec .

[0084] Perform online identification of valve characteristics, specifically, using the execution period dataset D. _execTo obtain novel valve characteristics without compromising normal navigation safety, a very small-amplitude random micro-step signal can be injected into a very short, safety-free micro-window (e.g., at the end of valve operation or under predetermined operating conditions). A recursive identification algorithm (e.g., recursive least squares method) is then used to identify key hydraulic characteristic parameters of the valve online based on the flow rate and differential pressure responses corresponding to the micro-excitation, such as the valve's equivalent flow area-opening curve A. _i (u) and local loss coefficient K _i The identified updated parameter A _update_i (u) and K _update_i It will be written back to update the Model. _MPC The state-output model makes MPC predictions more closely resemble physical reality.

[0085] Perform an impulse response update, specifically, using the execution-time dataset D. _exec and the updated valve characteristic A _update_i (u), K _update_i Due to changes in valve characteristics (A) _i The change in (u) will directly lead to a change in the hydraulic shock response H[n]. The system can update the response matrix based on the new identification data or through model correction to obtain H. _update [n]. Updated H _update [n] will be written back for the next round of interference micropulse optimization, ensuring that the vibration suppression template is always optimized for the latest system dynamic characteristics.

[0086] The mooring distribution and radius are adaptively adjusted; specifically, newly acquired video is used. _new IMU _new p _up_new Based on the data and following the logic of non-contact observation and probabilistic fusion, a new probability distribution F of the mooring force was recalculated. _line_dist_new (t) and its confidence level conf _line_new (t). The system compares the predicted values ​​of the new distribution with the actual observed values ​​(or residuals) and adaptively adjusts the Wasserstein radius ρ based on the magnitude of the residuals. Specifically, when the observed residuals increase significantly (indicating a decline in model or data quality and increased uncertainty), the system automatically and preventively increases the Wasserstein radius ρ. _new This makes robust barrier constraints (C) _barrier The residual becomes more conservative to ensure safety; conversely, if the residual remains very small, ρ can be appropriately reduced to improve operating efficiency.

[0087] All updated parameter A _update_i (u), K _update_i H _update [n]、F_line_dist_new (t) and ρ _new All were written back to the digital twin state Twin _state (t) and Model _MPC In this process, as the starting point for the next rolling optimization cycle, a complete adaptive closed loop of perception-decision-execution-learning is formed.

[0088] Example 8 provides a specific and reproducible numerical calculation example of hydraulic interference vibration suppression and hydraulic phase-locked loop (HIS+H-PLL).

[0089] Consider a simplified application scenario: there is a switching operation between two water-saving tank valves (i=1,2), and the hydraulic disturbances mainly converge at the critical section (k=1). The sampling period Δt of the system control is set to 0.1 seconds.

[0090] Input: During the operation of the H-PLL (Hydraulic Phase-Locked Loop) system, the switching operation of the water-saving tank triggered hydraulic oscillations in the pipeline network. The H-PLL module started working. Assume the system acquires the pressure sequence p from the pressure sensor at the critical section k=1. _down (t). The H-PLL module performs bandpass filtering, Hilbert transform, and phase unwrapping on the signal, and successfully locks onto a dominant water hammer mode through a second-order phase-locked loop filter. The output phase-locked reference PLL _ref (t) indicates: dominant frequency f _dom =0.2Hz (i.e., oscillation period T=5 seconds), and outputs a continuous phase-locked phase φ synchronized with the dominant mode in real time. _dom (t).

[0091] While the H-PLL is operating, the system superimposes pseudo-random sequence (PRBS) micro-excitations onto valves 1 and 2 within the safe injection window. The recursive least squares (RLS) algorithm is employed, with valve position s as the basis for the impulse response identification. _gate_1 (t), s _gate_2 (t) is the input, with the critical section pressure p _down (t) represents the output, identifying the discrete impulse response H[n] from each valve to the critical section. Assume (exemplary) the identified FIR (finite impulse response) sequence (n=0,1,2... corresponding to 0s,0.1s,0.2s...) is as follows:

[0092] h _1 ,1[n]={0,0,0.5,0.8,0.3,-0.2,...}(Response of valve 1, starting to respond at 0.2s);

[0093] h _1,2[n]={0,0,0,0.2,0.6,0.4,-0.1,...}(The response of valve 2 has an additional delay of 0.1s);

[0094] Furthermore, the interference micropulse optimization is performed. The system is structured as an optimization problem, and the interference micropulse sequence v is solved. _1 [n] and v _2 [n], the objective is to minimize the synthetic pressure energy at the critical section: J _v =min _v Σ _n w _1 [n]•∣p _k_pred [n]∣ 2 Among them, predicted pressure:

[0095] p _k_pred [n]=(h _1 ,1*v _1 )[n]+(h _1 ,2*v _2 [n] (* represents convolution).

[0096] The weights are set, and the system determines the phase-locked phase φ output by the H-PLL. _dom (t), dynamically setting time weight w _1 [n]. For example, when φ _dom The phase φ corresponding to (t) _dom When [n] falls within the interval [π / 2, 3π / 2] (indicating that a pressure peak is about to occur or is currently occurring), w will... _1 [n] is set to a high weight (e.g., w) _1 [n]=10.0); in other phase intervals (troughs), set to low weight (e.g., w). _1 [n]=1.0). This weight setting forces the optimization algorithm to concentrate its efforts to counteract the high-energy pressure pulsations that occur at the peak phase, achieving more efficient destructive interference.

[0097] At the same time, optimization is subject to strict device constraints, such as (example values):

[0098] Velocity constraint: |v _i [n]-v _i [n-1]∣ / ]]beaming is strictly subject to unit stroke / second);

[0099] Jerk constraint: |v _i [n]-2*v _i [n-1]+v _i [n-2]∣ / Δt 2 ≤Δtvk) unit travel / second 3 );

[0100] Furthermore, the system (e.g., using a QP solver) solves a constrained optimization problem to obtain the optimal interferometric micropulse sequence, for example:

[0101] v _1 [n]={...,0,0.00,0.01,-0.03,0.02,0,...};

[0102] v _2 [n]={...,0,0.00,0.00,0.02,0.01,-0.04,0.01,...} (These sequences are small, rapidly changing gate position adjustments that, when superimposed in opposite phases, can generate destructive waves at critical sections).

[0103] Furthermore, template generation and shaping are performed, based on the PLL. _ref Using the phase of (t), the trough-first-open, peak-last-close criterion is applied to determine the overlap time window W between valve 1 closing and valve 2 opening. _overlap (For example, 2.0 seconds). The system generates a slow opening template θ within this window. _i_base (t) (e.g., θ) _1_base (t) decreases from 1.0 to 0, θ _2_base (t) rises from 0 to 1.0, all as smooth curves), and the micropulse sequence v _1 [n] and v _2 [n] is superimposed onto the corresponding slow template. An S-shaped shaping algorithm is applied to the superimposed trajectory. This algorithm ensures that the final output trajectory strictly satisfies the set jerk and velocity constraints, making it relatively smooth and avoiding high-frequency excitation introduced by the control signal itself. The final generated vibration-suppressing opening template θ _i_prime (t), which is θ _i_base (t) is superimposed with v _i [n] and a smooth control curve that has been S-shaped and can be directly issued for execution.

[0104] Example 9: A multi-level water-saving ship lock navigation safety assurance system based on digital twins is provided, specifically including:

[0105] The data aggregation unit is used to aggregate multi-source sensor data, generate aligned data streams and quality assessment reports, and update the digital twin status. In a specific implementation, the data aggregation unit can be deployed on the server or edge computing node of the lock control center, responsible for communication and data access with field programmable logic controllers (PLCs), water level gauges, pressure sensors, video surveillance, IMUs, AIS receivers, etc. It performs strict time alignment (e.g., using causal splines or piecewise cubic Hermite interpolation), coordinate and scale registration (e.g., using the EPnP+RANSAC algorithm), and data quality assessment (e.g., removing glitches through Hampel filtering and estimating noise power spectral density), ultimately outputting a highly synchronized aligned data stream D that can be used by subsequent units. _sync Quality Control (QC) report that quantifies uncertainty _report .

[0106] The probabilistic observation unit is used to obtain the mooring force probability distribution through non-contact observation and probabilistic fusion, utilizing aligned data streams and quality assessment reports. Specifically, the probabilistic observation unit can be implemented as a set of software algorithm modules running on a server, receiving D... _sync Visual and inertial data are used to obtain the cable geometry sequence through semantic segmentation and line structure tracking. Prior tension is calculated based on the cable material properties, and hydrodynamically corrected tension (i.e., estimation of relative velocity and pulsating pressure components) is estimated by combining pressure, water level, and ship dynamic data. Within a probabilistic fusion framework (preferably using unscented Kalman filtering (UKF)), tension estimation and QC are fused. _report Sensor uncertainty information is generated, and the real-time output includes the mean μ. _line and standard deviation σ _line The probability distribution of mooring force F _line_dist (t).

[0107] A vibration suppression template generation unit is used to identify the hydraulic shock response based on digital twin state and aligned data stream, and generate a vibration suppression opening template using hydraulic phase-locked loop tracking. Specifically, it includes two key sub-modules: a hydraulic shock response (HIS) identification module and a hydraulic phase-locked loop (H-PLL) module. The HIS module applies a micro-excitation (such as PRBS) within the safe injection time window and uses a recursive least squares (RLS) algorithm (preferably including Tikhonov regularization) to identify the shock response matrix H[n] online. The H-PLL module extracts the instantaneous phase of the pressure signal through bandpass filtering and Hilbert transform, and uses a second-order phase-locked loop filter to dynamically track the frequency f of the dominant water hammer mode. _dom With phase φ _dom To form a phase-locked reference PLL _ref(t). By solving a quadratic programming (QP) optimization problem, the interferometric micropulse sequence v with high weight at the peak phase and minimizing pressure energy is calculated. _i [n]. Based on this, v _i [n] is merged with the slow opening template and an S-shaped shaping constraint is applied to generate a smooth opening template θ with destructive interference effect. _i_prime (t).

[0108] The sub-bulb control unit is used in a digital twin system to integrate the probability distribution of mooring force with the vibration damping opening template, perform sub-bulb stochastic control barrier optimization, and output an executable valve trajectory. The sub-bulb control unit operates within a model predictive control (MPC) framework, equipped with an MPC state-output model. _MPC Define the safety barrier function (B(t)=α*F). _break -F _line (t)), and based on QC _report Define the Wasserstein radius ρ and construct the uncertainty distribution set U(ρ). Construct a multi-objective function J (including water hammer index, flow smoothness, and water saving cost) and equip it with hard constraints (including template tracking constraint |u). _i -θ _i_prime |≤ε _track The worst-case probability constraint of the mooring force (inf Pr{B≥0}≥1-δ) is transformed into a deterministic robust barrier constraint (C). _barrier Preferably, a second-order cone programming (SOCP) equivalent form is adopted. Employing techniques such as the sequential convex approximation (SQP) and hot-start technology, the optimization problem is solved in the rolling time domain, outputting an executable valve trajectory u that satisfies both vibration suppression requirements and robust safety. _i_exec (t).

[0109] Furthermore, in some alternative implementations, the system may also include an online calibration unit for performing an adaptive closed-loop function. This involves injecting a small stimulus to identify the updated valve characteristics (A) online. _update_i (u) and impulse response (H) _update [n]), and adaptively adjust the Wasserstein radius ρ based on the observed residuals. _new .

[0110] It is understood that, in terms of physical implementation, the aforementioned unit may be a software program module, firmware, application-specific integrated circuit (ASIC) or field-programmable gate array (FPGA) logic running on one or more central servers, edge controllers or cloud computing platforms, or any combination thereof.

[0111] Example 10 provides a specific implementation method for obtaining real-time dynamic information on navigation water levels, supplements the hydraulic model, and describes the specific technical methods of the navigation water level spatiotemporal forecasting module.

[0112] In this embodiment, the system accesses forecast data released by water conservancy and meteorological departments through a data interface, as well as discrete station data collected in real time by a large number of sensors (such as water level gauges, flow meters, and anemometers) deployed upstream and downstream of the lock, in the water-saving pool, and in the lock chamber.

[0113] To fuse these discrete station data into a continuous dynamic field covering the entire waterway, forming a digital water surface, Kriging interpolation is preferably used. Kriging interpolation is an optimal linear unbiased estimation method based on spatial (or spatiotemporal) autocorrelation.

[0114] Specifically, its main formula can be expressed as: Z _est (s0,t0)=Σ _i (w _i *Z _obs (si,ti)); where: Z _est (s0,t0) represents the water level (or wave intensity) value at the spatiotemporal point (location s0, time t0) to be predicted. _obs (si,ti) represents the known observation value of the i-th monitoring station (location si, time ti). _i This represents the weighting coefficient assigned to the i-th monitoring station.

[0115] Unlike simple inverse distance weighting, the weight coefficient w _i Determining the weighting factor is crucial for Kriging interpolation. The system needs to analyze historical data to construct a spatiotemporal variability function or covariance function to quantify the correlation (or difference) between observations at any two spatiotemporal points (e.g., (si,ti) and (sj,tj)). The weighting coefficient w... _i The value Z is determined by solving a system of linear equations. _est (s0,t0) satisfies the properties of unbiasedness (i.e., the expected value of the estimation error is zero) and optimality (i.e., the estimation variance is minimized).

[0116] Using the methods described above, the system can fully utilize the spatiotemporal correlation of data to generate a high spatiotemporal resolution real-time dynamic information map of navigable water levels. It can also intuitively display short-term (e.g., 1-3 hours) water level change trends and wave intensity distribution, which can be used in Modeling. _MPC The boundary condition input, or the constraint used for intelligent gear shifting decision.

[0117] Example 11 provides a specific implementation of the intelligent lock passage model, and describes the algorithm of the lock safety operation calculation module to calculate the optimal ship passage sequence and berth allocation scheme in the lock chamber.

[0118] Solving a complex multi-objective optimization problem using a model, the objective function can be expressed as: Minimize: F(X), where X is the berthing scheme (i.e., the sequence of ship positions in the lock chamber). F(X) is the comprehensive cost function, preferably considering at least the following three conflicting objectives:

[0119] Gate chamber area utilization rate f _area (X), the goal is to minimize the wasted area of ​​the gate chamber.

[0120] Total waiting time for all ships f _wait (X), the goal is to minimize the total time and improve efficiency.

[0121] Safety margin between ships f _safety (X) aims to maximize the minimum safe distance to ensure safety.

[0122] Therefore, F(X) = k1*f _area (X)+k2*f _wait (X)-k3*f _safety (X), where k1, k2, k3 are weighting coefficients. The optimization problem is subject to many constraints, such as ships cannot overlap, lock chamber size restrictions must be met, and loaded and unloaded ships need to be properly balanced to ensure mooring equilibrium.

[0123] Since optimization problems are high-dimensional, multi-constrained, and nonlinear (NP-hard problems), it is preferable to use an improved particle swarm optimization (PSO) algorithm for intelligent optimization solutions.

[0124] The standard PSO algorithm iteratively updates the particle velocity and position by simulating the foraging behavior of bird flocks:

[0125] Speed ​​update: v _id (t+1)=w*v _id (t)+c1*r1*(pbest _id -x _id (t))+c2*r2*(gbest _d -x _id (t));

[0126] Location update: x _id (t+1)=x _id (t)+v _id (t+1);

[0127] Where i is the particle index, d is the dimension, t is the iteration number, and x is the particle index. _id It's a position, v _idHere, w is the velocity, c1 and c2 are the inertia weights, r1 and r2 are random numbers between (0 and 1), and pbest is the learning factor. _id It is the best position in the history of an individual particle, gbest _d It is the globally optimal position for the population.

[0128] To address the issues of standard PSO easily getting trapped in local optima and having unstable convergence speed, the following measures are taken:

[0129] 1. Set adaptive inertia weights. The inertia weight w is no longer a fixed value, but decreases dynamically with the number of iterations t: w(t) = w _max -(w _max -w _min )*(t / t _max )where w _max It is the maximum inertia weight (e.g., 0.9), w _min It is the minimum inertia weight (e.g., 0.4), t _max This represents the maximum number of iterations. In the early stages of the search (when w is larger), it grants particles stronger global exploration capabilities, while in the later stages (when w is smaller), it enhances the algorithm's local fine-grained search capabilities.

[0130] 2. Set asynchronously changing learning factors. Learning factors c1 and c2 are no longer fixed values, but change asynchronously: c1(t) = c1 _start -(c1 _start -c1 _end )*(t / t _max c2(t) = c2 _start +(c2 _end -c2 _start )*(t / t _max For example, we can linearly decrease c1 from 2.5 to 0.5, while simultaneously linearly increasing c2 from 0.5 to 2.5. This allows the algorithm to emphasize independent individual exploration in the early stages (individual cognition c1 is large), and to emphasize learning from the optimal group in the later stages (social cognition c2 is large), achieving a smooth transition from divergent exploration to convergent refinement. _start Let c1 be the initial value of the learning factor. _end This is the end value of c1. c2 _start Let c2 be the initial value of the learning factor c2. _end This is the end value of c2.

[0131] 3. Introduce a contraction factor. To ensure the convergence of the algorithm and effectively control the particle velocity to prevent its explosive growth, a contraction factor K is introduced.

[0132] The speed update formula becomes: v _id (t+1)=K*[v _id (t)+c1*r1*(pbest_id -x _id (t))+c2*r2*(gbest _d -x _id (t))];

[0133] Where K = 2 / ∣2 - φ - sqrt(φ) 2 -4*φ)∣, φ=c1+c2, and φ>4. For example, if we take c1=c2=2.05, then φ=4.1, and we can calculate that K is approximately equal to 0.729.

[0134] 4. Perform mutation operations using a hybrid genetic algorithm. To help the population escape local optima, after each iteration, mutations are performed with a certain probability (e.g., mutation probability p). _m =0.1) for the globally optimal particle gbest _d Perform mutation operation:

[0135] gbest _d_mut =gbest _d *(1+N(0,1)*σ _mut ); where N(0,1) is a random number from a standard normal distribution, σ _mut It is a variable-length time that decreases with the number of iterations. It is preferably triggered when the global optimal solution has not shown significant improvement for several consecutive iterations.

[0136] Through the above improvements, the intelligent shifting model can efficiently and robustly solve the intelligent shifting problem of ships in large water-saving locks, providing scientific scheduling decisions for ship passage scheduling and water-saving pool control modules.

[0137] Example 12 provides a specific implementation method for the water-saving pool control safety decision-making model, and describes another important algorithm of the lock safety operation calculation module, which can be used as an alternative to the sub-bar control unit or a pre-verification scheme.

[0138] Given a valve operation plan for a water-saving tank (e.g., a valve opening curve), the decision model uses a deterministic hydrodynamic model to determine whether the operation plan will cause the ship's mooring force inside the lock chamber to exceed a safety threshold. Preferably, it includes the following steps:

[0139] Step 1: Perform hydraulic disturbance prediction. Specifically, the system acquires a valve operation scheme for the water-saving tank to be evaluated, which can be represented as a curve u(t) showing the valve opening change over time. Based on u(t), the system predicts the water flow rate Q(t) flowing into or out of the gate chamber. For example, the prediction can be calculated based on the valve characteristic curve and Bernoulli's equation: Q(t) = C _d *A(u(t))*sqrt(2*g*Δh(t)); where C _dis the flow coefficient, A(u(t)) is the flow area related to the valve opening u(t) (which can be a lookup value or a fitting function), g is the gravitational acceleration, and Δh(t) is the real-time water level difference between the water-saving pool and the gate chamber at that moment.

[0140] Step 2: Perform hydrodynamic calculations for the lock chamber. Specifically, the system uses the flow rate Q(t) calculated in Step 1 as the input boundary condition and, using the hydrodynamic model, calculates the rise and fall rate dV / dt of the water level in the lock chamber and the longitudinal or transverse flow velocity V at key points (e.g., near the bow, stern, or cable guide opening). boat (t). In a preferred embodiment, the hydrodynamic model may employ shallow water equations, such as the one-dimensional or two-dimensional Saint-Venant equations, which describe the dynamic processes of water surface fluctuations and velocity fields within the gate chamber as the flow rate Q(t) changes. In another alternative, computationally simpler embodiment, a simplified lumped parameter model may be used, for example, treating the gate chamber as a large water tank, calculating only the average rise and fall rates of the water level, and estimating the velocity at key points based on flow conservation.

[0141] Step 3: Perform ship hydrodynamic calculations. Specifically, the system calculates the flow velocity V at key points based on the values ​​obtained in Step 2. boat Let F(t) be the hydrodynamic force (primarily drag force) exerted by the water flow on the ship. For example, the hydrodynamic force F(t) can be calculated using hydrodynamic formulas, typically proportional to the square of the flow velocity:

[0142] F(t) = 0.5 * ρ _w *C _D_ship *A _ship *(V boat (t)) 2 ; where ρ _w It is the density of water, C _D_ship A is the drag coefficient of the ship at that current speed and attitude (which can be determined according to the ship type). _ship It is the lateral or longitudinal stress-bearing area of ​​the underwater part of a ship.

[0143] Step 4: Calculate the mooring force and determine safety. Specifically, the system uses the hydrodynamic force F(t) calculated in Step 3 as an external load and inputs it into the preset static and dynamic models of the ship mooring system. The model considers the ship's mass, moment of inertia, and the geometric arrangement, length, and material properties (such as stiffness) of each mooring line. After solving the model, the force Fm(t) distributed on each mooring line can be obtained. Safety determination is then performed: the maximum mooring force max(Fm(t)) among all mooring lines is compared with the breaking strength F of the mooring line. _break (Or multiply by a safety factor α, for example, α=0.5). If max(Fm(t))<α*F _breakIf the system determines that the water-saving operation plan u(t) is safe and can be executed, then the system can proceed. If max(Fm(t)) ≥ α*F _break If the system fails to meet the safety constraints, it will issue an early warning and automatically generate a more lenient valve operation plan (e.g., reduce the valve opening speed, use step-by-step opening, extend the switching time, etc.), and return to step 1 to recalculate until a plan that meets the safety constraints is found. The model outputs a safety-verified and optimized sequence of water-saving tank valve control commands, u. _safe (t).

[0144] Example 13: For a large water-saving ship lock with three-stage water-saving tanks (i.e., low-level, middle-level, and high-level water-saving tanks), a specific implementation method of preset operating rules for the water-saving tanks is provided. The vibration suppression and robust control technology is applied to the valve switching process defined by the rules. The preset operating rules include:

[0145] The lock filling process (for upstream vessels) specifically includes: opening the low-level water-saving pool valve, at which point the water level in the low-level pool is higher than the initial water level in the lock chamber, and the low-level pool begins to fill the lock chamber. When the water stored in the low-level pool is about to be depleted (e.g., as monitored by a level gauge), the low-level pool valve is quickly closed, and simultaneously (or with a very short delay) the middle-level water-saving pool valve is opened, allowing the middle-level pool to continue filling the lock chamber. When the water stored in the middle-level pool is about to be depleted, the middle-level pool valve is quickly closed, and simultaneously the high-level water-saving pool valve is opened, allowing the high-level pool to continue filling the lock chamber. When the water stored in the high-level pool is about to be depleted, the high-level pool valve is quickly closed, and simultaneously the main passage filling valve of the lock chamber is opened, allowing water to continue filling the lock chamber from the upstream approach channel. At the end of the filling period, when the difference between the upstream water level and the gate chamber water level is only a small difference (e.g., 0.25 meters to 0.5 meters), the water flow is started to close the filling valve (i.e., the valve is closed before the water flow has completely stopped), so that when the valve is closed, the reverse head borne by the upper gate miter gate does not exceed the preset threshold (e.g., 0.25 meters).

[0146] The lock water discharge procedure (for downstream vessels) specifically includes: opening the high-level water-saving pool valve, at which point the lock chamber water level is higher than the high-level pool, and the lock chamber begins discharging water into the high-level pool. When the high-level pool is about to reach the set water level, the high-level pool valve is quickly closed, and simultaneously the middle-level water-saving pool valve is opened, allowing the lock chamber to continue discharging water into the middle-level pool. When the middle-level pool is about to reach the set water level, the middle-level pool valve is quickly closed, and simultaneously the low-level water-saving pool valve is opened, allowing the lock chamber to continue discharging water into the low-level pool. When the low-level pool is about to reach the set water level, the low-level pool valve is quickly closed, and simultaneously the lock chamber main corridor discharge valve is opened, allowing the lock chamber to continue discharging water into the downstream approach channel. At the end of the discharge phase, when the lock chamber water level and the downstream approach channel water level have only a small difference, the discharge valves are closed by kinetic energy flow, ensuring that when the valves are fully closed, the reverse head borne by the lower lock head miter gate does not exceed a preset threshold (e.g., 0.25 meters).

[0147] It is understandable that the rapid closing and simultaneous opening switching operations in the above process are the main sources of severe water hammer. Vibration-damping opening templates and executable valve trajectories are preferably applied to these critical switching moments to replace simple rapid switching actions, achieving active vibration damping and safety assurance.

[0148] According to one aspect of this application, the selection of the key section and injection window is specifically as follows:

[0149] Obtain digital twin status and historical operation data, filter and sort them according to the sensitivity of the cross section to valve position disturbance and the peak event occurrence rate, and form a list of key cross sections as direct objects for subsequent identification and optimization.

[0150] Obtain the aligned data stream (current operating conditions) and equipment constraints, establish the micro-excitation injection window without affecting navigation, and specify the maximum injection energy and allowable frequency band to ensure that online identification has both observability and safety boundaries.

[0151] Set injection interlock conditions to disable injection by default during berthing and mooring phases, provide manual overriding paths, and output injection strategies for subsequent online identification and direct reference, and verify them during execution.

[0152] According to one aspect of this application, online micro-stimulus and impulse response identification specifically includes:

[0153] Based on the injection time window and equipment constraints, a low-amplitude micro-excitation scheme (including timing and amplitude arrangement) is designed to ensure that it does not exceed the safety ratio of the gate speed and gate jerk upper limit, thus obtaining a micro-excitation plan for each valve.

[0154] While executing the micro-excitation plan, the upstream pressure, downstream pressure, and valve position of the valve are collected, uniformly bandpassed to the water hammer frequency range, and denoised to obtain filtered identification data, which is used to improve the signal-to-noise ratio of subsequent parameter estimation.

[0155] The memory length, forgetting factor, and regularization strength of the online identification model are set, and a micro-incentive program is used to ensure sufficient persistent incentives, providing robust initial values ​​for continuous updates.

[0156] The filtered identification data is input into the online algorithm, and the discrete impulse response is continuously updated according to the valve-section channel. The maximum order is limited and a regularization term is introduced to suppress overfitting. The responses of each channel are combined according to the time sequence and the section dimension to generate a response matrix, which serves as the basis for subsequent interference optimization.

[0157] The newly obtained response matrix is ​​compared with the response under similar historical conditions. When the matching degree is insufficient or the residual is abnormal, the injection is extended or the forgetting factor is adjusted until it passes the verification and is written back to the digital twin state for use in interference optimization and trajectory shaping.

[0158] According to one aspect of this application, a hydraulic phase-locked ring specifically comprises:

[0159] The pressure sequence is obtained from the aligned data stream, and bandpass processing and signal analysis are performed for the main frequency range of water hammer to construct the signal. The instantaneous phase and instantaneous amplitude are extracted to provide measurement quantities for phase tracking.

[0160] A second-order discrete phase-locked loop is constructed. A suitable closed-loop bandwidth and damping are selected between response speed and jitter suppression. The dominant frequency is obtained through sliding estimation, and the loop is used in conjunction with a loop filter to output the phase-locked phase. A locking criterion is then established.

[0161] When the locking criterion is not met, the lock is unlocked according to the established procedure and a small-range frequency search or multi-band parallel search is performed. After the lock is quickly restored, the phase-locked reference continues to be output.

[0162] According to one aspect of this application, the interference micropulse optimization specifically includes:

[0163] By combining the response matrix, the list of key sections, and the phase-locked reference, the optimization objective is to minimize the pressure energy of key sections within the switching window, and weights are set according to the importance and safety margin of the sections to form the objective definition.

[0164] By combining equipment constraints with execution boundaries such as gate accelerometer limit and gate speed limit, the mapping between interference commands and the final gate trajectory is incorporated into the same constraint system, and template tracking error limits are set to form a constraint set.

[0165] Choose either quadratic programming or second-order cone programming based on the objective definition and constraint set, and set the solution tolerance, iteration upper limit and time budget to ensure that the rolling solution meets the real-time requirements.

[0166] Output interference micropulses for each valve; if an infeasibility occurs, shrink the cross-sectional weights or shorten the window in a predetermined order to restore a feasible solution; then implement amplitude and frequency band limiting to ensure compatibility with the execution boundary and obtain the interference result after limiting.

[0167] According to one aspect of this application, the overlapping flexible switching and trajectory shaping specifically includes:

[0168] Based on phase-locked reference, the overlap time window between valves is determined according to the principle of opening first at the trough and closing last at the peak, and the necessary advance and delay amounts are calculated.

[0169] A slow-changing gate position template is generated within the overlapping time window to maintain the continuity of the total flow; the interference result after amplitude limiting is superimposed on it to form a candidate trajectory, leaving a boundary margin for execution.

[0170] The candidate trajectory is shaped with gate jerk constraints, and the gate velocity and gate jerk are checked to see if they are within the allowable range. The vibration suppression gate template is injected into the digital twin scene for rapid pre-simulation. If peak pressure, local scour or water level difference exceeds the limit, the section weight or sliding time window is adjusted backtracked until the verification is passed. The final vibration suppression gate template and overlapping time window are output and written into the twin state.

[0171] According to one aspect of this application, the barrier function and distribution set are constructed as follows:

[0172] The mooring force distribution is summarized within a rolling window to form an empirical distribution for each cable; the ensemble radius is set to cover observation noise and modeling bias by combining the channel confidence and residual levels in the quality assessment report.

[0173] Based on the allowable mooring safety threshold, the safety barrier is defined as the difference between the allowable threshold and the estimated mooring force, and this barrier is discretized into the optimization time domain to form a discrete barrier sequence.

[0174] Centered on the experience distribution, construct a distribution set according to the set radius; combined with the default upper limit, form a barrier set oriented towards robust transformation.

[0175] According to one aspect of this application, the objective function is assembled with hard constraints, specifically as follows:

[0176] Obtain pressure information and digital twin status from the aligned data stream, and define water hammer indicators (which can be time-domain pulsating energy or master mode energy); define the change in total flow rate between adjacent time points as a smoothing term based on the basic model.

[0177] Obtain operating parameters and map water-saving demand into water consumption costs; based on this, combine vibration damping gate templates and overlapping time windows to install hard constraints such as gate accelerator limit, gate speed limit, gate travel limit, water level difference limit, and scour limit, and set template tracking error limits.

[0178] The water hammer index, smoothing term, and water usage cost are weighted and aggregated to form an objective function, which serves as the performance metric for real-time optimization.

[0179] According to one aspect of this application, the deterministic transformation of the Brux barrier is as follows:

[0180] Obtain the barrier set, and based on the set definition and numerical conditions, prioritize the use of the cone-shaped equivalent form for deterministic transformation; when there are stronger tails or correlations, moment bounds or semidefinite relaxation can be used.

[0181] Map the set radius and the upper limit of default to key parameters in the transformation, check the numerical conditions to avoid ill-conditioned conditions, and obtain the transformation parameters after completion.

[0182] Constraints on multiple cables are constructed in parallel; when different cables have coupling effects, they are handled in a joint or grouped manner to form the final robust barrier set.

[0183] According to one aspect of this application, real-time solving and warm start-up are specifically as follows:

[0184] The state and output model, objective function, set of hard constraints, and set of robust barriers are integrated to form a rolling optimization problem. In each cycle, a sequential convex approximation is used to perform linear or second-order approximation near the previous solution, and a trust region scheduling is set.

[0185] Using the executed gate trajectory from the previous cycle as the initial solution, we set the single-cycle solution time and upper tolerance bound, and configured the corresponding optimized solver to ensure the real-time performance of the online application.

[0186] After solving the problem, the optimal gate trajectory is obtained. When it is not feasible, the set radius is tightened or the water hammer and smoothing weights are increased in a predetermined order, and the template tracking error is appropriately relaxed within a safe range to restore feasibility. At the same time, the worst-case distribution safety margin (i.e. the difference between the allowable fracture safety boundary and the estimated risk boundary) is calculated for operation monitoring and archiving.

[0187] The optimal gate position trajectory is injected into the digital twin scenario for rapid consistency verification; after passing the verification, executable gate position instructions are generated and the worst-case distribution safety margin is recorded for use by the monitoring module and adaptive adjustment.

[0188] This invention proposes an active vibration suppression scheme based on a hydraulic phase-locked loop (H-PLL) and online system identification. Using H-PLL technology, the instantaneous phase and frequency of the pressure sequence are extracted in real time, enabling dynamic tracking of the dominant water hammer mode. Based on a real-time phase-locked reference and combined with the online identified discrete impulse response matrix, the system can actively generate a series of interference micro-pulse sequences through optimization. After being fused and shaped into a vibration suppression opening template, the resulting pressure wave and the original water hammer main wave form destructive interference at the key section, actively suppressing hydraulic oscillations. This solves the problem that passive vibration suppression strategies cannot track the instantaneous phase and frequency drift of the dominant water hammer mode in real time, making it difficult to achieve active destructive interference.

[0189] This invention proposes a non-contact observation and probabilistic fusion technique. Utilizing multi-source data such as visual and inertial sensors, and combining uncertainty information from sensor quality assessment reports, it outputs a mooring force probability distribution including the mean and standard deviation through nonlinear filtering (such as UKF). This provides crucial uncertainty quantification input for subsequent robust control. Building upon this, the invention further performs distributed stochastic control barrier optimization (DR-SCBF-MPC). Instead of relying on a single deterministic threshold, it constructs a set of uncertainty distributions around an empirical distribution (whose radius ρ is determined by data quality QC). _report Adaptive setting is used to ensure that the safety barrier function still satisfies the safety constraints under the worst-case probability distribution within the set. The method of transforming probabilistic constraints into deterministic robust barrier constraints (such as second-order cone constraints) provides provable safety guarantees and solves the problem that deterministic safety strategies lack effective quantification and robust control methods for the uncertainty of mooring forces.

[0190] Building upon this, the present invention further integrates the two aforementioned schemes, applying a tracking constraint on the vibration damping opening template during optimization of the sub-bulb control unit. This ensures that the system, while performing active vibration damping, also meets the robust safety requirements for uncertainties in mooring forces.

[0191] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A multi-stage water-saving ship lock navigation safety guarantee method based on digital twinning, characterized in that, The method comprises the following steps: Collecting multi-source sensing data, generating aligned data stream and quality evaluation report, updating digital twin state, the multi-source sensing data comprising: gate chamber water level sequence, pressure before and after the valve, valve position, visual and inertial measurement data of ship position and attitude, historical operation data; Using the aligned data stream and the quality evaluation report, the mooring force probability distribution is obtained through non-contact observation and probability fusion; Based on the digital twin state and the aligned data stream, the hydraulic shock response is identified and the hydraulic phase-locked tracking is adopted to generate the vibration suppression opening template; In the digital twin system, the mooring force probability distribution and the vibration suppression opening template are integrated to execute the distribution robust stochastic control barrier optimization, and an executable valve trajectory is output; Wherein, based on the digital twin state and the aligned data stream, the hydraulic shock response is identified and the hydraulic phase-locked tracking is adopted to generate the vibration suppression opening template, comprising: Applying a micro-excitation in a safe injection time window, identifying the discrete shock response matrix of the valve to the key section according to the pressure and position response in the aligned data stream; Extracting the pressure sequence in the aligned data stream, applying the hydraulic phase-locked loop to dynamically track the frequency and instantaneous phase of the water hammer dominant mode to form a phase-locked reference; Based on the discrete shock response matrix and the phase-locked reference, the interference micro-pulse sequence is optimized and solved to minimize the key section pressure energy; According to the phase-locked reference, the overlap time window is determined, and in the overlap time window, the slow opening template is generated, the interference micro-pulse sequence is superimposed on the slow opening template, and the trajectory shaping constraint is applied to generate the vibration suppression opening template.

2. The method of claim 1, wherein, Applying the hydraulic phase-locked loop to dynamically track the frequency and instantaneous phase of the water hammer dominant mode to form a phase-locked reference, comprising: Applying band-pass filtering and Hilbert transform to the pressure sequence to obtain an analytical signal; Solving the analytical signal to obtain a continuous instantaneous phase trajectory; Setting a second-order phase-locked loop filter containing proportional gain and integral gain; Real-time calculating the phase error of the instantaneous phase trajectory and the reference phase inside the filter; Using the phase error to drive the second-order phase-locked loop filter to dynamically update the tracking frequency and the phase-locked phase to form the phase-locked reference.

3. The method of claim 1, wherein, Based on the discrete shock response matrix and the phase-locked reference, the interference micro-pulse sequence is optimized and solved to minimize the key section pressure energy, comprising: Constructing an objective function to represent the weighted sum of the synthesized pressure energy generated at the key section after the interference micro-pulse sequence is convolved by the discrete shock response matrix; According to the instantaneous phase of the phase-locked reference, higher weights are applied to the phase interval corresponding to the expected pressure peak in the weighted sum; Solving the objective function under the device constraints of valve stroke, speed and jerk to obtain the interference micro-pulse sequence.

4. The method of claim 1, wherein, According to the phase-locked reference, the overlap time window is determined, and in the overlap time window, the slow opening template is generated, the interference micro-pulse sequence is superimposed on the slow opening template, and the trajectory shaping constraint is applied to generate the vibration suppression opening template, comprising: According to the instantaneous phase in the phase-locked reference, the criterion of opening before trough and closing after peak is applied to determine the overlap time window of valve switching; In the overlap time window, the slow opening template is generated, and the interference micro-pulse sequence is superimposed on the slow opening template; Applying S-shaped shaping to the superimposed trajectory to make it meet the preset jerk and speed constraints to form the vibration suppression opening template.

5. The method of claim 1, wherein, applying a micro-excitation within a safety injection time window, identifying a discrete impulse response matrix of the valve to a critical section according to pressure and gate position responses in the aligned data stream, including: superimposing a pseudo-random sequence or a micro-step to the valve as a micro-excitation within the safety injection time window; extracting pressure responses triggered by the micro-excitation in the aligned data stream, and corresponding valve gate position sequences; obtaining the discrete impulse response matrix online according to the valve gate position sequences and the pressure responses by using a recursive least squares algorithm.

6. The method of claim 1, wherein, outputting executable valve trajectories, including: assembling an MPC state-output model based on a digital twin state; defining a safety barrier function based on a mooring force probability distribution; setting a Wasserstein radius according to a quality assessment report, and constructing an uncertainty distribution set with an empirical distribution of the mooring force probability distribution as a center; assembling a hard constraint set containing a vibration suppression opening template tracking constraint, and constructing an objective function; transforming a probabilistic constraint of the safety barrier function under the uncertainty distribution set into a deterministic robust barrier constraint; solving an optimization problem that minimizes the objective function under the MPC state-output model, the hard constraint set and the deterministic robust barrier constraint in a rolling horizon, and obtaining executable valve trajectories.

7. The method of claim 6, wherein setting the Wasserstein radius according to the quality assessment report includes adaptively increasing the Wasserstein radius when the quality assessment report indicates that a sensor channel confidence decreases or a residual increases; transforming the probabilistic constraint of the safety barrier function under the uncertainty distribution set into the deterministic robust barrier constraint includes using a second-order cone programming equivalent form, a moment bound or a semi-definite relaxation method to reconstruct a worst-case probabilistic constraint under the uncertainty distribution set into a deterministic constraint that does not depend on a specific probability distribution.

8. The method of claim 6, wherein, constructing the objective function containing a water hammer index, a flow smoothness and a water saving cost, including: defining a time domain fluctuation variance or a frequency domain main mode energy of the pressure sequence in the aligned data stream as the water hammer index; defining a total flow change amount predicted by the MPC state-output model at adjacent time points as the flow smoothness; defining a water cost mapped based on a working condition parameter in the digital twin state as the water saving cost; constructing the objective function as a preset weighted sum of the water hammer index, the flow smoothness and the water saving cost.

9. A multi-stage water-saving ship lock navigation safety guarantee system based on digital twinning, characterized in that, including: a data aggregation unit for aggregating multi-source sensor data, generating an aligned data stream and a quality assessment report, and updating a digital twin state; a probabilistic observation unit for obtaining a mooring force probability distribution by non-contact observation and probabilistic fusion using the aligned data stream and the quality assessment report; a vibration suppression template generation unit for identifying a hydraulic impulse response and generating a vibration suppression opening template by hydraulic lock-in tracking based on the digital twin state and the aligned data stream; a distribution robust control unit for executing a distribution robust stochastic control barrier optimization in a digital twin system by integrating the mooring force probability distribution and the vibration suppression opening template, and outputting executable valve trajectories; The multi-source sensing data includes: gate chamber water level sequence, pressure before and after the valve, valve door position, visual and inertial measurement data of ship position and attitude, and historical operation data; Based on the digital twin state and the aligned data stream, a hydraulic shock response is identified and a hydraulic phase-locked loop is tracked to generate a vibration suppression opening template, including: A micro-excitation is applied within a safe injection time window, and a discrete shock response matrix of the valve to the key section is identified according to the pressure and door position response in the aligned data stream; A pressure sequence in the aligned data stream is extracted, a hydraulic phase-locked loop is dynamically tracked to track the frequency and instantaneous phase of the water hammer dominant mode, and a phase-locked reference is formed; Based on the discrete shock response matrix and the phase-locked reference, a sequence of interference micro-pulses is optimized and solved to minimize the key section pressure energy; According to the phase-locked reference, an overlap time window is determined, a slow opening template is generated in the overlap time window, and the vibration suppression opening template is generated by fusing the sequence of interference micro-pulses, applying trajectory shaping constraints.

Images