CN-122022379-A - Marine fishery production element management and control method based on digital twin architecture

CN122022379ACN 122022379 ACN122022379 ACN 122022379ACN-122022379-A

Abstract

The invention discloses a marine fishery production element management and control method based on a digital twin architecture, and belongs to the technical field of intelligent ocean engineering. The method comprises the steps of realizing stereoscopic perception of sea ice, wave flow and biological characteristics by deploying an anti-interference multi-source sensor array, constructing an air-space-sea integrated heterogeneous network, realizing multi-mode data fusion and trusted storage by using a flow batch integrated engine and a blockchain technology, integrating ice-wave-flow-structure multi-field coupling simulation capability and an aquaculture vertical large model, constructing a high-fidelity digital twin body, carrying out disaster deduction and growth prediction, and generating a collaborative operation instruction of an heterogeneous equipment group based on multi-agent reinforcement learning, thereby realizing closed-loop accurate management and control. The invention obviously improves the disaster resistance and the intelligent operation efficiency of the cultivation facilities.

Inventors

SHI YANJUN
LI YUKUN
ZHAO YUNPENG

Assignees

大连理工大学

Dates

Publication Date: 20260512
Application Date: 20260317

Claims (10)

1. The marine fishery production element management and control method based on the digital twin architecture is characterized by comprising the following steps of: acquiring multi-source heterogeneous fishery production data according to a three-dimensional perception network deployed in a culture facility and a surrounding sea area thereof; according to the fishery production data, carrying out transmission and edge calculation through a heterogeneous fusion network to obtain data after cleaning and feature extraction; according to the data after cleaning and feature extraction, fusing and storing the data by using a flow batch integrated engine and a blockchain technology, and constructing a culture full-factor dynamic knowledge graph and a trusted data chain; Constructing a digital twin body comprising a multi-field coupling simulation and a vertical large model according to the dynamic knowledge graph and the trusted data chain, and carrying out full life cycle deduction and prediction on the breeding objects and the environment; generating a collaborative operation instruction aiming at the heterogeneous intelligent equipment group through a multi-agent reinforcement learning algorithm according to the deduction and prediction result of the digital twin body; and performing closed loop control on a physical execution terminal according to the verified collaborative operation instruction, and feeding back an execution effect in real time to correct the digital twin body.
2. The method of claim 1, wherein the collecting multi-source heterogeneous fishery production data comprises: According to an X-band radar and machine vision fusion unit deployed on a cultivation facility, texture features and motion tracks of sea ice are obtained; Acquiring structural ice load distribution and ice-induced vibration frequency according to a film pressure sensor array and a fiber bragg grating strain sensor which are arranged on an upright post of a cultivation facility; Acquiring fish swarm behavior and biomass data in a high-turbidity water environment according to an optical camera and a double-frequency identification sonar which are deployed in a culture facility; and acquiring a dynamic response mode of the structure under storm surge according to the triaxial accelerometer and the inclination sensor which are arranged at key nodes of the cultivation facility.
3. The method of claim 1, wherein the transmitting and edge computing over the heterogeneous converged network comprises: According to the intelligent SD-WAN gateway deployed on the main control platform, monitoring the link quality of a plurality of communication links in real time, and dynamically scheduling transmission paths or aggregation links according to service priorities; an underwater autonomous vehicle is used as a mobile relay node, and automatically floats to the water surface when the thermocline shielding causes the blocking of underwater acoustic communication, and relays data through radio; And automatically executing preset emergency control logic according to the edge intelligent gateway with the local computing power when communication is interrupted, and carrying out local caching and breakpoint continuous transmission on the data.
4. The method of claim 1, wherein constructing a digital twin comprising a multi-field coupled simulation and a vertical large model comprises: Based on real-time perceived environment and structure data, simulating a sea ice dynamics process by using a discrete element method, solving a wave field by combining computational fluid dynamics, and performing real-time data interaction with a finite element structure model through a bidirectional fluid-solid coupling interface to obtain a multi-field coupling simulation result of the structure under the combined action of ice-wave-flow-structure; According to the multi-field coupling simulation result and unstructured culture data, a culture suggestion and risk early warning in a natural language form are obtained through a vertical large model optimized for aquaculture; according to the real-time environmental data and the fish physiological indexes, a dynamic prediction result of the growth and health state of the fish is obtained through a dynamic energy budget theoretical model.
5. The method of claim 1, wherein the generating a collaborative job instruction for a heterogeneous intelligent equipment group comprises: According to the fish shoal distribution and biomass detected in real time, cooperatively scheduling a fish suction pump, a classifying screen and a transport ship through a multi-agent reinforcement learning algorithm, and generating a cooperative operation instruction for optimizing start-stop time sequence and operation parameters; Generating a compensation control instruction through an active heave compensation algorithm according to the hull heave displacement measured in real time by the motion reference unit, and driving a servo valve or a servo motor to adjust an actuating mechanism according to the compensation control instruction so as to offset the relative motion caused by waves; And generating a cooperative control instruction for synchronizing the inspection path and the data relay through a distributed consistency protocol according to the cluster cooperation requirement of the underwater inspection robot and the unmanned ship on the water surface.
6. The method of claim 1, wherein the closed loop control of the physical execution terminal comprises: according to the digital twin bodies in the facility construction stage, virtual previewing is carried out on the offshore installation process under different sea conditions, risk points are identified, and a construction scheme is optimized; And predicting the residual service life of the key parts through the Weibull distribution and long-short time memory network according to the equipment operation data and the environmental load history, and automatically generating a spare part purchasing plan and a maintenance work order.
7. The method of claim 2, wherein the collecting multi-source heterogeneous fishery production data further comprises: And carrying out real-time enhancement processing on the underwater video stream under the high-turbidity water body according to the prior of the dark channel and a turbidity removing algorithm for generating the fusion of the opposite network.
8. The method of claim 4, wherein constructing a digital twin comprising a multi-field coupled simulation and a vertical large model further comprises: and according to the multi-field coupling simulation result, analyzing and outputting a decision suggestion in a natural language form by the vertical large model.
9. The method of claim 5, wherein generating the compensation control command comprises operating on a heave displacement signal fed back by the motion reference unit via a PID controller to generate the compensation control command.
10. The method of claim 1, wherein the constructing a farmed full element dynamic knowledge-graph and trusted data chain comprises: And according to the extracted digital fingerprints of the key operation events of feeding, medication, capturing and maintaining, performing uplink storage through a blockchain technology to form a tamper-proof trust anchor point.

Description

Marine fishery production element management and control method based on digital twin architecture Technical Field The invention belongs to the technical field of intelligent ocean engineering, and particularly relates to a digital twin architecture-based ocean fishery production element management and control method. Background With the increasing shortage of global marine fishery resources, mariculture is accelerating the expansion from near-shore estuaries to deep open seas. The natural environment of the Bohai sea area is significant in specificity and challenges as an important aquaculture base in the north of China, and extremely high requirements are provided for aquaculture equipment and management and control technology. Prior art limited depth analysis: 1. lack of environmental adaptability and perceptibility: The existing intelligent cultivation system is designed for a bay or a low sea condition area, and the extreme environment of the yellow Bohai sea is not fully considered for sensor selection and deployment. Sea ice threatens that serious sea ice is frequently generated in the north of the Bohai sea in winter, and huge load (ice pressure) generated by impact and accumulation of the flowing ice is extremely easy to cause deformation and even damage of the net cage structure. The existing system lacks a special monitoring module aiming at sea ice thickness, drift speed and structural ice load, and can not provide effective ice disaster early warning. Extreme stormy waves, the sea area is often influenced by storm surge and typhoon in a warm zone, the wind speed of the extreme value in one hundred years can reach 37m/s, and the effective wave height can reach 6m. Existing monitoring systems often fail under severe sea conditions and lack real-time safety assessment for structural stress, mooring tension, resulting in "blind images". High turbidity water, namely high suspended matters in water at the offshore and estuary areas of Bohai sea and high turbidity. The traditional optical camera images and blurs in turbid water bodies, and a fish behavior analysis and disease identification algorithm based on common Computer Vision (CV) is basically invalid. 2. Equipment collaborative operation level is low and 'island effect': although single automatic equipment such as automatic bait casting machines, fish sucking pumps, net washing robots and the like exists at present, the single automatic equipment usually operates as independent systems, and lacks of unified control protocols and cooperative logic, so that the phenomenon of 'business function island' is serious. And (3) capturing-transferring and uncoupling, namely, when fishing is recovered, the power adjustment of the fish suction pump is in linkage with the berthing speed of the transferring ship and the speed of the conveying belt is lacked. Often need manual operation respectively, can't be adjusted in real time according to fish current density, very easily cause fish body mechanical damage (abrasion, extrusion), and the operating efficiency is low. The feeding strategy is rigidified, namely the feeding machine usually operates according to a preset time schedule and cannot carry out closed-loop linkage with real-time water quality (such as dissolved oxygen dip) or shoal feeding feedback (such as judgment by sonar), so that feed waste or water quality pollution is caused. 3. Digital twinning application depth is not enough: The digital twin is declared to stay at the three-dimensional visual display level, namely, a game engine is utilized to render a realistic scene, but the kernel is lacked. The mechanism model is missing, and a depth calculation model based on hydrodynamics, structural mechanics and bioenergy is absent. The system cannot perform virtual trial and error (such as simulating the performance of different anchoring schemes in a twin body under 14-stage typhoons), and cannot predict the long-term influence of environmental mutation on fish growth. Full life cycle fault, namely building data (BIM) of facilities are often discarded after delivery and cannot be continued to an operation and maintenance stage, so that operation and maintenance personnel lack of structural original data support, and accurate life prediction and maintenance are difficult to perform. Disclosure of Invention In order to solve the technical problems, the invention provides a marine fishery production element management and control method based on a digital twin architecture, which comprises the following steps: acquiring multi-source heterogeneous fishery production data according to a three-dimensional perception network deployed in a culture facility and a surrounding sea area thereof; according to the fishery production data, carrying out transmission and edge calculation through a heterogeneous fusion network to obtain data after cleaning and feature extraction; according to the data after cleaning and feature extraction, fusing and storing the da