Marine culture multi-parameter online detection system and method based on edge computing

Abstract

The application discloses a kind of based on edge computing's marine culture multi-parameter on-line detection system and method, belong to marine culture data processing technical field.The device includes adaptive hierarchical synchronous acquisition device including multiple collection cavities, negative pressure shunt transmission pipeline, water quality detection warehouse equipment and multi-source data acquisition equipment;Multiple collection cavities include first cavity, second cavity and third cavity;Negative pressure shunt transmission pipeline, water quality detection warehouse equipment detect water quality and obtain multi-cavity water quality data;Multi-source data acquisition equipment collects net cage deformation value and fish population density;Edge computing equipment deduces water evolution information;Remote server stores water evolution information, and generates marine culture suggestion.The application is coupled by collecting water sample in different water layers of culture net cage, and fish population, net cage deformation disturbance deduces water quality evolution information, realizes the accurate prediction and local closed loop disposal of culture water risk, improves the reliability and early warning accuracy of deep sea three-dimensional culture monitoring.

Description

Technical Field

[0001] This invention relates to the field of marine aquaculture data processing technology, and more specifically to a multi-parameter online detection system and method for marine aquaculture based on edge computing. Background Technology

[0002] Currently, three-dimensional marine aquaculture relies on the vertical space of the water body to carry out stratified farming. The water quality indicators, such as dissolved oxygen, water temperature, and ammonia nitrogen, vary significantly across different water layers, and the farmed organisms exhibit a vertical distribution. This places extremely high demands on in-situ stratified monitoring and rapid closed-loop response to emergencies. Current marine three-dimensional environmental monitoring mainly falls into two mainstream technical approaches, both of which have inherent shortcomings, and the industry has yet to develop a complete solution adapted to enclosed aquaculture cages.

[0003] The first category is large-scale macroscopic remote sensing monitoring of marine areas, which relies on aerial observation platforms to collect water surface spectral information and indirectly invert underwater environmental parameters through mathematical interpolation. This mode can only output average estimated data at the regional scale, with a macroscopic observation granularity. It cannot penetrate into the closed aquaculture cages to obtain in-situ real-time water quality data of the aquaculture organisms' habitat, nor can it simultaneously collect aquaculture-specific information such as fish school behavior and cage structural deformation. The data update delay is relatively long, making it impossible to link oxygenation and feeding equipment on-site for emergency control. It is naturally unsuitable for the urgent needs of localized water layer hypoxia and immediate treatment in three-dimensional aquaculture, and it does not offer any inspiration for combining with the technical route of in-situ monitoring and edge closed-loop control.

[0004] The second category is the application of scattered, standalone monitoring equipment: Although the industry has independent hardware functions such as layered sensor unit deployment, edge device data processing, dual-link marine communication, and probe ultrasonic cleaning, these devices are used in isolation without customized operating rules for the integrated aquaculture scenario. This results in multiple structural defects: First, sensor units are mostly arranged uniformly vertically, without a depth-differentiated deployment strategy based on the habitat patterns of aquaculture organisms. There is also a lack of layered frequency conversion sampling and staggered cleaning mechanisms, making it prone to missed detections in the core habitat layer and gaps in vertical gradient data caused by multi-layer synchronous cleaning. Second, edge computing power is concentrated in a single gateway, with underwater acquisition nodes simultaneously loading intelligent computing tasks. Under conditions of high humidity and limited power supply in seawater, device power consumption is high, and the latency for hazard inference generally exceeds 30ms. Third: Early warning generally adopts single-parameter threshold judgment logic, and three types of heterogeneous data, water quality, fish behavior and cage stress, are alarmed separately and independently, often resulting in false alarms when dissolved oxygen fluctuates slightly but the fish are in normal condition; Fourth: Dual-link communication only realizes simple switching between different signal channels, and does not allocate transmission resources according to the severity of the danger, and the layered monitoring data is easily permanently lost after the offshore network is disconnected.

[0005] Therefore, how to conduct systematic monitoring of three-dimensional aquaculture scenarios is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of the above problems, the present invention is proposed to provide an edge computing-based online multi-parameter detection system and method for marine aquaculture that overcomes or at least partially solves the above problems.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] In a first aspect, embodiments of the present invention provide a multi-parameter online monitoring system for marine aquaculture based on edge computing, comprising: An adaptive hierarchical synchronous acquisition device is used to collect multi-source heterogeneous sensing parameters within aquaculture cages, including multiple acquisition chambers, negative pressure diversion transmission pipelines, water quality testing chamber equipment, and multi-source data acquisition equipment. Wherein: the multi-collection cavity includes a first cavity disposed on the surface of the aquaculture water layer, a second cavity disposed on the middle layer of the aquaculture water layer, and a third cavity disposed on the bottom layer of the aquaculture water layer; The negative pressure diversion transmission pipeline is connected at one end to the bottom of the multi-collection chamber and is used to extract water samples from the multi-collection chamber. The water quality testing chamber is connected to the other end of the negative pressure diversion transmission pipeline and is used to test the water sample to obtain multi-chamber water quality data. The multi-source data acquisition device is set at a preset position in the multi-acquisition cavity to simultaneously acquire the deformation value of the aquaculture net cage and the fish density in the second cavity; An edge computing device is electrically connected to the multi-source data acquisition device and the water quality testing chamber device, respectively, and is configured to infer water body evolution information based on the multi-cavity water quality data, the fish density, the cage deformation value and the water body evolution risk prediction model. The water body evolution information is used to characterize the changes in water quality in the vertical spatial dimension under the joint disturbance of fish population and cage deformation. A remote server, which is communicatively connected to the edge computing device, is used to store the water body evolution information and generate marine aquaculture recommendations based on the water body evolution information.

[0009] In one embodiment, the adaptive hierarchical synchronous acquisition device further includes a differential pressure adaptive linkage lifting mechanism composed of two static pressure bellows, wherein: The first static pressure bellows connects the first cavity and the second cavity; The second static pressure bellows connects the second cavity and the third cavity.

[0010] In one embodiment, the multi-source data acquisition device specifically includes: A sonar probe is placed on either side of the second cavity to obtain the fish density inside the second cavity by transmitting and receiving the returned sound waves; A cage stress sensor is installed at the bottom of the third cavity to collect the deformation value of the aquaculture cage.

[0011] In one embodiment, the detection system further includes: a plurality of distributed passive differential pressure preprocessing devices connected to the edge computing device; The distributed passive differential pressure preprocessing device is located above the cavity and is configured to preprocess the multi-cavity water quality data, the fish density, and the cage deformation value to generate a multi-source heterogeneous feature dataset. The cavity is the first cavity, the second cavity, or the third cavity. The edge computing device infers water body evolution information based on the multi-source heterogeneous feature dataset and the water body evolution risk prediction model.

[0012] In one embodiment, the distributed passive differential pressure pretreatment device specifically includes: The vertical water quality pretreatment module is used to obtain multiple vertical water quality gradient features through the multi-cavity water quality data, dissolved oxygen concentration conversion function and ammonia nitrogen concentration conversion function. The vertical water quality gradient features include the dissolved oxygen concentration difference between two adjacent aquaculture water layers and the ammonia nitrogen concentration difference between two adjacent aquaculture water layers. The fish density preprocessing module is used to normalize the fish density to obtain a normalized fish density, wherein the normalized fish density is:

[0013] in, The normalized fish density To output real-time echo voltage values ​​for underwater sonar probes. The lower limit voltage for normalizing fish density. The upper limit voltage for normalizing fish density; The wire mesh cage deformation preprocessing module is used to normalize the wire mesh cage deformation value to obtain a normalized wire mesh cage deformation value, wherein the normalized wire mesh cage deformation value is:

[0014] in, Normalized mesh cage deformation value To provide real-time voltage signal output for the stress sensor of the wire mesh cage. This represents the minimum voltage reference value output by the stress sensor in the undeformed state of the wire mesh cage. The maximum voltage reference value output by the stress sensor when the cage reaches the maximum allowable deformation; The feature dataset construction module is used to construct the multi-source heterogeneous feature dataset based on the multiple water quality vertical gradient features, the normalized fish density, and the normalized cage deformation value.

[0015] In one embodiment, the edge computing device specifically includes: The MCU unit has a built-in water body evolution risk prediction model and a water body evolution risk degree determination module. The water body evolution risk prediction model is used to deduce the water body evolution risk value within a preset time period through differential iteration. The water body evolution risk degree determination module is used to determine the water body evolution risk degree based on the water body evolution risk value and the corresponding relationship, wherein the corresponding relationship is the correspondence between the water body evolution risk value and the water body evolution risk degree. The communication module is used to push push parameters corresponding to the water body evolution risk level; A hierarchical cache storage module is used to store the water body evolution risk value when the communication module is offline; The control module is used to generate control signals for the underwater aeration equipment and control signals for the equipment's output rate based on the degree of risk of water body evolution.

[0016] In one embodiment, the risk level includes at least mild evolution risk and severe evolution risk, and the control module specifically includes: an oxygenation equipment control unit and a discharge equipment control unit; The oxygenation equipment control unit is connected to the underwater oxygenation equipment control unit. When the water body evolution risk level is the severe evolution risk, a control signal for turning on the underwater oxygenation equipment is generated. The control unit of the feeding device is connected to the control unit of the fish bait feeding device. When the water body evolution risk level is the mild evolution risk, a control signal is generated to reduce the feeding amount of the fish bait feeding device.

[0017] Secondly, embodiments of the present invention provide an online detection method for multiple parameters in marine aquaculture based on edge computing, applied to an online detection system for multiple parameters in marine aquaculture based on edge computing as described in the first aspect, comprising: The adaptive hierarchical synchronous acquisition device collects water samples from multiple acquisition chambers through a negative pressure diversion transmission pipeline and obtains multi-chamber water quality data by detecting the water samples through a water quality testing chamber; it also collects the deformation value of the aquaculture net cage and the fish density of the second chamber in the multi-acquisition chamber through a multi-source data acquisition device. The edge computing device infers water body evolution information based on the multi-cavity water quality data, the fish density, the cage deformation value, and the water body evolution risk prediction model. The water body evolution information is used to characterize the changes in water quality in the vertical spatial dimension under the combined disturbance of fish population and cage deformation. The remote server stores the water body evolution information and generates marine aquaculture recommendations based on the water body evolution information.

[0018] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following: An adaptive hierarchical synchronous acquisition module collects water samples from the surface, middle, and bottom layers of aquaculture cages, while simultaneously collecting data on fish activity and cage deformation. This allows for the quantification of water disturbance effects through fish activity and cage deformation. A risk differential transmission linkage module infers water evolution information based on multi-cavity water quality data, fish density, and cage deformation values ​​to accurately predict water evolution and achieve local edge closed-loop management of potential hazards. Simultaneously, a data archiving module can store water evolution information in real time and generate marine aquaculture recommendations to optimize aquaculture strategies. Attached Figure Description

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

[0020] Figure 1 This is a schematic diagram of the structure of the multi-acquisition cavity provided in an embodiment of the present invention; Figure 2 This invention provides a distributed passive differential pressure pretreatment device. Figure 3 A schematic diagram illustrating the principle of online multi-parameter detection in marine aquaculture based on edge computing, provided in an embodiment of the present invention. Figure 4 The flowchart illustrates the online detection method for multiple parameters in marine aquaculture based on edge computing, as provided in this embodiment of the invention. Detailed Implementation

[0021] 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] Example 1: This invention discloses an online multi-parameter monitoring system for marine aquaculture based on edge computing. It can sense water samples from different water layers in aquaculture cages and couple them with fish density and cage deformation values. It can also extrapolate water evolution information through a water body evolution risk prediction model to achieve accurate prediction of water body evolution information.

[0023] like Figure 1 As shown, the system includes an adaptive hierarchical synchronous acquisition device, an edge computing device, and a remote server.

[0024] in: (a) Adaptive hierarchical synchronous acquisition device Specifically, the adaptive layered synchronous acquisition module consists of multiple acquisition chambers, negative pressure diversion transmission pipelines, water quality testing chamber equipment, and multi-source data acquisition equipment. It relies entirely on the natural power of seawater to complete the acquisition and self-cleaning actions. There are no motors or ultrasonic power supply modules underwater, and it abandons the traditional electric lifting support, layered independent water quality probes, and ultrasonic electronically controlled descaling structure.

[0025] The adaptive stratified synchronous acquisition device is located on the outside of the aquaculture cage. It is tied to the outer ring of the upper frame of the cage with a flexible mooring cable, floats with the waves, and monitors the vertical stratified water environment outside the cage. Combined with the deformation sensor data of the cage itself, it calculates the risk of the aquaculture environment inside the cage without the need for a probe to be inserted into the cage.

[0026] (1) Multiple acquisition chambers Specifically, such as Figure 1 As shown, from top to bottom, there are three chambers: the first chamber located on the surface of the aquaculture water layer, the second chamber located in the middle layer of the aquaculture water layer, and the third chamber located at the bottom of the aquaculture water layer.

[0027] The surface layer can be a water layer with a depth of 0.5m, the middle layer can be a water layer with a depth of 8m or 10m, and the bottom layer is a water layer 2m away from the bottom of the aquaculture cage.

[0028] Each cavity has symmetrical flow inlets on both the left and right sides, and passive rotating turbulence blades are installed inside the cavity. When the seawater inside the aquaculture cage flows, the water flows into the cavity from the flow inlets on both sides, driving the passive rotating turbulence blades to rotate continuously and form turbulence. This continuously washes the surface of the sensing membrane and automatically peels off the attached algae and shellfish. The entire process requires no external power supply or preset timed cleaning program, achieving long-lasting passive scale inhibition.

[0029] Understandably, the first, second, and third chambers are independent of each other and are watertight, with no water passages between the three chambers, so seawater samples will not flow between the chambers.

[0030] (2) Differential pressure adaptive linkage lifting mechanism Specifically, it includes two sealed static pressure bellows. The first and second chambers are connected by the sealed static pressure bellows, and the second and third chambers are also connected by the sealed static pressure bellows. The static pressure differential linkage bellows is connected between the three acquisition chambers and is only used to transmit the water pressure difference signal between each water layer, without transporting water.

[0031] The sealed static pressure corrugated pipe is connected to the internally filled atmospheric pressure seawater. Since the middle layer of the aquaculture water layer is the main habitat of fish, the water pressure will fluctuate in real time with the fish's grouping activities and water depth changes. The pressure changes will squeeze or stretch the corrugated pipe, which will simultaneously drive the corresponding sampling chambers of the surface and bottom layers to automatically fine-tune the water depth. It can adaptively match the habitat water depth of the aquaculture organisms without the need for motor drive, reducing the blind spots of stratified monitoring.

[0032] (3) Negative pressure diversion transmission pipeline Specifically, the bottoms of the first, second, and third chambers are connected to the same set of negative pressure diversion pipelines. Electrically controlled time-sharing diversion valves are installed inside the pipelines to extract water samples in a fixed order of surface, middle, and bottom layers and collect them into the public water quality testing chamber.

[0033] (4) Water quality testing chamber equipment The water quality testing chamber equipment is equipped with only one set of multi-parameter water quality sensor cores. That is, a single set of cores can complete the detection of multiple parameters (i.e., multi-cavity water quality data) of three layers of water. Compared with related technologies, the embodiments of the present invention can significantly reduce equipment procurement costs by eliminating the need for independent sensing hardware for each layer.

[0034] It is understood that the adaptive stratified integrated module in this embodiment of the invention includes three unique mechanical structures: a multi-collection chamber with self-cleaning function, a differential pressure adaptive linkage lifting mechanism, and a three-layer negative pressure diversion pipeline; the entire structure relies on seawater flow and static pressure to complete self-cleaning and stratified collection actions, abandoning the industry-standard motor lifting, ultrasonic electronic descaling, and independent water quality probe solutions for each layer, thereby reducing underwater power consumption and maintenance frequency from the hardware level. Existing stratified collection brackets do not have the same structural design concept.

[0035] (5) Multi-source data acquisition equipment Specifically, the multi-source data acquisition equipment includes sonar and bottom cage stress sensors. In this embodiment of the invention, there are a total of 8 cage stress sensors. Four directional measuring points are set at 90° intervals along the outer circumference of the aquaculture cage netting. At each directional measuring point, one stress sensor is arranged at the upper netting top line and the bottom netting bottom line. The upper stress sensor is placed at a water depth of 1.0-2.0m to sense the netting tension caused by the impact of surface water flow; the lower stress sensor is placed at the bottom netting bottom line to sense the bottom settlement and compression contraction of the cage. The sensors are clamped and fixed to the outside of the netting load-bearing cable using anti-corrosion clamps to avoid damaging the netting structure. The sensor body faces the outer water body to avoid collision with the farmed fish. Eight stress sensors collect tension deformation signals from different orientations and water depths. The underwater signal cables are connected to the distributed passive differential pressure preprocessing unit inside the adaptive stratified synchronous acquisition platform. The edge end performs interpolation fitting on the data from four upper and four lower measuring points to obtain the deformation profiles of the upper and lower parts of the cage. The overall deformation characteristics of the cage are further calculated and used as the deformation correction input for the stratified water risk spatiotemporal coupling differential prediction model.

[0036] In this embodiment of the invention, the underwater sonar probe and the stress sensor are both independently deployed on the side of the cavity. The underwater sonar probe is used to collect fish swarm density data in the middle layer of the second cavity in real time, and the stress sensor records the deformation value of the net cage. The two types of data are independently transmitted to the host at the edge of the sea surface via wire.

[0037] (ii) Distributed passive differential pressure pretreatment equipment like Figure 2 As shown, the distributed passive differential pressure pretreatment device includes a vertical water quality pretreatment module, a fish density pretreatment module, a net cage deformation pretreatment module, and a feature dataset construction module. In this embodiment of the invention, the distributed passive differential pressure pretreatment unit can be a simple differential pressure signal conversion element, respectively arranged in the first cavity, the second cavity, and the third cavity.

[0038] After the water flow and water pressure drive data acquisition is completed, the distributed passive differential pressure preprocessing unit extracts only three types of simplified feature data packets: the difference in water quality parameters between the three layers, fish density, and cage deformation. The massive amount of raw water samples and image data are directly discarded underwater, and only the feature data is uploaded to the surface host, reducing the amount of communication data transmitted at sea from the source.

[0039] Specifically, the vertical water quality pretreatment module converts the analog voltage signals of multi-chamber water quality data into dissolved oxygen and ammonia nitrogen stratified concentrations using a hardware-based, fixed linear conversion formula. It then calculates the vertical gradient difference between water layers using differential hardware circuitry, generating two sets of vertical gradient characteristics for dissolved oxygen and ammonia nitrogen. The dissolved oxygen concentration conversion formula can be expressed as:

[0040] in, Let represent the dissolved oxygen concentration in the i-th layer, where i = 1, 2, and 3 correspond to the surface, middle, and bottom layers, respectively. Let a be the conversion factor between dissolved oxygen and voltage, a = 0.08 mg / (L·V). Let be the analog voltage signal of the i-th layer, and b be the offset value, b = -0.02 mg / L.

[0041] The formula for converting ammonia nitrogen concentration can be expressed as:

[0042] in, Let be the ammonia nitrogen concentration in the i-th layer, where i = 1, 2, and 3 correspond to the surface, middle, and bottom layers, respectively. Let c be the conversion factor between ammonia nitrogen and voltage, where c = 0.05 mg / (L·V). Let d be the analog voltage signal of the i-th layer, and d be the offset value, d = 0.001 mg / L.

[0043] Next, determine the dissolved oxygen concentration difference and ammonia nitrogen concentration difference between two adjacent aquaculture water layers. The dissolved oxygen concentration difference can be expressed as:

[0044]

[0045]

[0046]

[0047] in, This represents the difference in dissolved oxygen concentration between the surface and middle layers. This represents the difference in dissolved oxygen concentration between the middle and bottom layers. This represents the difference in ammonia nitrogen concentration between the surface and middle layers. This indicates the difference in ammonia nitrogen concentration between the middle and bottom layers.

[0048] The fish density preprocessing module normalizes the fish density using a normalization function. The normalized fish density is:

[0049] in, The normalized fish density To output real-time echo voltage values ​​for underwater sonar probes. The lower limit voltage for normalizing fish density. Normalized upper limit voltage for fish density =0.2V, =4.8V 。

[0050] The cage deformation preprocessing module normalizes the cage deformation values ​​using a normalization function. The normalized cage deformation values ​​are as follows:

[0051] in, Normalized mesh cage deformation value To provide real-time voltage signal output for the stress sensor of the wire mesh cage. This represents the minimum voltage reference value output by the stress sensor in the undeformed state of the wire mesh cage. =0.5V, The maximum voltage reference value output by the stress sensor when the cage reaches its maximum allowable deformation. =4.5V.

[0052] Among them, the normalized fish density and the normalized cage deformation value are the dimensionless cluster density and deformation characteristics in the interval [0, 1].

[0053] After the feature dataset construction module completes its processing, only three types of simplified feature data packages—vertical gradient, fish density, and net cage deformation—are packaged and uploaded to the sea surface differential computing host. All original simulated voltage and complete stratified concentration data are discarded on-site. The simplified feature set is as follows: .

[0054] (III) Detailed Implementation of Edge Computing Devices Specifically, the internal components include an MCU unit, which integrates a water body risk evolution module, a water body evolution risk level determination module, a communication module, a hierarchical cache storage module, and a control module. The MCU unit is equipped with a differential model computation chip. The transmission scheduling logic is dynamically adjusted entirely based on the water body evolution risk value output by the differential model, without employing fixed storage partitions or fixed transmission priorities.

[0055] (1) Water body risk evolution module Specifically, the water risk evolution module receives the feature dataset and assigns a value to the differential term of vertical water diffusion, which can be expressed as:

[0056] in, These are the vertical diffusion coefficients of surface-middle layer and middle-bottom layer water, respectively. =0.4, =0.6.

[0057] Then, by introducing correction terms for oxygen consumption due to fish density and water disturbance due to cage deformation, and using the fish density F collected by sonar as a correction coefficient, the differential equation for dissolved oxygen evolution is constructed as follows:

[0058] in, The oxygen consumption coefficient per unit fish population for the corresponding farmed species can be manually adjusted according to the stocking density. The normalized fish density The disturbance coefficient of the deformed water body is fixed at 0.09 mg / (L·h). This is the normalized deformation value of the wire mesh cage. This is a correction term for cage deformation disturbance.

[0059] The differential equation for the evolution of ammonia nitrogen is constructed as follows:

[0060] in, The upward diffusion coefficient of ammonia nitrogen in the substrate is given. , The difference in ammonia nitrogen concentration between the middle and bottom layers. The ammonia nitrogen release coefficient from fish excrement. , The coefficient for deformation of the substrate disturbance. The deformation of the cage will disturb the bottom silt and accelerate the upward diffusion of ammonia nitrogen.

[0061] Based on the differential equations for dissolved oxygen and ammonia nitrogen evolution, using the dissolved oxygen concentration, ammonia nitrogen concentration, fish density, and cage deformation values ​​collected at the current moment for each water layer as initial conditions, and performing 60 consecutive differential iterations with an iteration step size of 1 minute, the k-th iteration yields the predicted dissolved oxygen concentration C for the k-th minute in the future. DO (k) and the predicted ammonia nitrogen concentration N(k), where k = 1, 2, ..., 60. Based on the preset dissolved oxygen safety threshold, dissolved oxygen hazard threshold, ammonia nitrogen safety threshold, and ammonia nitrogen hazard threshold, C DO Normalization of N(k) and N(k) yields the dissolved oxygen risk sub-value V at minute k. DO (k) and ammonia nitrogen risk sub-value V N (k), and calculate the water body evolution risk value V(k) at minute k based on the water body evolution risk function, the expression of which is:

[0062] Among them, V DO (k) is the normalized dissolved oxygen risk sub-value at minute k, V N (k) is the normalized ammonia nitrogen risk sub-value at minute k. and The corresponding weighting coefficients are set to 0.6 and 0.4, respectively. A water body evolution risk curve for the next 60 minutes is constructed from V(1) to V(60), and Vmax = max{V(1), V(2), ..., V(60)} is taken as the final water body evolution risk value for this prediction cycle. Using the maximum risk value within the next 60 minutes can identify the most unfavorable moment and achieve advance risk prediction.

[0063] In this embodiment of the invention, the dissolved oxygen safety threshold is 6.0 mg / L, the warning threshold is 4.5 mg / L, and the danger threshold is 3.0 mg / L; the ammonia nitrogen safety threshold is 0.20 mg / L, the warning threshold is 0.35 mg / L, and the danger threshold is 0.50 mg / L.

[0064] It is understood that the embodiments of the present invention predict the risk of stratified water bodies through a spatiotemporally coupled differential prediction model, abandoning the traditional single-parameter threshold alarm and static linear weighted judgment method; by coupling the differential equation of vertical diffusion of water bodies with the oxygen consumption correction term of fish groups, the water quality evolution trend of the next hour is iteratively extrapolated locally at the edge, thereby achieving risk prediction in advance.

[0065] (2) Water body evolution risk zone module Specifically, when Vmax < 0.3, it indicates that the water condition is normal within the next 60 minutes, corresponding to no evolution risk; when 0.3 ≤ Vmax < 0.6, it indicates that there is slight water quality fluctuation within the next 60 minutes, corresponding to slight evolution risk; when Vmax ≥ 0.6, it indicates that severe hypoxia or bottom sediment pollution may occur within the next 60 minutes, corresponding to severe evolution risk.

[0066] (3) Communication module Specifically, the communication module includes a 4G / 5G main communication unit and a BeiDou short message backup unit. When the water body evolution risk value for the next hour is determined to be a severe evolution risk, 60% of the communication bandwidth is allocated to allow the 4G / 5G main communication subunit and the BeiDou short message backup subunit to simultaneously push the complete water quality evolution differential curve. When the water body evolution risk is mild, layered gradient simplified data is transmitted, and the original video images are compressed and stored locally. When there is no water body evolution risk, layered average data is uploaded once per hour. When there is no water body evolution risk, the water quality status remains stable. The edge computing device calculates the arithmetic mean of the minute-level water quality detection data of the surface, middle, and bottom layers over the past 60 minutes every hour to generate layered average data.

[0067] (4) Hierarchical cache storage module Specifically, when the 4G / 5G main communication subunit signal in the communication unit is interrupted, the hierarchical cache storage unit stores all differential iteration calculation parameters completely. When the 4G / 5G main communication subunit signal is restored, all stored differential iteration calculation parameters are transmitted to the intervals corresponding to the water body evolution risk, prioritizing the retransmission of high-risk period data. Among them, high-risk period data is the complete monitoring, simulation and disposal data corresponding to the period when the edge computing device locally determines that the water body evolution risk level reaches mild or above during the communication interruption. It is the highest priority data in the hierarchical retransmission strategy. The core purpose is to prioritize the transmission of core information of the dangerous period after the signal is restored, rather than transmitting ordinary steady-state data in chronological order, rather than simply transmitting data in the order of collection time.

[0068] (5) Control module Specifically, it includes the oxygenation equipment control unit and the discharge equipment control unit, which can directly output a response control signal within six seconds.

[0069] When the water body evolution risk is severe, the oxygenation equipment control unit generates a control signal to turn on the underwater oxygenation equipment; when the water body evolution risk is mild, the discharge equipment control unit generates a control signal to reduce the equipment's discharge rate, without requiring cloud-based relay commands throughout the process.

[0070] (iv) Detailed implementation of remote server Specifically, the remote server serves only as a long-term data storage and monthly aquaculture pattern review platform, and does not participate in differential deduction or equipment control. The data archiving module receives the differential evolution dataset uploaded daily by the risk differential transmission linkage module, statistically analyzes the water body change patterns throughout the aquaculture cycle based on year-round stratified water quality data, and only provides management personnel with medium- to long-term aquaculture optimization suggestions, without on-site emergency response authority.

[0071] The remote server serves only as a long-term data storage and monthly aquaculture pattern review platform. It does not participate in differential deduction or on-site equipment control. It receives uploaded differential evolution datasets, risk records, and hourly stratified mean data. Based on the annual stratified water quality data, it statistically analyzes the water body change patterns throughout the aquaculture cycle, including: seasonal vertical stratification patterns, such as monthly changes in dissolved oxygen and ammonia nitrogen gradients, and identification of thermocline cycle characteristics; diurnal rhythm patterns of water quality, such as intraday dissolved oxygen peaks and troughs, and diurnal release characteristics of ammonia nitrogen; nutrient accumulation patterns in the sediment, such as the correspondence between feeding load and bottom ammonia nitrogen accumulation rate; and the linkage between aquaculture operations and water quality, such as the impact of aeration, feeding, and cage maintenance on water quality gradients.

[0072] The aforementioned correspondence may include: the rate of increase of bottom ammonia nitrogen under different feeding loads; the higher the feeding load, the more uneaten feed and excrement settle to the bottom, and the faster the accumulation rate of bottom ammonia nitrogen, thereby assessing the degree of driving effect of feeding on bottom sediment pollution.

[0073] The correlation between aquaculture operations and water quality includes: statistically linking historical aeration, feeding, cage maintenance records with the vertical gradients of dissolved oxygen and ammonia nitrogen; feeding increases the vertical gradient of water quality; aeration can weaken the water quality gradient between layers; cage cleaning and net replacement disturb the bottom sediment, which will instantly increase the ammonia nitrogen gradient; the statistical results are used to generate medium- and long-term aquaculture optimization suggestions, and the remote server does not participate in on-site real-time simulation and equipment control.

[0074] The remote server analyzes the aforementioned water quality change patterns in conjunction with the tolerance thresholds of the aquaculture species. It then outputs medium- to long-term aquaculture optimization suggestions to management personnel through the management backend, without issuing on-site emergency control commands. On-site emergency response is handled locally in a closed loop by edge computing devices. Examples of medium- to long-term aquaculture optimization suggestions include: staggered aeration during periods of strong stratification; feeding recommendations to avoid periods of low dissolved oxygen and adjust feeding amounts based on the rate of sediment accumulation; developing cage cleaning and maintenance plans based on sediment pollution cycles; and appropriately adjusting stocking density during seasons of strong stratification to reduce the probability of severe risks (V≥0.6).

[0075] The following embodiment illustrates an online multi-parameter monitoring device for marine aquaculture based on edge computing according to the present invention: like Figure 2 The diagram shown is a schematic of the principle of online detection of multiple parameters in marine aquaculture based on edge computing provided in an embodiment of the present invention. The multi-collection chambers in the adaptive hierarchical synchronous acquisition device include a first chamber located at the surface, a second chamber located in the middle layer, and a third chamber located at the bottom layer. Water samples are then collected from the three chambers to obtain water quality data, and the deformation value of the net cage and the fish density in the second chamber are collected. The data are then preprocessed to achieve accurate detection of the three types of data and obtain a multi-source heterogeneous feature dataset, thereby effectively reducing computational complexity and consumption.

[0076] The real-time sonar echo voltage of the fish swarm in the middle layer was 2.5V, corresponding to the original fish density signal. The weighted fusion of stress sensors at the top and bottom of the cage yielded an equivalent voltage of 2.6V for the cage deformation. The original voltages for water quality detection in the three chambers were 2.1V for the surface layer, 1.6V for the middle layer, and 1.1V for the bottom layer. Using preset water quality calibration coefficients, the actual concentrations of the stratified water were calculated as follows: dissolved oxygen 6.16 mg / L in the surface layer, 4.96 mg / L in the middle layer, and 3.76 mg / L in the bottom layer; ammonia nitrogen 0.106 mg / L in the surface layer, 0.081 mg / L in the middle layer, and 0.056 mg / L in the bottom layer.

[0077] Further, normalization calculations were used to eliminate sensor range differences: the normalized fish density F≈0.50 was calculated based on the fish density normalization formula; the normalized net cage deformation value N was calculated based on the net cage deformation normalization formula. def =0.525. Meanwhile, based on the concentration of water quality in the three layers, the surface layer was calculated... Middle layer, middle layer The underlying dissolved oxygen gradient and ammonia nitrogen gradient are used to construct a complete multi-source heterogeneous feature dataset that includes stratified water quality concentration, vertical water quality gradient, normalized fish density, and normalized cage deformation characteristics.

[0078] The aforementioned multi-source heterogeneous feature dataset was input into the water body evolution risk prediction model. A 60-minute differential extrapolation was performed with an iteration step of 1 minute to obtain the dynamic evolution trend of dissolved oxygen and ammonia nitrogen in the water body over the next hour. Finally, the water body evolution risk value V within the preset time period was calculated. In this embodiment, under the actual measured conditions, the extrapolated water body evolution risk value V=0.34, which was determined to be a mild water body evolution risk. This fully realizes the quantitative assessment of water quality stratification evolution and the coupled evolution of aquaculture stress factors.

[0079] Then, the vertical evolution information of water bodies is inferred through the water body evolution risk prediction model and multi-source heterogeneous feature dataset to obtain the water body evolution risk value V within a preset time period. At the same time, the data archiving module stores the water body evolution risk and generates aquaculture suggestions.

[0080] The following comparative experiment illustrates the application effect of the online multi-parameter monitoring device for marine aquaculture based on edge computing of the present invention: A 90-day controlled experiment was conducted on a deep-sea large yellow croaker farming cage with a circumference of 80m and a water depth of 30m. The control group used industry-standard electric layered support, ultrasonic timed cleaning, independent layered probes, and a single-parameter threshold alarm scheme. The experimental group used the complete integrated hardware and software device of this invention. The actual measurements are as follows: First, the control group relied on monthly manual cleaning of probe deposits using ultrasonic sensors, and had multiple independent sensor cores underwater, resulting in frequent hardware replacements. In contrast, the multi-collection chambers in this invention eliminate the need for electrically controlled descaling, with one core completing three layers of water quality detection. The manual maintenance cycle for sensors is extended to 4.1 months, reducing underwater hardware procurement costs by 62% and offshore cage maintenance manpower and vessel costs by 78%. Furthermore, the entire underwater acquisition structure eliminates the need for lifting motors and ultrasonic power supply components, reducing the underwater mechanical failure rate by 85% and significantly lowering underwater equipment maintenance costs.

[0081] Second, the control group only triggered alarms after water quality parameters exceeded the standard, which is a post-event response, with an average alarm delay of more than 30 seconds; the differential prediction model in this embodiment of the invention can predict the evolution trend of low oxygen in the middle layer of water and the accumulation of ammonia nitrogen in the bottom layer 60 minutes in advance, reducing the false alarm rate of hypoxia to zero, and can start the control equipment in advance before the water quality deteriorates, thus avoiding mass fish deaths from the root cause.

[0082] Third, the single-parameter threshold alarm false alarm rate of the control group was 31.8%, and invalid alarms were pushed even with slight water fluctuations, increasing the workload of manual inspections; while the online detection device in this embodiment of the invention can combine water diffusion patterns and fish oxygen consumption behavior with differential deduction, resulting in a comprehensive early warning false alarm rate of only 2.8%, reducing the manpower input for invalid inspections by 43%, and significantly improving the accuracy of early warnings.

[0083] Fourth, the control group directly uploaded all original water samples and video data underwater, resulting in high data consumption, frequent BeiDou message pushes, and high costs; the online detection device in this invention only uploads simplified feature data packets of water quality gradient underwater, reducing overall communication traffic by 76%, and pushes data according to evolution risk classification, significantly reducing the frequency of BeiDou short message usage.

[0084] Fifth, when the online detection device in this embodiment of the invention is applied in the nearshore fish, shellfish and algae stratified aquaculture area, only the ammonia nitrogen diffusion coefficient of the bottom sediment in the differential equation needs to be adjusted to predict the risk of bottom sediment pollution in the bottom shellfish aquaculture area. The eutrophication trend of the water body can be warned 12 to 24 hours in advance. No large-scale diseases occurred during the entire aquaculture cycle, and the total output of aquatic products increased by 18% compared with the conventional monitoring plan in previous years.

[0085] Example 2: Based on the same inventive concept, this invention also provides an online detection method for multiple parameters in marine aquaculture based on edge computing. Since the principle of solving the problem by these methods is similar to that of the aforementioned online detection system for multiple parameters in marine aquaculture based on edge computing, the implementation of this method can refer to the implementation of the aforementioned system, and the repeated parts will not be described again.

[0086] like Figure 3 As shown, the method includes S1-S3: S1: The adaptive stratified synchronous acquisition device collects water samples from multiple acquisition chambers through a negative pressure diversion transmission pipeline and obtains multi-chamber water quality data by detecting the water samples through a water quality testing chamber; and collects the cage deformation value of the aquaculture cage and the fish density of the second chamber in the multi-acquisition chamber through a multi-source data acquisition device. S2: Edge computing devices infer water evolution information based on multi-cavity water quality data, fish density, cage deformation value, and water body evolution risk prediction model. The water body evolution information is used to characterize the changes in water quality in the vertical spatial dimension under the joint disturbance of fish population and cage deformation. S3: The remote server stores water body evolution information and generates marine aquaculture recommendations based on this information.

[0087] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0088] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A multi-parameter online monitoring system for marine aquaculture based on edge computing, characterized in that, include: An adaptive hierarchical synchronous acquisition device is used to collect multi-source heterogeneous sensing parameters within aquaculture cages, including multiple acquisition chambers, negative pressure diversion transmission pipelines, water quality testing chamber equipment, and multi-source data acquisition equipment. Wherein: the multi-collection cavity includes a first cavity disposed on the surface of the aquaculture water layer, a second cavity disposed on the middle layer of the aquaculture water layer, and a third cavity disposed on the bottom layer of the aquaculture water layer; The negative pressure diversion transmission pipeline is connected at one end to the bottom of the multi-collection chamber and is used to extract water samples from the multi-collection chamber. The water quality testing chamber is connected to the other end of the negative pressure diversion transmission pipeline and is used to test the water sample to obtain multi-chamber water quality data. The multi-source data acquisition device is set at a preset position in the multi-acquisition cavity to simultaneously acquire the deformation value of the aquaculture net cage and the fish density in the second cavity; An edge computing device is electrically connected to the multi-source data acquisition device and the water quality testing chamber device, respectively, and is configured to infer water body evolution information based on the multi-cavity water quality data, the fish density, the cage deformation value and the water body evolution risk prediction model. The water body evolution information is used to characterize the changes in water quality in the vertical spatial dimension under the joint disturbance of fish population and cage deformation. A remote server, which is communicatively connected to the edge computing device, is used to store the water body evolution information and generate marine aquaculture recommendations based on the water body evolution information.

2. The online multi-parameter monitoring system for marine aquaculture based on edge computing as described in claim 1, characterized in that, The adaptive hierarchical synchronous acquisition device further includes a differential pressure adaptive linkage lifting mechanism composed of two static pressure bellows, wherein: The first static pressure bellows connects the first cavity and the second cavity; The second static pressure bellows connects the second cavity and the third cavity.

3. The online multi-parameter monitoring system for marine aquaculture based on edge computing as described in claim 1, characterized in that, The multi-source data acquisition device specifically includes: A sonar probe is placed on either side of the second cavity to obtain the fish density inside the second cavity by transmitting and receiving the returned sound waves; A cage stress sensor is installed at the bottom of the third cavity to collect the deformation value of the aquaculture cage.

4. The online multi-parameter monitoring system for marine aquaculture based on edge computing as described in claim 1, characterized in that, The detection system further includes: multiple distributed passive differential pressure preprocessing devices connected to the edge computing device; The distributed passive differential pressure preprocessing device is located above the cavity and is configured to preprocess the multi-cavity water quality data, the fish density, and the cage deformation value to generate a multi-source heterogeneous feature dataset. The cavity is the first cavity, the second cavity, or the third cavity. The edge computing device infers water body evolution information based on the multi-source heterogeneous feature dataset and the water body evolution risk prediction model.

5. The online multi-parameter monitoring system for marine aquaculture based on edge computing as described in claim 4, characterized in that, The distributed passive differential pressure pretreatment device specifically includes: The vertical water quality pretreatment module is used to obtain multiple vertical water quality gradient features through the multi-cavity water quality data, dissolved oxygen concentration conversion function and ammonia nitrogen concentration conversion function. The vertical water quality gradient features include the dissolved oxygen concentration difference between two adjacent aquaculture water layers and the ammonia nitrogen concentration difference between two adjacent aquaculture water layers. The fish density preprocessing module is used to normalize the fish density to obtain a normalized fish density, wherein the normalized fish density is: in, The normalized fish density To output real-time echo voltage values ​​for underwater sonar probes. The lower limit voltage for normalizing fish density. The upper limit voltage for normalizing fish density; The wire mesh cage deformation preprocessing module is used to normalize the wire mesh cage deformation value to obtain a normalized wire mesh cage deformation value, wherein the normalized wire mesh cage deformation value is: in, Normalized mesh cage deformation value To provide real-time voltage signal output for the stress sensor of the wire mesh cage. This represents the minimum voltage reference value output by the stress sensor in the undeformed state of the wire mesh cage. The maximum voltage reference value output by the stress sensor when the cage reaches the maximum allowable deformation; The feature dataset construction module is used to construct the multi-source heterogeneous feature dataset based on the multiple water quality vertical gradient features, the normalized fish density, and the normalized cage deformation value.

6. The online multi-parameter monitoring system for marine aquaculture based on edge computing as described in claim 1, characterized in that, The edge computing device specifically includes: The MCU unit has a built-in water body evolution risk prediction model and a water body evolution risk degree determination module. The water body evolution risk prediction model is used to deduce the water body evolution risk value within a preset time period through differential iteration. The water body evolution risk degree determination module is used to determine the water body evolution risk degree based on the water body evolution risk value and the corresponding relationship, wherein the corresponding relationship is the correspondence between the water body evolution risk value and the water body evolution risk degree. The communication module is used to push push parameters corresponding to the water body evolution risk level based on the water body evolution risk level. A hierarchical cache storage module is used to store the water body evolution risk value when the communication module is offline; The control module is used to generate control signals for the underwater aeration equipment and control signals for the equipment's output rate based on the degree of risk of water body evolution.

7. The online multi-parameter monitoring system for marine aquaculture based on edge computing as described in claim 6, characterized in that, The risk level includes at least mild evolution risk and severe evolution risk. The control module specifically includes: an oxygenation equipment control unit and a discharge equipment control unit. The oxygenation equipment control unit is connected to the underwater oxygenation equipment control unit. When the water body evolution risk level is the severe evolution risk, a control signal for turning on the underwater oxygenation equipment is generated. The control unit of the feeding device is connected to the control unit of the fish bait feeding device. When the water body evolution risk level is the mild evolution risk, a control signal is generated to reduce the feeding amount of the fish bait feeding device.

8. A method for online detection of multiple parameters in marine aquaculture based on edge computing, applied to the online detection system for multiple parameters in marine aquaculture based on edge computing as described in any one of claims 1-7, characterized in that, include: The adaptive hierarchical synchronous acquisition device collects water samples from multiple acquisition chambers through a negative pressure diversion transmission pipeline and obtains multi-chamber water quality data by detecting the water samples through a water quality testing chamber. The deformation value of the aquaculture cage and the fish density of the second cavity in the multi-source data acquisition cavity are collected by a multi-source data acquisition device. The edge computing device infers water body evolution information based on the multi-cavity water quality data, the fish density, the cage deformation value, and the water body evolution risk prediction model. The water body evolution information is used to characterize the changes in water quality in the vertical spatial dimension under the combined disturbance of fish population and cage deformation. The remote server stores the water body evolution information and generates marine aquaculture recommendations based on the water body evolution information.

Images