Water temperature monitoring system for mariculture environment

Abstract

The invention relates to the technical field of marine environment monitoring, in particular to a marine aquaculture environment water temperature monitoring system which comprises a dynamic depth adjusting module, a sensor unit, a multi-parameter fusion analysis module, a passive heat shielding module, a heat compensation algorithm module and a central control unit. The passive heat shielding module is used for inhibiting parasitic heat conduction to an internal probe under a strong sunlight condition and comprises a heat flow sensing unit, a composite layer and a heat lag compensation subunit, the heat flow sensing unit collects shell surface temperature rise gradient data in real time, and the heat hysteresis compensation subunit transmits shell surface heat flow data to the thermal compensation algorithm module in real time. Through the synergistic effect of the dynamic depth adjustment module and the thermal compensation algorithm module, the problems of thermocline misjudgment and evaporative cooling pseudo signal interference caused by biological activities are solved, and accurate monitoring of the temperature of the aquaculture water layer and environment regulation and control are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of marine environment monitoring, and in particular to a water temperature monitoring system for a marine aquaculture environment. Background Art

[0002] The marine aquaculture industry relies heavily on accurate water temperature monitoring. Water temperature fluctuations directly impact fish metabolic rates, feed conversion efficiency, and disease incidence. In deepwater cage aquaculture, vertical stratification of water temperature can lead to uneven dissolved oxygen distribution. Fish's adaptability to specific water temperature layers is closely related to their growth stage. Therefore, real-time temperature data for different water layers is a key technical foundation for optimizing aquaculture management and preventing hypoxia and heat stress.

[0003] Existing technology, such as a marine water temperature monitoring device and monitoring method for marine environment monitoring proposed in publication number CN116907668A, includes a sedimentation box, a traction rope is provided in the sedimentation box, a detection box is provided on the side wall of the traction rope, a control component is provided in the sedimentation box, a temperature sensor is fixedly installed on the top surface of the detection box, a sleeve is fixedly connected to the inside of the detection box, and the traction rope passes through the sleeve; two lifting components are provided in the detection box, the top end of the traction rope is fixedly connected to an annular plate, and the top of the annular plate is fixedly connected to an annular airbag. It can achieve, effectively solve the problem that the existing marine water temperature monitoring range is small and it is inconvenient to monitor seawater at different depths.

[0004] While this invention addresses the limited ocean temperature monitoring range and enables monitoring at varying depths, it can easily overlook microscale temperature perturbations caused by the movement of densely populated aquacultured fish when sensing vertical ocean temperature gradients. This biological behavior can disrupt the natural thermocline structure. Existing technologies lack multi-source data fusion capabilities and are unable to distinguish between natural temperature gradients and bioturbation signals, potentially leading to misjudgments by the aeration system or missed detection of heat stress migration. Furthermore, while surface-floating sensors are waterproof, prolonged exposure to strong sunlight can cause parasitic heat conduction from infrared radiation absorbed by the outer shell, causing the measured value to deviate from the true water temperature. Furthermore, after rainfall or splashing waves, the evaporation of droplets on the outer shell, resulting in short-term "pseudo-cooling" signals, can mask true temperature trends. Due to the lack of heat shielding and dynamic compensation mechanisms, existing technologies can suffer data distortion rates as high as 30% in areas with alternating sunny and rainy weather, severely limiting the reliability of aquaculture decisions. Therefore, a marine aquaculture environment water temperature monitoring system is proposed that accurately decouples bioturbation from natural environmental variables in marine aquaculture water temperature monitoring. Summary of the Invention

[0005] To solve the above problems, the present invention provides a water temperature monitoring system for a marine aquaculture environment. Through the synergistic effect of a dynamic depth adjustment module and a thermal compensation algorithm module, it solves the problems of thermocline misjudgment and evaporative cooling false signal interference caused by biological activities, and realizes accurate monitoring of the aquaculture water layer temperature and environmental regulation.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows: A water temperature monitoring system for a marine aquaculture environment, comprising a dynamic depth adjustment module, a sensor unit, a multi-parameter fusion analysis module, a passive heat shield module, a heat compensation algorithm module and a central control unit, wherein:

[0007] The dynamic depth adjustment module is used to collect data and dynamically adjust the vertical distribution position of the sensor unit based on the local temperature rise caused by the swimming of fish clusters. Based on the soundprint characteristics of the fish's tail swing frequency, a convolutional neural network is used to extract the swimming energy parameters in real time. By calculating the temperature rise rate of biological disturbance at the bottom of the cage, when the temperature rise rate exceeds a preset threshold, the sensor unit's depth dynamic adaptation algorithm is triggered, and a vertical position adjustment instruction is generated and transmitted to the sensor unit;

[0008] The sensor unit is used to adjust the vertical distribution position according to the dynamic adjustment instruction, collect the temperature data of the target water layer in real time through the distributed temperature probe, and synchronously transmit the data to the multi-parameter fusion analysis module; it is also used to collect temperature data, droplet distribution morphology data, fish swimming sound pattern data, salinity-depth profile data and raindrop particle size distribution data, and transmit them to the dynamic depth adjustment module, multi-parameter fusion analysis module and thermal compensation algorithm module respectively;

[0009] The multi-parameter fusion analysis module combines temperature data transmitted by sensor units with data on fish vertical migration behavior. When vertical temperature stratification occurs, it uses a cross-scale spatiotemporal feature fusion model to distinguish between natural thermoclines and micro-temperature differences caused by biological disturbances. A dual-channel convolutional network is constructed. The first channel analyzes density mutation points in the salinity-depth profile to locate natural thermoclines, while the second channel analyzes the spatial correlation between the water layer where fish reside and temperature disturbances. Adversarial training is used to eliminate temperature noise caused by biological behavior, and the corrected temperature distribution map is output to the central control unit.

[0010] The passive heat shield module is used to suppress parasitic heat conduction to the internal probe under strong sunlight conditions. The passive heat shield module includes a heat flow sensing unit, a composite layer, and a thermal hysteresis compensation subunit. The heat flow sensing unit collects the shell surface temperature rise gradient data in real time and transmits the shell surface heat flow data to the thermal compensation algorithm module in real time through the thermal hysteresis compensation subunit.

[0011] The thermal compensation algorithm module receives the shell surface temperature rise gradient data transmitted by the passive heat shield module and eliminates evaporative cooling false signals in areas with alternating sunny and rainy weather. Specifically, it uses the droplet distribution morphology data collected by the sensor unit, segments the droplet coverage area through the U-Net network, establishes a non-steady-state evaporation dynamics model, and calculates the total amount of evaporative heat absorption. It uses the second-order inertial link transfer function to predict the transient cooling curve caused by rainfall splashing, combines it with a Kalman filter to eliminate noise, and outputs the true water temperature signal to the central control unit.

[0012] The central control unit is used to integrate the corrected temperature distribution map and the real water temperature signal, generate aquaculture environment control instructions and trigger the thermocline abnormality warning, and at the same time feedback the target water layer optimization parameters to the dynamic depth adjustment module; it is also used to fuse the heat flow data of the passive heat shield module and the output signal of the heat compensation algorithm module, iteratively optimize the real water temperature value through the weighted least squares method, and feed back the corrected parameters to the passive heat shield module.

[0013] Furthermore, the dynamic depth adjustment module includes an acoustic monitoring subunit, which is used to extract the tail swing frequency characteristics based on the fish swimming sound pattern data collected by the sensor unit using a convolutional neural network, and calculate the heat Q generated by the biological disturbance at the bottom of the cage using the metabolic heat release formula:

[0014] Q=α·f 2 ·ρ·V

[0015] Among them, α is the biomechanical coefficient, f is the swing frequency, ρ is the fish density, and V is the volume of the cage. Combined with the heat capacity parameter of the cage water, the heating rate per unit time is calculated:

[0016]

[0017] Among them, ρ water represents the water density, C is the specific heat capacity of water, Δt is the monitoring time interval, and when ΔT>0.5℃ / h, a bioturbation temperature rise warning signal is generated and transmitted to the deep dynamic adaptation algorithm to trigger the sensor unit adjustment instruction.

[0018] Furthermore, the dynamic depth adjustment module also includes an inertial navigation subunit, which is used to track the water layer where the fish are located based on the vertical motion trajectory data collected by the acceleration sensor implanted in the fish body. When ΔT exceeds the threshold and the water layer where the fish gather changes, the inertial navigation subunit is used to prioritize tracking the water layer where the fish are located, and use the track inference algorithm to reconstruct the distribution of the water layers where the fish live. By calculating the proportion of the residence time in each water layer, a water layer priority list is generated, and the list is transmitted to the sensor unit to dynamically update the monitoring depth.

[0019] Furthermore, the dynamic depth adjustment module also includes an inertial navigation subunit, which is used to track the water layer where the fish are located based on the vertical motion trajectory data collected by the micro-acceleration sensor implanted in the fish body. When ΔT exceeds the threshold and the water layer where the fish gather changes, the inertial navigation subunit is used to prioritize tracking the water layer where the fish are located, and use the track inference algorithm to reconstruct the distribution of the water layers where the fish live. By calculating the proportion of the residence time in each water layer, a water layer priority list is generated, and the list is transmitted to the sensor unit to dynamically update the monitoring depth.

[0020] Furthermore, the multi-parameter fusion analysis module includes a thermocline boundary identification subunit. The thermocline boundary identification subunit is used to identify the natural thermocline boundary based on the salinity-depth profile data collected by the sensor unit, using a density mutation point detection algorithm, combined with the fish habitat water layer data collected by the inertial navigation subunit, and filtering out the pseudo-thermocline signal caused by biological disturbance through a double-constrained optimization function, generating corrected thermocline position data and transmitting it to the central control unit.

[0021] Furthermore, the composite layer is integrated on the surface of the sensor unit housing, including from the outside to the inside:

[0022] Spectrally selective reflective layer: Made of ceramic-based composite material, with a reflectivity of ≥92% in the infrared band;

[0023] Dynamic phase change layer: Made of paraffin and graphene composite, with a phase change temperature range of 35-45°C, it is used to absorb the residual radiant heat that penetrates the reflective layer;

[0024] Adaptive thermal insulation layer: Using temperature-sensitive aerogel material, the expansion mechanism is triggered when the shell surface temperature rise rate is greater than 0.5℃ / min. After expansion, the thermal conductivity is ≤0.018W / (m·K);

[0025] The dynamic phase change layer calculates the heat absorption delay time of the phase change material based on the temperature rise gradient data of the shell surface using a thermal resistance network model. When the temperature rise rate is greater than 0.5°C / min, a phase change layer saturation signal is generated, triggering the adaptive insulation layer expansion instruction and transmitting it to the thermal compensation algorithm module.

[0026] Furthermore, the thermal compensation algorithm module includes an evaporative cooling modeling subunit. The evaporative cooling modeling subunit uses the U-Net network to segment the droplet coverage area based on the droplet distribution morphology data collected by the sensor unit, extracts the droplet equivalent diameter d, contact angle θ, and coverage S, and calculates the total amount of evaporative heat absorption:

[0027]

[0028] Where K represents the mass transfer coefficient, P sat and P air Represents the saturated vapor pressure of water at the current temperature and the actual water vapor partial pressure in the air, R thermairepresents the thermal resistance of heat transfer from the sensor unit housing to the droplet, and β represents the droplet morphology correction factor;

[0029] The dynamic compensation coefficient is generated by the total amount of evaporation heat absorption, and the probe measurement value is corrected by the Kalman filter.

[0030] Furthermore, the thermal compensation algorithm module also includes a rainfall transient response subunit. The rainfall transient response subunit uses an impact dynamics model to simulate the water film formation process on the shell surface based on the raindrop particle size distribution data collected by the sensor unit, predicts the transient cooling response curve, eliminates high-frequency noise components through a Kalman filter, and transmits the corrected water temperature signal to the central control unit.

[0031] Furthermore, it also includes a data verification module, which is used to calculate the node temperature difference threshold based on the temperature data collected by adjacent sensor units using a spatial correlation test algorithm. When the temperature difference is ≥0.3℃, an abnormal alarm signal is generated and transmitted to the central control unit to trigger the sensor self-test program.

[0032] Furthermore, the data verification module includes a bioturbation simulation subunit. The bioturbation simulation subunit uses the discrete element method to simulate the turbulent field generated by the stirring of fish schools based on the fish swimming frequency data of the acoustic monitoring subunit and the fish habitat water layer data of the inertial navigation subunit. The dynamic response correction parameters are generated by calculating the eddy diffusion coefficient and transmitted to the multi-parameter fusion analysis module to optimize the temperature noise filtering algorithm.

[0033] Furthermore, the method of constructing the fish school motion model using the discrete element model is as follows:

[0034] The equivalent mechanical parameters of the fish body are defined as follows: swing amplitude A = k1·f, stirring range R = k2·f, where k1 and k2 are calibration coordinate coefficients, and v is the migration speed;

[0035] Generate the energy spectral density E(k) of the turbulent field generated by the fish school stirring and calculate the equivalent eddy diffusion coefficient

[0036] According to the eddy diffusion coefficient D v , modify the temperature noise filtering algorithm of the multi-parameter fusion analysis module, and add a biological perturbation constraint term to the adversarial training loss function:

[0037]

[0038] Among them, λ is the weight factor, For the temperature gradient, the convolution kernel parameters of the dual-channel convolutional network are updated by back propagation, which reduces the natural thermocline positioning error to ±0.3m;

[0039] The corrected temperature noise filtering parameters are transmitted to the central control unit to optimize the generation logic of the aquaculture environment control instructions: when the deviation between the corrected thermocline position and the target water layer is detected to be greater than 1.2m, the aerator power adaptive adjustment mode is triggered; according to the bioturbation intensity D v Dynamically adjust the sampling frequency of the sensor unit, in D v >0.1m 2 / s and increase the sampling rate from 1Hz to 5Hz.

[0040] The above scheme has the following beneficial effects:

[0041] 1. This solution uses an acoustic monitoring subunit to analyze the frequency characteristics of fish tail swings in real time, combines it with a metabolic thermodynamic model to accurately calculate the temperature rise rate caused by biological disturbance, and links it with an inertial navigation subunit to track the vertical migration trajectory of fish schools. It also dynamically adjusts the monitoring depth of the sensor unit. This effectively overcomes the problem of traditional fixed sensors missing key ecological niche data due to their inability to adapt to changes in the water layer where farmed species live. It ensures that in the scenario of localized temperature rise caused by the movement of fish schools, the sensor array always focuses on the target water layer, avoiding the failure of hypoxia warning caused by misjudgment of the thermocline.

[0042] 2. This solution, based on a cross-scale spatiotemporal feature fusion model using a dual-channel convolutional network, combines the detection of density mutation points in salinity-depth profiles with the analysis of spatial correlation between fish behavior and temperature disturbances. Through an adversarial training mechanism, it eliminates the micro-temperature difference noise caused by biological activities, achieves high-precision positioning of the natural thermocline boundary, and completely solves the problem of thermocline misjudgment caused by the traditional single-parameter analysis method that ignores the biothermodynamic effects.

[0043] 3. This solution adopts a composite structure design of a spectrally selective reflection layer, a dynamic phase change layer, and an adaptive thermal insulation layer to achieve three-level attenuation of solar radiation heat flux (reflection → phase change absorption → aerogel insulation), so that the temperature rise of the probe inside the shell is ≤0.3°C. Combined with the non-steady-state evaporation dynamics model of the thermal compensation algorithm module and the inertial-Kalman hybrid filtering technology, the total amount of heat absorption by droplet evaporation is quantified and the transient pseudo-cooling signal of rainfall is separated. The measurement deviation under alternating sunny and rainy conditions is reduced from ±1.2°C to ±0.15°C, breaking through the data distortion bottleneck caused by the thermal hysteresis effect and evaporative cooling coupling interference of traditional sensors.

[0044] 4. In this solution, the central control unit dynamically optimizes sensor parameters and control instructions by iteratively fusing the heat flow data of the heat shield module and the correction signal of the thermal compensation algorithm through the weighted least squares method. The data verification module identifies hardware anomalies and corrects temperature response errors in real time based on the spatial correlation test of redundant sensors and the simulation of bioturbation turbulent fields, forming a fully closed-loop control system of "monitoring-compensation-verification-control", which significantly improves the long-term operational stability of the system in complex marine environments.

[0045] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 Schematic diagram of the system framework of an embodiment of a water temperature monitoring system for a marine aquaculture environment according to the present invention. DETAILED DESCRIPTION

[0047] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0048] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0049] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0050] The following is further described in detail through specific implementation methods:

[0051] Example:

[0052] As attached Figure 1 Shown: A water temperature monitoring system for a marine aquaculture environment, including a dynamic depth adjustment module, a sensor unit, a multi-parameter fusion analysis module, a passive heat shielding module, a heat compensation algorithm module and a central control unit, wherein:

[0053] The dynamic depth adjustment module is used to collect data and dynamically adjust the vertical distribution position of the sensor unit based on the local temperature rise caused by the swimming of fish clusters. Based on the soundprint characteristics of the fish's tail swing frequency, a convolutional neural network is used to extract the swimming energy parameters in real time. By calculating the temperature rise rate of biological disturbance at the bottom of the cage, when the temperature rise rate exceeds a preset threshold, the sensor unit's depth dynamic adaptation algorithm is triggered, and a vertical position adjustment instruction is generated and transmitted to the sensor unit;

[0054] The dynamic depth adjustment module includes an acoustic monitoring subunit. The acoustic monitoring subunit is used to extract the tail swing frequency characteristics based on the fish swimming sound pattern data collected by the sensor unit using a convolutional neural network, and calculate the heat Q generated by the biological disturbance at the bottom of the cage using the metabolic heat release formula:

[0055] Q=α·f 2 ·ρ·V

[0056] Among them, α is the biomechanical coefficient, f is the swing frequency, ρ is the fish density, and V is the volume of the cage. Combined with the heat capacity parameter of the cage water, the heating rate per unit time is calculated:

[0057]

[0058] Among them, ρ water represents the water density, C is the specific heat capacity of water, Δt is the monitoring time interval, and when ΔT>0.5℃ / h, a bioturbation temperature rise warning signal is generated and transmitted to the deep dynamic adaptation algorithm to trigger the sensor unit adjustment instruction;

[0059] The dynamic depth adjustment module also includes an inertial navigation subunit, which is used to track the water layer where the fish are located based on the vertical motion trajectory data collected by the acceleration sensor implanted in the fish body. When ΔT exceeds the threshold and the water layer where the fish gather changes, the inertial navigation subunit prioritizes tracking the water layer where the fish are located, and uses the track estimation algorithm to reconstruct the distribution of the water layers where the fish live. By calculating the proportion of the residence time in each water layer, a water layer priority list is generated, and the list is transmitted to the sensor unit to dynamically update the monitoring depth.

[0060] The sensor unit is used to adjust the vertical distribution position according to the dynamic adjustment instructions, collect the temperature data of the target water layer in real time through distributed high-precision temperature probes, and synchronously transmit the data to the multi-parameter fusion analysis module; it is also used to collect temperature data, droplet distribution morphology data, fish swimming soundprint data, salinity-depth profile data and raindrop particle size distribution data, and transmit them to the dynamic depth adjustment module, multi-parameter fusion analysis module and thermal compensation algorithm module respectively.

[0061] The multi-parameter fusion analysis module combines temperature data transmitted by sensor units with data on fish vertical migration behavior. When vertical temperature stratification occurs, it uses a cross-scale spatiotemporal feature fusion model to distinguish between natural thermoclines and micro-temperature differences caused by biological disturbances. A dual-channel convolutional network is constructed. The first channel analyzes density mutation points in the salinity-depth profile to locate natural thermoclines, while the second channel analyzes the spatial correlation between the water layer where fish reside and temperature disturbances. Adversarial training is used to eliminate temperature noise caused by biological behavior, and the corrected temperature distribution map is output to the central control unit.

[0062] The multi-parameter fusion analysis module includes a thermocline boundary identification subunit. The thermocline boundary identification subunit is used to identify the natural thermocline boundary based on the salinity-depth profile data collected by the sensor unit, using a density mutation point detection algorithm. Combined with the fish habitat water layer data collected by the inertial navigation subunit, the pseudo-thermocline signal caused by biological disturbance is filtered out through a double-constrained optimization function, and the corrected thermocline position data is generated and transmitted to the central control unit.

[0063] The passive heat shield module is used to suppress parasitic heat conduction to the internal probe under strong sunlight conditions. The passive heat shield module includes a heat flow sensing unit, a composite layer, and a thermal hysteresis compensation subunit. The heat flow sensing unit collects the shell surface temperature rise gradient data in real time and transmits the shell surface heat flow data to the thermal compensation algorithm module in real time through the thermal hysteresis compensation subunit. The composite layer is integrated into the surface of the sensor unit shell, including from the outside to the inside:

[0064] Spectrally selective reflective layer: Made of ceramic-based composite material, with a reflectivity of ≥92% in the infrared band;

[0065] Dynamic phase change layer: Made of paraffin and graphene composite, with a phase change temperature range of 35-45°C, it is used to absorb the residual radiant heat that penetrates the reflective layer;

[0066] Adaptive thermal insulation layer: Using temperature-sensitive aerogel material, the expansion mechanism is triggered when the shell surface temperature rise rate is greater than 0.5℃ / min. After expansion, the thermal conductivity is ≤0.018W / (m·K);

[0067] The dynamic phase change layer calculates the heat absorption delay time of the phase change material based on the temperature rise gradient data of the shell surface using a thermal resistance network model. When the temperature rise rate is greater than 0.5°C / min, a phase change layer saturation signal is generated, triggering the adaptive insulation layer expansion instruction and transmitting it to the thermal compensation algorithm module.

[0068] The thermal compensation algorithm module receives the shell surface temperature rise gradient data transmitted by the passive heat shield module and eliminates evaporative cooling false signals in areas with alternating sunny and rainy weather. Specifically, it uses the droplet distribution morphology data collected by the high-resolution infrared thermal imager in the sensor unit, segments the droplet coverage area through the U-Net network, establishes a non-steady-state evaporation dynamics model, and calculates the total amount of evaporative heat absorption. It uses the second-order inertial link transfer function to predict the transient cooling curve caused by rainfall splashing. It then uses a Kalman filter to eliminate high-frequency noise and outputs the true water temperature signal to the central control unit.

[0069] The thermal compensation algorithm module includes an evaporative cooling modeling subunit. The evaporative cooling modeling subunit uses a U-Net network to segment the droplet coverage area based on the droplet distribution morphology data collected by a high-resolution infrared thermal imager, extract the droplet equivalent diameter d, contact angle θ, and coverage S, and calculate the total evaporative heat absorption:

[0070]

[0071] Where K represents the mass transfer coefficient, P sat and P air Represents the saturated vapor pressure of water at the current temperature and the actual water vapor partial pressure in the air, R thermai represents the thermal resistance of heat transfer from the sensor unit housing to the droplet, and β represents the droplet morphology correction factor;

[0072] Generate dynamic compensation coefficients based on the total amount of evaporation heat absorption and correct probe measurement values ​​using a Kalman filter;

[0073] The thermal compensation algorithm module also includes a rainfall transient response subunit. The rainfall transient response subunit uses an impact dynamics model to simulate the water film formation process on the shell surface based on the raindrop particle size distribution data collected by the raindrop spectrometer, predicts the transient cooling response curve, eliminates high-frequency noise components through a Kalman filter, and transmits the corrected water temperature signal to the central control unit.

[0074] The central control unit is used to integrate the corrected temperature distribution map and the real water temperature signal, generate aquaculture environment control instructions and trigger the thermocline abnormality warning, and at the same time feedback the target water layer optimization parameters to the dynamic depth adjustment module; it is also used to fuse the heat flow data of the passive heat shield module and the output signal of the heat compensation algorithm module, iteratively optimize the real water temperature value through the weighted least squares method, and feed back the corrected parameters to the passive heat shield module.

[0075] It also includes a data verification module, which is used to calculate the node temperature difference threshold using a spatial correlation test algorithm based on the temperature data collected by adjacent sensor units. When the temperature difference is ≥0.3°C, an abnormal alarm signal is generated and transmitted to the central control unit to trigger the sensor self-test program;

[0076] The data verification module includes a bioturbation simulation subunit. Based on the fish swimming frequency data from the acoustic monitoring subunit and the fish habitat data from the inertial navigation subunit, the bioturbation simulation subunit uses the discrete element method to simulate the turbulent flow field generated by the fish school. The dynamic response correction parameters are generated by calculating the eddy diffusion coefficient and transmitted to the multi-parameter fusion analysis module to optimize the temperature noise filtering algorithm.

[0077] The method of constructing fish school motion model using discrete element model is as follows:

[0078] The equivalent mechanical parameters of the fish body are defined as follows: swing amplitude A = k1·f, stirring range R = k2·f, where k1 and k2 are calibration coordinate coefficients, and v is the migration speed;

[0079] Generate the energy spectral density E(k) of the turbulent field generated by the fish school stirring and calculate the equivalent eddy diffusion coefficient

[0080] According to the eddy diffusion coefficient D v , modify the temperature noise filtering algorithm of the multi-parameter fusion analysis module, and add a biological perturbation constraint term to the adversarial training loss function:

[0081]

[0082] Among them, λ is the weight factor, For the temperature gradient, the convolution kernel parameters of the dual-channel convolutional network are updated by back propagation, which reduces the natural thermocline positioning error to ±0.3m;

[0083] The corrected temperature noise filtering parameters are transmitted to the central control unit to optimize the generation logic of the aquaculture environment control instructions: when the deviation between the corrected thermocline position and the target water layer is detected to be greater than 1.2m, the aerator power adaptive adjustment mode is triggered; according to the bioturbation intensity D v Dynamically adjust the sampling frequency of the sensor unit, in D v >0.1m 2 / s and increase the sampling rate from 1Hz to 5Hz.

[0084] The specific implementation process is as follows: This example uses a high-density breeding cage of golden pomfret in the South China Sea, with a cage volume of 50m 3 Due to the frequent alternation of sunny and rainy weather in the South China Sea in summer, the average daily sunshine intensity is 1200W / m 2, with instantaneous rainfall intensities reaching 50 mm / h. Traditional water temperature monitoring systems suffer from data distortion due to the following issues: 1. After rainfall, droplets evaporate from the sensor housing, absorbing heat and causing a short-term "false cooling" with measurement deviations of up to ±1.2°C. 2. Fish movement causes localized temperature rises (0.5-0.8°C / h), making it impossible for traditional fixed sensors to distinguish between natural thermoclines and bioturbation. 3. Under strong sunlight, the probe temperature inside the sensor housing rises by up to 1.5°C, masking true water temperature fluctuations.

[0085] Therefore, the water temperature monitoring system for the marine aquaculture environment proposed in this scheme collects the soundprint data of fish swimming through the hydrophone array in the sensor unit. The collected soundprint data has a frequency of 4-6 Hz, and the tail swing energy characteristics are extracted through the convolutional neural network. The metabolic heat release Q = 90 J / s is calculated, and the temperature rise rate is calculated by combining the heat capacity formula to obtain ΔT≈0.12℃ / h. At this time, ΔT does not exceed the threshold and is not adjusted for the time being.

[0086] The inertial navigation subunit collects the acceleration sensor implanted in the fish body of the sensor unit, which monitors the migration of fish to the 8-10m water layer, where the residence time is greater than 70%. A priority list is generated and transmitted to the sensor unit. The sensor unit adjusts the distributed probe to a depth of 8m according to the priority list, collects real-time temperature data of 25.3℃±0.2℃, and transmits it synchronously to the multi-parameter fusion analysis module.

[0087] The thermocline boundary identification subunit collects salinity profile data based on the CTD sensor in the sensor unit. The density mutation point detection algorithm identifies that the natural thermocline is located at a depth of 12m. Combined with the water layer where fish gather at 8-10m, the bioturbation signal is filtered out through the double-constraint optimization function, and the corrected thermocline position is output to the central control unit.

[0088] At a sunlight intensity of 1100W / m 2 Under the condition of , the surface temperature rise rate of the composite layer shell in the passive thermal shielding module is monitored to be 0.6℃ / min. At this time, the dynamic phase change layer absorbs heat and the aerogel expands, the thermal conductivity drops to 0.016W / (m·K), the internal probe temperature rises by only 0.2℃, and the thermal hysteresis compensation subunit converts the shell heat flow data Q=85W / m 2 Transmitted to the thermal compensation algorithm module.

[0089] The evaporative cooling modeling subunit captures the droplet distribution on the shell surface through the high-resolution infrared thermal imager in the sensor unit, and obtains the coverage S = 35%, the equivalent diameter d = 1.2 mm, and the contact angle θ = 60°. The heat absorption Q is calculated using the non-steady-state evaporation dynamics model. evap =0.032W, a compensation coefficient of +0.28°C is generated to correct the probe measurement value, and the evaporation heat absorption is quantified based on the droplet segmentation and unsteady evaporation dynamics model of U-Net.

[0090] When rainfall occurs, the rainfall transient response subunit detects a rainfall intensity of 40 mm / h and a raindrop size of 3-5 mm based on the raindrop spectrometer in the sensor unit. The impact dynamics model is used to predict the water film formation time T1 = 10 s and T2 = 20 s. The second-order inertia link transfer function is used to calculate:

[0091]

[0092] After filtering, high-frequency noise is eliminated and the actual water temperature signal is output as 25.5℃±0.15℃, which is transmitted to the central control unit.

[0093] The central control unit integrates the corrected thermocline position (12m) and the actual water temperature (25.5℃), detects that the deviation between the target water layer (8m) and the thermocline is greater than 1.2m, triggers the aerator power to increase to 80%, and uses weighted least squares iterative optimization to feed back the thermal compensation parameters to the passive heat shield module, adjusting the aerogel expansion threshold to 0.4℃ / min.

[0094] Through the data verification module, when some redundant sensors in the sensor unit detect a temperature difference of 0.4°C between adjacent nodes, and the temperature difference is greater than the threshold of 0.3°C, the self-check program is triggered and the backup probe is switched. The eddy diffusion coefficient D is calculated based on the fish swimming data (f = 5Hz, v = 0.3m / s). v =0.18m 2 / s, and modified the adversarial training weight λ of the multi-parameter fusion module to 0.6, reducing the thermocline positioning error from 1.1m to 0.3m. During 30 consecutive days of testing at a South China Sea aquaculture farm, the system successfully eliminated 92% of evaporative cooling artifacts, ensuring water temperature monitoring accuracy of ±0.2°C and reducing fish mortality by 19%.

[0095] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will readily appreciate that other variations or modifications based on the above descriptions are possible. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. A water temperature monitoring system for a marine aquaculture environment, characterized in that: It includes a dynamic depth adjustment module, a sensor unit, a multi-parameter fusion analysis module, a passive heat shielding module, a heat compensation algorithm module and a central control unit, wherein: The dynamic depth adjustment module is used to collect data and dynamically adjust the vertical distribution position of the sensor unit based on the local temperature rise caused by the swimming of fish clusters. Based on the soundprint characteristics of the fish's tail swing frequency, a convolutional neural network is used to extract the swimming energy parameters in real time. By calculating the temperature rise rate of biological disturbance at the bottom of the cage, when the temperature rise rate exceeds a preset threshold, the sensor unit's depth dynamic adaptation algorithm is triggered, and a vertical position adjustment instruction is generated and transmitted to the sensor unit; The sensor unit is used to adjust the vertical distribution position according to the dynamic adjustment instruction, collect the temperature data of the target water layer in real time through the distributed temperature probe, and synchronously transmit the data to the multi-parameter fusion analysis module; it is also used to collect temperature data, droplet distribution morphology data, fish swimming sound pattern data, salinity-depth profile data and raindrop particle size distribution data, and transmit them to the dynamic depth adjustment module, multi-parameter fusion analysis module and thermal compensation algorithm module respectively; The multi-parameter fusion analysis module combines temperature data transmitted by sensor units with data on fish vertical migration behavior. When vertical temperature stratification occurs, it uses a cross-scale spatiotemporal feature fusion model to distinguish between natural thermoclines and micro-temperature differences caused by biological disturbances. A dual-channel convolutional network is constructed. The first channel analyzes density mutation points in the salinity-depth profile to locate natural thermoclines, while the second channel analyzes the spatial correlation between the water layer where fish reside and temperature disturbances. Adversarial training is used to eliminate temperature noise caused by biological behavior, and the corrected temperature distribution map is output to the central control unit. The passive heat shield module is used to suppress parasitic heat conduction to the internal probe under strong sunlight conditions. The passive heat shield module includes a heat flow sensing unit, a composite layer, and a thermal hysteresis compensation subunit. The heat flow sensing unit collects the shell surface temperature rise gradient data in real time and transmits the shell surface heat flow data to the thermal compensation algorithm module in real time through the thermal hysteresis compensation subunit. The thermal compensation algorithm module receives the shell surface temperature rise gradient data transmitted by the passive heat shield module and eliminates evaporative cooling false signals in areas with alternating sunny and rainy weather. Specifically, it uses the droplet distribution morphology data collected by the sensor unit, segments the droplet coverage area through the U-Net network, establishes a non-steady-state evaporation dynamics model, and calculates the total amount of evaporative heat absorption. It uses the second-order inertial link transfer function to predict the transient cooling curve caused by rainfall splashing, combines it with a Kalman filter to eliminate noise, and outputs the true water temperature signal to the central control unit. The central control unit is used to integrate the corrected temperature distribution map and the real water temperature signal, generate aquaculture environment control instructions and trigger the thermocline abnormality warning, and at the same time feedback the target water layer optimization parameters to the dynamic depth adjustment module; it is also used to fuse the heat flow data of the passive heat shield module and the output signal of the heat compensation algorithm module, iteratively optimize the real water temperature value through the weighted least squares method, and feed back the corrected parameters to the passive heat shield module.

2. The water temperature monitoring system for a marine aquaculture environment according to claim 1, characterized in that: The dynamic depth adjustment module includes an acoustic monitoring subunit. The acoustic monitoring subunit is used to extract the tail swing frequency characteristics based on the fish swimming sound pattern data collected by the sensor unit using a convolutional neural network, and calculate the heat Q generated by the biological disturbance at the bottom of the cage using the metabolic heat release formula: Q=α·f 2 ·p·V Among them, α is the biomechanical coefficient, f is the swing frequency, ρ is the fish density, and V is the volume of the cage. Combined with the heat capacity parameter of the cage water, the heating rate per unit time is calculated: Among them, ρ water represents the water density, C is the specific heat capacity of water, Δt is the monitoring time interval, and when ΔT>0.5℃ / h, a bioturbation temperature rise warning signal is generated and transmitted to the deep dynamic adaptation algorithm to trigger the sensor unit adjustment instruction.

3. The water temperature monitoring system for a marine aquaculture environment according to claim 2, characterized in that: The dynamic depth adjustment module also includes an inertial navigation subunit, which is used to track the water layer where the fish are located based on the vertical motion trajectory data collected by the acceleration sensor implanted in the fish body. When ΔT exceeds the threshold and the water layer where the fish gather changes, the inertial navigation subunit prioritizes tracking the water layer where the fish are located, and uses the track estimation algorithm to reconstruct the distribution of the water layers where the fish live. By calculating the proportion of the residence time in each water layer, a water layer priority list is generated, and the list is transmitted to the sensor unit to dynamically update the monitoring depth.

4. The water temperature monitoring system for a marine aquaculture environment according to claim 3, characterized in that: The multi-parameter fusion analysis module includes a thermocline boundary identification subunit. The thermocline boundary identification subunit is used to identify the natural thermocline boundary based on the salinity-depth profile data collected by the sensor unit, using a density mutation point detection algorithm. Combined with the fish habitat water layer data collected by the inertial navigation subunit, the pseudo-thermocline signal caused by biological disturbance is filtered out through a double-constrained optimization function, and the corrected thermocline position data is generated and transmitted to the central control unit.

5. The water temperature monitoring system for a marine aquaculture environment according to claim 4, characterized in that: The composite layer is integrated on the surface of the sensor unit housing, including from the outside to the inside: Spectrally selective reflective layer: Made of ceramic-based composite material, with a reflectivity of ≥92% in the infrared band; Dynamic phase change layer: Made of paraffin and graphene composite, with a phase change temperature range of 35-45°C, it is used to absorb the residual radiant heat that penetrates the reflective layer; Adaptive thermal insulation layer: Using temperature-sensitive aerogel material, the expansion mechanism is triggered when the shell surface temperature rise rate is greater than 0.5℃ / min. After expansion, the thermal conductivity is ≤0.018W / (m·K); The dynamic phase change layer calculates the heat absorption delay time of the phase change material based on the temperature rise gradient data of the shell surface using a thermal resistance network model. When the temperature rise rate is greater than 0.5°C / min, a phase change layer saturation signal is generated, triggering the adaptive insulation layer expansion instruction and transmitting it to the thermal compensation algorithm module.

6. The water temperature monitoring system for a marine aquaculture environment according to claim 5, characterized in that: The thermal compensation algorithm module includes an evaporative cooling modeling subunit. The evaporative cooling modeling subunit uses the U-Net network to segment the droplet coverage area based on the droplet distribution morphology data collected by the sensor unit, extracts the droplet equivalent diameter d, contact angle θ, and coverage S, and calculates the total amount of evaporative heat absorption: Where K represents the mass transfer coefficient, P sat and P air Represents the saturated vapor pressure of water at the current temperature and the actual water vapor partial pressure in the air, R thermai represents the thermal resistance of heat transfer from the sensor unit housing to the droplet, and β represents the droplet morphology correction factor; The dynamic compensation coefficient is generated by the total amount of evaporation heat absorption, and the probe measurement value is corrected by the Kalman filter.

7. The water temperature monitoring system for a marine aquaculture environment according to claim 6, characterized in that: The thermal compensation algorithm module also includes a rainfall transient response subunit. The rainfall transient response subunit uses an impact dynamics model to simulate the water film formation process on the shell surface based on the raindrop particle size distribution data collected by the sensor unit, predicts the transient cooling response curve, eliminates the noise component through the Kalman filter, and transmits the corrected water temperature signal to the central control unit.

8. The water temperature monitoring system for a marine aquaculture environment according to claim 7, characterized in that: It also includes a data verification module, which is used to calculate the node temperature difference threshold based on the temperature data collected by adjacent sensor units using a spatial correlation test algorithm. When the temperature difference is ≥0.3℃, an abnormal alarm signal is generated and transmitted to the central control unit to trigger the sensor self-test program.

9. The water temperature monitoring system for a marine aquaculture environment according to claim 8, characterized in that: The data verification module includes a bioturbation simulation subunit. The bioturbation simulation subunit uses the discrete element method to simulate the turbulent field caused by the stirring of fish schools based on the fish swimming frequency data from the acoustic monitoring subunit and the fish habitat water layer data from the inertial navigation subunit. The dynamic response correction parameters are generated by calculating the eddy diffusion coefficient and transmitted to the multi-parameter fusion analysis module to optimize the temperature noise filtering algorithm.

10. The water temperature monitoring system for a marine aquaculture environment according to claim 9, characterized in that: The method of constructing fish school motion model using discrete element model is as follows: The equivalent mechanical parameters of the fish body are defined as follows: swing amplitude A = k1·f, stirring range R = k2·f, where k1 and k2 are calibration coordinate coefficients, and v is the migration speed; Generate the energy spectral density E(k) of the turbulent field generated by the fish school stirring and calculate the equivalent eddy diffusion coefficient According to the eddy diffusion coefficient D v , modify the temperature noise filtering algorithm of the multi-parameter fusion analysis module, and add a biological perturbation constraint term to the adversarial training loss function: Among them, λ is the weight factor, For the temperature gradient, the convolution kernel parameters of the dual-channel convolutional network are updated by back propagation, which reduces the natural thermocline positioning error to ±0.3m; The corrected temperature noise filtering parameters are transmitted to the central control unit to optimize the generation logic of the aquaculture environment control instructions: when the deviation between the corrected thermocline position and the target water layer is detected to be greater than 1.2m, the aerator power adaptive adjustment mode is triggered; according to the bioturbation intensity D v Dynamically adjust the sampling frequency of the sensor unit, in D v >0.1m 2 / s and increase the sampling rate from 1Hz to 5Hz.

Images