An atlantic salmon farming system under an lng-ifv system

Abstract

The present application relates to the Atlantic salmon breeding technical field, disclose a kind of LNG-IFV system under the Atlantic salmon breeding system.The system includes water body monitoring module, real-time collection LNG-IFV heat seawater temperature gradient after, dissolved oxygen concentration, flow velocity distribution and Atlantic salmon feeding activity, population distribution density;Breeding partition module is based on temperature gradient, flow velocity distribution divides breeding pond fluid dynamic area, in combination with population distribution density Set each area breeding density threshold value;State portrait module extracts historical seawater parameter and biological behavior data to construct historical state portrait, and obtains current real-time state portrait;Stress response module calculates water quality and biological behavior deviation vector, based on water quality deviation vector identifies abnormal stress factor;Feeding optimization module is associated with abnormal stress factor and feeding activity to generate feeding strategy adjustment instruction;Parameter decision module outputs LNG cold energy input control parameter and feeding machine control parameter, realizes the accurate control of Atlantic salmon breeding.

Description

Technical Field

[0001] This invention relates to the field of Atlantic salmon farming technology, specifically to an Atlantic salmon farming system under an LNG-IFV system. Background Technology

[0002] Atlantic salmon, as a globally important economically farmed fish, is experiencing a shift towards intensive and large-scale aquaculture, leading to an increasing demand for precise environmental control. Currently, Atlantic salmon farming largely relies on traditional static aquaculture methods, often facing low energy efficiency in seawater temperature control. Some projects have attempted to introduce cold energy utilization technologies, but existing technologies struggle to precisely match cold energy with the needs of the aquaculture water, resulting in uneven seawater temperature distribution after heat exchange and the emergence of localized temperature anomalies, which negatively impacts the normal growth of Atlantic salmon.

[0003] In water quality monitoring, traditional methods mostly focus on timed collection of single parameters, failing to acquire multi-dimensional parameters such as seawater temperature gradient, dissolved oxygen concentration, and flow velocity distribution in real time, making it difficult to comprehensively reflect the dynamic changes in aquaculture water. Simultaneously, monitoring of Atlantic salmon biological behavior data is scarce, relying solely on manual observation and recording of feeding activities, which cannot accurately capture subtle changes in feeding activity and population density, leading to a lag in assessing the suitability of the aquaculture environment for fish growth.

[0004] Regarding the setting of stocking density, existing models mostly adopt fixed density standards without considering the differences in fluid dynamics within the stocking pond. Due to the uneven distribution of seawater flow velocity, the water exchange capacity and dissolved oxygen supply vary in different areas of the stocking pond. A uniform stocking density can easily lead to excessively high fish density in some areas, resulting in problems such as insufficient dissolved oxygen and accumulation of excrement, while other areas may suffer from excessively low density, leading to a waste of space resources.

[0005] In terms of stress response management, existing technologies lack systematic integration and analysis of historical and real-time data. When anomalies occur in the aquaculture environment, problems can only be judged based on a single real-time parameter. It is impossible to combine seawater parameters and biological behavior data from historical periods to accurately identify the type and cause of abnormal stress factors. This results in insufficiently targeted stress response measures and difficulty in quickly mitigating the adverse effects of environmental changes on Atlantic salmon.

[0006] In the feeding strategy formulation stage, existing feeders mostly operate according to preset times and feeding amounts, without establishing a correlation with fish feeding activity and abnormal environmental conditions. When water quality abnormalities or fish stress responses occur, the feeding amount and frequency cannot be adjusted in a timely manner, easily leading to feed waste. Furthermore, inappropriate feeding may increase the digestive burden on fish, affecting their growth. In addition, regarding the coordination of LNG cold energy utilization and aquaculture parameter control, existing systems lack a unified decision-making mechanism. The adjustment of cold energy input and feeder control parameters is independent, failing to achieve synergistic optimization of the aquaculture environment and feeding strategy, thus hindering the efficiency and quality improvement of Atlantic salmon farming. Summary of the Invention

[0007] The purpose of this invention is to provide an Atlantic salmon farming system under an LNG-IFV system to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides an Atlantic salmon farming system under an LNG-IFV system, the system comprising:

[0009] The water monitoring module is used to collect seawater parameters after LNG-IFV heat exchange in real time. The seawater parameters include temperature gradient, dissolved oxygen concentration, and flow velocity distribution. It also monitors the biological behavior data of Atlantic salmon, including feeding activity and population density.

[0010] The aquaculture zoning module is used to divide the fluid dynamic zone of the aquaculture pond based on the temperature gradient and the flow velocity distribution, and to set the aquaculture density threshold for each dynamic zone in combination with the population distribution density.

[0011] The status profile module is used to extract the seawater parameters and biological behavior data in the historical period, construct the historical status profile, and obtain the real-time status profile of the current period.

[0012] The stress response module is used to calculate water quality deviation vector and biological behavior deviation vector based on the historical state profile and the real-time state profile, and to identify abnormal stress factors based on the water quality deviation vector.

[0013] The feeding optimization module is used to associate the abnormal stress factors with the feeding activity level and generate feeding strategy adjustment instructions.

[0014] The parameter decision module is used to integrate the fluid dynamic region, the breeding density threshold and the feeding strategy adjustment command, and output LNG cold energy input regulation parameters and feeder control parameters.

[0015] Preferably, the aquaculture zoning module divides the fluid dynamic region by:

[0016] Based on the temperature gradient, the high and low temperature transition zone of the aquaculture pond is located, and combined with the flow velocity distribution, the eddy generation zone and the steady flow zone are marked.

[0017] Calculate the density variance of each region based on the population distribution density;

[0018] The aquaculture density threshold is dynamically adjusted based on the density variance and the preset density tolerance.

[0019] Preferably, the state profiling module constructs the historical state profile by including:

[0020] Extract the mean sequence of the seawater parameters and the peak sequence of the biological behavior data for N consecutive aquaculture cycles;

[0021] Based on the mean and peak sequences, the water quality stability coefficient and biological activity coefficient for each period are fitted.

[0022] By associating the water quality stability coefficient with the biological activity coefficient, the historical state profile is generated.

[0023] Preferably, the stress response module calculates the water quality deviation vector by:

[0024] Decompose the baseline dissolved oxygen concentration and baseline temperature fluctuation range in the historical state profile;

[0025] The dissolved oxygen concentration is compared with the baseline dissolved oxygen concentration in the real-time status image to obtain the dissolved oxygen deviation value;

[0026] Analyze the deviation of the current temperature fluctuation range from the reference temperature fluctuation range;

[0027] The dissolved oxygen deviation value and the deviation degree are combined to construct the water quality deviation vector.

[0028] Preferably, the stress response module identifies the abnormal stress factors by including:

[0029] When the magnitude of the water quality deviation vector exceeds the first threshold, the dissolved oxygen compensation mechanism is activated;

[0030] When the rate of change of the biological behavior deviation vector exceeds the second threshold, the group behavior correction mechanism is activated.

[0031] The dissolved oxygen compensation mechanism and the group behavior correction mechanism are correlated, and the abnormal stress factors are labeled.

[0032] Preferably, the feeding optimization module generates the feeding strategy adjustment instruction in the following ways:

[0033] Map the abnormal stress factors to a feeding response curve library and output the feeding attenuation coefficient;

[0034] The metabolic rate weights of each region are calculated based on the temperature distribution of the fluid dynamic region.

[0035] The feeding amount per unit time is dynamically adjusted by combining the feed attenuation coefficient and the metabolic rate weight.

[0036] Preferably, the parameter decision module outputs the LNG cold energy input control parameters including:

[0037] Based on the temperature distribution of the fluid dynamic region, the target heat transfer efficiency of each region is calculated.

[0038] Based on the aforementioned stocking density threshold, temperature control priority is set;

[0039] By combining the target heat exchange efficiency and the temperature control priority, the LNG vaporization rate increment is calculated.

[0040] Preferably, the parameter decision module outputs the following control parameters for the feeder:

[0041] Decompose the regional feeding amount correction value in the feeding strategy adjustment instruction;

[0042] Calculate the diffusion radius of the feeding point based on the coordinates of the eddy generation zone in the fluid dynamic region;

[0043] Based on the diffusion radius and the regional feeding amount correction value, the feeder movement trajectory and feeding sequence are generated.

[0044] Preferably, the system further includes a data backtracking module;

[0045] The data backtracking module is used to set the data backtracking period based on the change frequency of the water quality deviation vector;

[0046] Based on the data backtracking cycle, redundant data in the historical status profile that has exceeded the storage time limit is cleaned up.

[0047] Preferably, the data backtracking module sets the data backtracking period as follows:

[0048] The cumulative frequency of the magnitude of the water quality deviation vector exceeding the third threshold is counted.

[0049] When the cumulative frequency reaches the warning frequency, the data backtracking cycle is shortened.

[0050] If the biological behavior deviation vector fails to trigger the correction mechanism for M consecutive cycles, the data backtracking cycle is extended.

[0051] Compared with the prior art, the beneficial effects of the present invention are:

[0052] This Atlantic salmon farming system under the LNG-IFV system achieves real-time acquisition of multi-dimensional parameters of seawater and Atlantic salmon biological behavior data after LNG-IFV heat exchange by setting up a water monitoring module. This module can simultaneously acquire seawater temperature gradients, dissolved oxygen concentrations, flow velocity distributions, as well as fish feeding activity and population density. This overcomes the limitations of traditional monitoring methods, which suffer from single parameters and data lag. It allows for comprehensive and timely monitoring of the dynamics of the aquaculture water and the growth status of the fish, making the assessment of the suitability of the aquaculture environment and biological behavior more timely and accurate.

[0053] The aquaculture zoning module, based on temperature gradients and flow velocity distributions obtained from the water monitoring module, divides the aquaculture ponds into fluid dynamic zones. It then sets a density threshold for each zone based on population distribution density. This zoning method fully considers the differences in water exchange capacity and dissolved oxygen supply in different areas of the aquaculture pond, avoiding the problems of wasted space resources or excessively high local densities found in traditional fixed-density models. It ensures that the aquaculture density is adapted to the environmental carrying capacity of each zone, providing Atlantic salmon with a more suitable growth space and reducing growth differences caused by unsuitable environments.

[0054] The status profiling module extracts seawater parameters and biological behavior data from historical periods to construct historical status profiles and obtains real-time status profiles for the current period, achieving systematic integration and analysis of aquaculture process data. By comparing historical and real-time data, the changing patterns of the aquaculture environment and fish growth status can be clearly presented, providing comprehensive data references for subsequent anomaly detection and strategy adjustments. This avoids the one-sidedness of judgments caused by relying solely on real-time data, making the understanding of the aquaculture system status more continuous and holistic.

[0055] The stress response module calculates water quality deviation vectors and biological behavior deviation vectors based on historical and real-time status profiles, and identifies abnormal stress factors based on the water quality deviation vectors. This process overcomes the limitations of traditional stress responses that rely on only a single parameter. Through multi-dimensional deviation vector analysis, it can accurately locate the type and cause of abnormal stress factors, making stress response measures more targeted, effectively shortening the anomaly handling time, reducing the adverse effects of environmental anomalies on Atlantic salmon growth, and minimizing growth stagnation or health problems caused by stress responses.

[0056] The feeding optimization module correlates abnormal stress factors with feeding activity to generate feeding strategy adjustment instructions, achieving dynamic adaptation of feeding strategies to the aquaculture environment and fish condition. When abnormal stress factors affect fish feeding, the module can promptly adjust the feeding amount and frequency, avoiding feed waste under traditional fixed feeding patterns. It also reduces the digestive burden on fish caused by environmental discomfort, ensuring that fish receive appropriate nutritional support under different environmental conditions and maintain a stable growth rhythm.

[0057] The parameter decision module integrates fluid dynamic regions, stocking density thresholds, and feeding strategy adjustment commands, outputting LNG cold energy input control parameters and feeder control parameters, achieving coordinated optimization of all aspects of the aquaculture system. This module breaks the limitation of traditional systems where cold energy regulation and feeding control are independent, enabling precise matching of LNG cold energy input to the temperature requirements of each fluid dynamic region. Simultaneously, the feeder control parameters are adapted to stocking density and fish feeding status, forming a complete closed-loop aquaculture control system. This improves the overall operational efficiency and resource utilization efficiency of the aquaculture system, ensuring the growth of Atlantic salmon in a stable and suitable environment, and enhancing the quality and yield stability of farmed products.

[0058] Through the synergistic effect of its various modules, the entire system fully leverages the cold energy utilization advantages of the LNG-IFV system, precisely combining cold energy with aquaculture needs to reduce energy waste and align with the development trend of green aquaculture. Simultaneously, the system's refined control of the aquaculture process reduces the intensity of human intervention, minimizes errors caused by human operation, and improves the standardization and reliability of aquaculture management. This provides a feasible technical solution for the large-scale and intensive development of the Atlantic salmon farming industry, promoting the transformation of aquaculture models from traditional experience-based to modern precision-based. Attached Figure Description

[0059] Figure 1 This is a timing diagram of an Atlantic salmon farming system under an LNG-IFV system as described in this invention;

[0060] Figure 2 Workflow diagram for dividing the aquaculture zoning module into fluid dynamic regions;

[0061] Figure 3 Workflow diagram for identifying abnormal stress factors in the stress response module;

[0062] Figure 4 A flowchart illustrating the workflow for generating feeding strategy adjustment instructions for the feeding optimization module. Detailed Implementation

[0063] 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.

[0064] Please see Figure 1 This invention provides an Atlantic salmon farming system under an LNG-IFV system. The system includes: a water monitoring module, a farming zone module, a status profiling module, a stress response module, a feeding optimization module, and a parameter decision module. Specific implementation methods are as follows:

[0065] The water monitoring module collects real-time seawater parameters and Atlantic salmon biological behavior data after LNG-IFV heat exchange. The aquaculture zoning module divides fluid dynamic zones based on temperature gradients and flow velocity distribution, and sets aquaculture density threshold for each zone in conjunction with population distribution density. The status profiling module extracts historical seawater parameters and biological behavior data to construct historical status profiles and compares them with real-time status profiles. The stress response module calculates water quality deviation vectors and biological behavior deviation vectors to identify abnormal stress factors. The feeding optimization module correlates abnormal stress factors with feeding activity to generate feeding strategy adjustment instructions. The parameter decision module integrates fluid dynamic zones, aquaculture density thresholds, and feeding strategy adjustment instructions to output LNG cold energy input control parameters and feeder control parameters, achieving precise regulation.

[0066] Example 1: See Figure 2 When the aquaculture zoning module performs fluid dynamic zone division, temperature gradient data is collected by 16 sets of PT100 temperature sensors equidistantly arranged along the long axis of the aquaculture pond, with the sensor spacing being 1 / 8 of the pond width. High and low temperature transition zone identification uses a gradient abrupt change detection method: when the temperature difference between adjacent sensors exceeds 0.5℃ for three consecutive samples, the system automatically marks this interval as the boundary of the transition zone. Flow velocity distribution is simultaneously monitored by eight acoustic Doppler current profilers, updating the three-dimensional flow velocity field data every 30 seconds. Eddy generation zones are determined based on eddy current intensity; when the local velocity vector rotation angular velocity exceeds 0.2 rad / s and persists for more than 5 minutes, the system marks it as an eddy zone. Steady flow zones are defined as areas with a flow velocity fluctuation standard deviation of less than 0.08 m / s and continuous stability; the stability of these zones is verified in real time using a sliding window algorithm.

[0067] The fish population density monitoring employs a top-mounted wide-angle underwater camera system, with one 2-megapixel camera deployed per square meter, working in conjunction with the YOLOv5 target detection model to count fish populations in real time. Density variance is calculated on a 15-minute cycle, performed separately within each fluid dynamic zone defined by the system. The preset density tolerance is dynamically configured according to the rearing stage: ±25% density fluctuation is allowed during the juvenile stage, tightened to ±15% during the adult stage. If the density variance of a certain area exceeds the upper tolerance limit for six consecutive cycles, the system automatically lowers the rearing density threshold for that area, with the reduction calculated using the formula (current density - threshold) × 0.12, simultaneously triggering an audible and visual alarm for manual verification. If the density variance remains below the lower tolerance limit, the system adopts a step-by-step upward adjustment strategy: the initial increase is no more than 5%, and subsequent increases require verification of the stability of fish behavior after the previous adjustment.

[0068] In constructing historical state profiles, the state profile module selected seven consecutive aquaculture cycles (24 hours per cycle) as the analysis window. The generation of seawater parameter mean sequences employed a three-stage processing method: after Kalman filtering for noise reduction, the raw data was calculated as a 15-minute moving average, and finally, the arithmetic mean of 96 moving averages within the cycle was taken. The dissolved oxygen concentration mean was further processed with diurnal segmentation, aggregating daytime (06:00-18:00) and nighttime data separately. The peak sequences of biological behavior data were obtained using a time-series extreme value extraction algorithm. The peak of feeding activity was located at the maximum value within a 40-60 minute window after feeding, while the extreme value of population distribution density was taken as the high-density state value that lasted for more than 2 hours per cycle.

[0069] The water quality stability coefficient calculation integrates multiple dimensions of indicators: dissolved oxygen concentration fluctuation amplitude is taken as the relative difference between the maximum and minimum values ​​within the period; the slope of temperature gradient change is obtained through linear regression analysis; and flow velocity stability is characterized by the dominant frequency energy ratio after Fourier transform. After min-max standardization, the three indicators are weighted at 0.4:0.3:0.3 to synthesize the stability coefficient. When solving for the biological activity coefficient, a 120-dimensional covariance matrix of feeding activity and population distribution density is first constructed. Principal component analysis is used to extract the variance contribution rate of the first principal component as the activity characterization value. Historical state profiles are stored using a hierarchical data structure: the bottom layer is the original time-series database, the middle layer stores aggregated feature vectors, and the top layer encapsulates profile snapshots containing timestamps. With each new aquaculture cycle, the system automatically removes the earliest cycle data and recalculates the profile features, maintaining a 7-cycle rolling update mechanism.

[0070] In a specific implementation case, a fish farming pond measuring 30 meters long and 15 meters wide detected a significant temperature jump of 1.2℃ within a range of 8-12 meters from the inlet, as detected by a temperature sensor array. Simultaneously, flow velocity monitoring showed a clockwise rotating flow field in this area with a vortex intensity of 0.35 rad / s. Based on this, the system delineated a 4m × 4m vortex generation zone. The flow velocity in adjacent areas remained stable at 0.15 ± 0.03 m / s, classifying this as a steady flow zone. Population distribution monitoring revealed a fish density of 35 fish / m³ in the vortex zone, exceeding the baseline threshold of 28 fish / m³, and the density variance consistently exceeded the set tolerance. The system automatically lowered the stocking density threshold for this area to 25 fish / m³ and guided some fish to migrate to the steady flow zone using a diversion device.

[0071] When constructing the historical state profile, the system extracted data from the most recent 7 days: average dissolved oxygen concentration of 6.8 mg / L during the day and 6.2 mg / L at night; average temperature gradient slope of -0.15℃ / m; and flow velocity dominant frequency energy accounting for 82%. The calculated water quality stability coefficient was 0.78 (out of 1.0). In the biological behavior data, the covariance matrix of the peak feeding activity and the extreme population density showed a strong positive correlation, with the first principal component extracted by principal component analysis contributing 91%, and the biological activity coefficient was rated at 0.85. The final generated historical state profile was labeled as "stable state - high activity" and stored as a structured data package for subsequent comparison. When new periodic data is input, the system removes the oldest periodic data and recalculates to maintain the timeliness and continuity of the profile.

[0072] The process of adjusting the stocking density threshold demonstrates dynamic adaptability. In a stable flow zone of an adult fish pond, the initial density threshold was set at 30 fish / m³. Continuous monitoring revealed that the actual density stabilized in the range of 26-28 fish / m³, with the density variance consistently below the tolerance lower limit of 15%. The system adjusted the threshold in three steps: initially to 31 fish / m³, with a 72-hour validation period monitoring the evenness of fish distribution; the second adjustment to 32 fish / m³ resulted in a 5% decrease in feeding activity, so the adjustment was paused and the current threshold was maintained; seven days later, activity recovered, and the threshold was further increased to 33 fish / m³ and remained stable.

[0073] The status profile update mechanism embodies intelligent learning characteristics. While continuously updating historical status profiles, the system simultaneously analyzes the feature weights of data from removed periods. If a feature maintains a high weight (>0.7) for three consecutive removal periods, the system automatically increases its weight in the profile. For example, if the dissolved oxygen concentration in a certain aquaculture pond remains consistently significant at night, the system increases its weight in the stability coefficient calculation from 0.3 to 0.4, making the profile more accurately reflect the actual aquaculture environment characteristics. This dynamic weight adjustment mechanism enables historical status profiles to progressively optimize, gradually adapting to the personalized characteristics of the current aquaculture system.

[0074] Example 2: See Figure 3 In calculating the water quality deviation vector, the stress response module uses a dynamic weighted moving average method to determine the baseline dissolved oxygen concentration. The system extracts dissolved oxygen data from the past seven aquaculture cycles from historical data profiles, assigning weights according to the time decay coefficient: the most recent cycle has a weight of 0.25, the earliest cycle has a weight of 0.05, and intermediate cycles are distributed in an arithmetic progression. The baseline temperature fluctuation range is calculated using quantile regression, taking the 5th and 95th percentiles of historical temperature gradient data as the boundaries of the fluctuation range. The comparison between the current dissolved oxygen concentration and the baseline value uses a sliding window mechanism, calculating the mean concentration deviation within a 60-minute window every 5 minutes to avoid interference from instantaneous fluctuations.

[0075] When the system detects that the aquaculture pond is within 2 hours of feeding, it automatically relaxes the deviation judgment threshold to ±8%; during the nighttime period (20:00-06:00), due to the cessation of photosynthesis, the threshold tightens to ±4%. Temperature deviation analysis employs a dynamic time warping algorithm, which non-linearly aligns the current temperature fluctuation curve with the historical baseline curve, calculating the cumulative distance under the optimal matching path. This algorithm can effectively identify phase differences in temperature changes, such as delayed temperature peaks or abnormally long temperature plateaus. The synthesis of the water quality deviation vector uses a variable weighting strategy: during the juvenile stage (body length <15cm), the temperature weight is set to 0.6 and the dissolved oxygen weight to 0.4; during the adult stage (body length >30cm), the weights are adjusted to 0.3 for temperature and 0.7 for dissolved oxygen.

[0076] During the identification of abnormal stress factors, the dissolved oxygen compensation mechanism is activated using a graded response strategy. When the magnitude of the water quality deviation vector exceeds 120% of the first threshold, a first-level response is initiated: the LNG vaporization rate is adjusted to lower the seawater temperature by 0.5°C, and two auxiliary aerators are activated simultaneously. If the magnitude continues to exceed the limit for 30 minutes, the response is upgraded to a second-level response: an additional backup pure oxygen injection system is activated, and the vaporization rate adjustment range is increased to 1°C. The triggering conditions for the group behavior correction mechanism are set with dual verification: in addition to the rate of change of the biological behavior deviation vector exceeding the limit, the fish density in the area must also exceed the threshold of 110% for 10 consecutive minutes. Correction measures include adjusting the LED lighting spectrum (increasing the proportion of 450nm blue light) and activating directional flow pumps, with the flow intensity proportional to the rate of change of deviation.

[0077] In a specific implementation case, a fishpond experienced a sudden situation on a summer afternoon: the dissolved oxygen concentration plummeted from 6.5 mg / L to 5.2 mg / L within 40 minutes, while the baseline value was 6.3 ± 0.4 mg / L. The system calculated a dissolved oxygen deviation of -17.5% and a temperature deviation of 12%. The magnitude of the synthetic water quality deviation vector was 1.8 times the historical average, immediately triggering the primary dissolved oxygen compensation mechanism. The LNG vaporization rate increased by 15%, and both aerators operated at full power. After 30 minutes, the dissolved oxygen level recovered to 5.8 mg / L. Simultaneously, a 40% decrease in fish feeding activity was observed in the eddy current generation area, a change rate 2.3 times the historical standard deviation. The system activated the group behavior correction mechanism: increasing the lighting intensity by 30% and starting the No. 3 flow pump (flow rate 0.3 m / s). Two hours later, the system detected that all parameters had returned to normal ranges, and the compensation measures were gradually withdrawn.

[0078] The dynamic adjustment of the water quality deviation vector demonstrates the system's adaptability. In a fishpond during the initial introduction of a new fish species, the system recorded a unique pattern of consistently high temperature deviation but stable dissolved oxygen levels. After a 7-day learning period, the system automatically lowered the temperature weight from 0.6 to 0.4, while correspondingly increasing the dissolved oxygen weight. During winter operation, when the system detects that the influent temperature is consistently lower than the historical baseline, it initiates a switch in the temperature deviation calculation mode: changing from absolute temperature difference comparison to relative change rate comparison, thus avoiding frequent false alarms during the low-temperature season.

[0079] When identifying abnormal stress factors, the system traces the changes in water quality parameters and the evolution of fish behavior over the preceding six hours. For example, in one event, the system detected abnormal fluctuations in inflow rate 90 minutes before a sudden temperature drop and stored this information in a knowledge base for subsequent early warnings. The optimization of the group behavior correction mechanism is reflected in its spatial targeting: when fish are detected gathering in a corner of the pond, the system prioritizes adjusting the lighting and flow in that area, rather than making uniform adjustments across the entire pond.

[0080] During the diurnal cycle, the system implements differentiated stress response strategies. At dawn (04:00-06:00), the dissolved oxygen compensation response is delayed by 10 minutes to avoid conflict with the fish's natural awakening process. Around sunset (18:00-20:00), the intensity of group behavior correction is set to 70% of the daytime value to reduce interference with the fish's rest rhythm. This fine-tuning of the time sequence is achieved through a 24-hour biological clock model, with historical behavioral data and real-time observations as inputs.

[0081] When dissolved oxygen compensation and group behavior correction are activated simultaneously, the system calculates their cross-influence coefficient: the flow field changes caused by aeration will correspondingly reduce the power requirement of the flow pump, while light adjustment will temporarily delay some of the temperature compensation. In a typical case, the system detected the need to simultaneously increase dissolved oxygen by 0.8 mg / L and enhance fish activity. The final solution was: increase the aeration rate by 12% (originally 15%), increase the aerator operating power by 80% (originally 100%), and combine this with a blue light enhancement scheme to achieve a comprehensive intervention effect.

[0082] The application of historical data in stress response is reflected in pattern recognition. The system periodically analyzes the precursory characteristics of water quality parameter changes during historical anomalies and updates the early warning rule base. For example, it was found that before the three most recent major stress events, there was a continuous slight decrease in dissolved oxygen in the early morning (a decrease of more than 0.15 mg / L per hour), and the system added this pattern as a secondary early warning indicator. Long-term tracking and analysis of group behavior data is used to optimize and correct parameters, such as determining the optimal current-generating intensity curve for fish groups of different body lengths, to achieve precise behavioral guidance.

[0083] The system's emergency handling of emergencies demonstrates intelligent features. During a certain equipment failure, the inlet water temperature dropped suddenly by 3°C. The system completed the following actions within 30 seconds: marking the affected area as an emergency regulation area, automatically overriding the regular control logic; starting all standby heating units; adjusting the feeding plan to suspend subsequent feeding for the day; sending an alarm message to the management personnel containing the fault location and emergency measures. Throughout the process, the system continuously monitors the stress performance of the fish population, evaluates the effectiveness of the measures every 5 minutes, and dynamically adjusts the response plan.

[0084] Through long-term data accumulation, the system establishes a dissolved oxygen demand model under different fish population densities. For example, it is found that the dissolved oxygen consumption rate in the high-density area (>35 fish / m 3 ) within 2 hours after feeding is 40% faster than that in the low-density area, and the monitoring frequency and compensation threshold of these areas are adjusted accordingly. Individual differences in temperature adaptation are also taken into account: the system records the temperature adaptation curves of each batch of fry after they enter the pond, which are used to set the temperature deviation alarm threshold for personalization.

[0085] Example 3: Refer to Figure 4 , when the feeding optimization module performs the feeding response curve matching, the system maintains a stress scenario database containing multi-dimensional features. This database stores typical stress patterns classified by combinations of season, farming stage, and water quality parameters, as well as the feeding attenuation characteristics corresponding to each pattern. When an abnormal stress factor is identified, the system first extracts its spatio-temporal features: including the occurrence time period (dawn / noon / dusk), duration, influence range (local / global), and the change rate of water quality parameters. These features are matched with the records in the database through the nearest neighbor algorithm to find the 3 most similar historical scenarios, and the weighted average of their feeding response characteristics is taken as the current attenuation coefficient.

[0086] The calculation of the metabolic rate weight considers the temperature non-uniformity in the fluid dynamic area. The system divides the breeding pond into grids of 0.5m×0.5m, and the temperature data of each grid point is used to generate a continuous temperature field through the Kriging spatial interpolation algorithm. The metabolic rate weight W m is calculated by the following formula:

[0087]

[0088] where: T represents the current temperature of the grid point, T0 is the standard metabolic temperature for this farming stage (set to 12°C for the juvenile fish stage and 16°C for the adult fish stage), and α is the adjustment coefficient (0.25 in summer and 0.15 in winter). The system calculates the weight mean for each dynamic area, and the weight in the high-temperature area (T>T0+2°C) automatically increases by 10%, while the weight in the low-temperature area (T<T0-2°C) decreases by 15%.

[0089] The basal feeding amount is derived from the standard formula for the rearing cycle, calculated based on the total biomass of the fish population at a ratio of 1.2%-3.5%. The first-level correction applies the feeding attenuation coefficient to adjust the basal value globally; the second-level correction is based on metabolic rate weighting, implementing regionally differentiated allocation. Adjustments to feeding frequency follow the principle of minimum disturbance: when the attenuation coefficient is in the range of 0.9-1.1, the original frequency is maintained; outside this range, each adjustment should not exceed ±1 times / day, and the change in feeding amount per instance should not exceed 20%.

[0090] In a specific implementation case, a fishpond experienced continuous rainy weather in spring, and the system detected increased water temperature fluctuations and decreased dissolved oxygen. The abnormal stress factor was categorized as "low temperature and low oxygen complex," and three similar scenarios in the database were matched (with matching degrees of 78%, 65%, and 72%, respectively), ultimately determining the feed attenuation coefficient to be 0.82. When calculating the metabolic rate weight, the temperature distribution within the pond showed a significant difference between the low-temperature zone in the northwest corner (14.2℃) and the high-temperature zone in the southeast corner (17.8℃), with corresponding weights of 0.87 and 1.13, respectively. The system adjusted the original feeding plan of 4 times a day, 120kg each time, to: 5 times a day, 85kg each time in the low-temperature zone; 4 times a day, 105kg each time in the high-temperature zone; and a gradient distribution in the transition zone.

[0091] The spatiotemporal optimization of the feeding strategy is reflected in precise delivery. The system plans feeding points based on fluid dynamics: the main feeding point is set 1.5m upstream of the eddy generation zone, utilizing the eddy diffusion effect; in the steady flow zone, equidistant feeding points are used (one every 3m). The feeding sequence is synchronized with the natural gathering rhythm of the fish: the first feeding begins 1 hour after sunrise, and subsequent feeding intervals are dynamically adjusted according to the decay curve of the previous feeding activity. When the rate of decrease in feeding activity in a certain area is detected to be more than 20% faster than the historical average, the next feeding interval in that area is automatically shortened by 10-15 minutes.

[0092] Once the water quality deviation vector magnitude falls below the threshold, the system initiates a 3-day recovery monitoring period: the first day maintains 80% of the adjusted feeding amount, the second day restores it to 90%, and the third day determines whether to fully return to the standard value based on the recovery of feeding activity. In a typical case, after the fish experienced 48 hours of low-temperature stress, the system detected that feeding activity had only recovered to 85% of normal levels. Therefore, the full recovery period was extended to 5 days, with the daily increment adjusted to 5%.

[0093] The system pre-adjusts the water flow velocity in relevant areas 30 minutes before each feeding: the flow rate is increased by 0.05 m / s in the steady flow zone to promote feed diffusion, and decreased by 0.1 m / s in the eddy flow zone to prevent feed from being lost too quickly. Drastic water temperature adjustments are paused for one hour after feeding to maintain environmental stability and facilitate digestion. During the high-temperature period in summer, a special "dawn feeding" mode is set up: the first feeding is moved forward to 30 minutes before sunrise to avoid the peak metabolic stress caused by the afternoon heat.

[0094] The system automatically records the feeding recovery curves of fish populations after each stress event, forming a species-specific recovery characteristic database. For example, it was found that the feeding activity recovery of a certain strain of Atlantic salmon exhibited an S-shaped curve after low-temperature stress, and the system adjusted the feeding recovery strategy for that strain accordingly. Long-term data also revealed differences in feeding rhythms among different fish sizes: fish with a body length of 20-30cm reached their feeding peak 45 minutes after feeding, while the peak for individuals over 30cm was delayed until 60 minutes. These findings were used to optimize the feeding sequence.

[0095] After each feeding, the system compares the expected feeding activity with the actual observed value. When the deviation exceeds 15% for three consecutive times, parameter calibration is triggered: either the weighting parameters of the feeding attenuation coefficient calculation formula are updated, or the baseline temperature value in the metabolic rate model is adjusted. During winter operation, the system found that the actual feed intake was consistently lower than the predicted value. After analysis, a seasonal correction factor of 0.9 was added to the metabolic rate weighting formula in the low-temperature zone.

[0096] During the initial acclimatization period (first 72 hours) of the fish fry, the system employs a "micro-volume, high-frequency" feeding mode: feeding 0.2% of the biomass of feed per hour, while simultaneously increasing the frequency of feeding activity monitoring to once every 15 minutes. During the breeding preparation period, a "nutritional fortification" program is activated: in addition to regular feeding, two supplementary feedings rich in specific amino acids are added daily, with the timing of these feedings automatically determined based on the gonadal development index. These special strategies are stored in an independent knowledge base and are automatically invoked when a change in the rearing stage marker is detected.

[0097] The system cross-analyzes feeding effects with recent medication records, changes in the light cycle, and other relevant factors to identify potential influencing factors. For example, if an abnormal feeding effect is found to be correlated with parasite control treatment performed the previous day, the system automatically adds this medication information to the stress factor database, allowing for pre-emptive adjustments to the feeding plan when similar situations arise. Light data is used to optimize the selection of feeding times, with special consideration given to the promoting effect of feeding during periods of increasing light intensity in artificial lighting systems.

[0098] Example 4: The parameter decision module employs a hierarchical and zoned control strategy when regulating LNG cold energy input. The system divides the aquaculture pond into multiple control units, each corresponding to an independent heat exchange efficiency calculation unit. The target heat exchange efficiency is determined based on a real-time temperature distribution map, which marks the difference between the current temperature and the target temperature in each area. The temperature regulation priority setting considers not only the stocking density but also the fish activity index, which is calculated by analyzing the displacement distance and directional change rate of the fish population per unit time. The calculation process for the LNG vaporization rate increment includes three steps: first, calculating the heat load demand of each control unit; second, summarizing the total demand of the entire pond; and finally, allocating the increment based on the current operating status of the LNG vaporization unit.

[0099] In a specific implementation case, a rectangular aquaculture pond (40m long × 20m wide) was divided into 32 control units. The system detected that the temperature in the six units in the northwest corner was 1.5℃ lower than the target value, and the stocking density in these units was 35 fish / m³. 3 The fish activity index was 0.85 (normal range 0.7-1.0), and the area was marked as a first-priority zone. The temperature in the four units on the southeast side was 0.8℃ higher than normal, with a density of 28 fish / m³. 3 The activity index is 0.65, classifying it as a secondary control zone. System calculations indicate that the primary priority zone requires an additional 12kW of cold energy input, while the secondary control zone requires a reduction of 8kW. The LNG vaporization unit is currently operating at 75% load, with a remaining regulation capacity of 30kW. The system allocates 15kW to the priority zone and 5kW to optimize the uniformity of other areas.

[0100] The regional feeding amount correction value comes from the output of the feeding optimization module, which the system converts into the feeding weight for each control unit. The coordinates of the eddy generation zone are determined using particle tracking technology. The system releases neutral buoyancy tracer particles in the water, and their trajectories are recorded by a high-speed camera. The calculation of the feeding point diffusion radius considers the eddy intensity and the feed settling velocity, with a typical value of 1.2-2.5m. The feeder's movement trajectory is generated using piecewise linear interpolation, employing a spiral approximation path in the eddy zone and a straight reciprocating path in the steady flow zone. The feeding sequence is synchronized with the water flow velocity field, starting feeding 10 minutes before the peak velocity, utilizing the water flow diffusion effect. The parameter decision results for five typical control units during a specific control process are shown in Table 1. The table contains key information such as unit location, temperature deviation, control priority, cold energy increment allocation, and feeding parameter adjustments.

[0101] Table 1: Decision table for control unit regulation parameters.

[0102]

[0103] For units where the cold energy input changes by more than 10%, adjustments are made in three tiers: an initial 50% increment, followed by another 30% after 30 minutes of evaluation, and the final 20% adjustment is completed while ensuring system stability. The feeder's movement speed is dynamically adjusted according to the diffusion radius: 0.8 m / s for areas with a radius of less than 1.5 m, 0.6 m / s for areas with a radius of 1.5-2.0 m, and 0.4 m / s for areas with a radius of more than 2.0 m. This speed matching ensures that the bait is evenly distributed in the target area.

[0104] The system analyzes temperature variation patterns during the same time period over the past 7 days to predict temperature trends over the next 2 hours. For example, if it detects a natural temperature rise of 0.3℃ in a certain area between 14:00 and 15:00 daily, the system will fine-tune the cold energy input 10 minutes in advance to compensate. Feeding records are used to optimize feeding point selection; the system marks the coordinates of the points with the highest feeding activity over the past 3 days and adds temporary feeding points in nearby locations.

[0105] A sudden power outage caused the main gasification unit to shut down. Within 90 seconds of the backup power being activated, the system made emergency decisions: marking the 10 most sensitive control units as life support zones and concentrating limited cooling energy supply; suspending all non-essential feeding; and adjusting the distribution of remaining feed to the core area. After power was restored, the system adopted a slow recovery strategy, restoring 20% ​​of the cooling energy supply every hour to avoid drastic temperature fluctuations.

[0106] During winter operation, the system activates low-temperature protection logic: when the inlet water temperature is below 4℃, it automatically increases the cold energy input at the inlet area by 5% to create a temperature buffer zone. In summer, a high-temperature emergency plan is activated: on days when the forecast temperature exceeds 30℃, the overall temperature of the aquaculture pond is lowered by 0.5℃ two hours in advance to alleviate heat stress in the fish. These seasonal strategies are stored in independent configuration files and automatically switch based on meteorological data.

[0107] During cold energy regulation, the system monitors parameters such as the gasification unit outlet pressure and seawater pump flow rate in real time to ensure the equipment operates within a safe range. When the feeder executes a movement command, it synchronously adjusts the matching relationship between the throwing motor speed and the traveling speed to ensure a constant feeding amount per unit area. When a response delay exceeding 5 seconds is detected, the system automatically reduces the adjustment range by 50% and issues a maintenance alarm.

[0108] The control interface displays a heat map of the entire pond's temperature distribution, using different colors to indicate the priority of each control unit. The feeder's movement trajectory is projected in real-time onto the plan view, marking completed and pending feeding points. Historical curves simultaneously display the trends in cold energy input and fish activity index, helping operators understand the system's decision-making logic. In one commissioning case, engineers observed that the activity index in a certain area was consistently low. They manually added additional water quality monitoring points and ultimately discovered a minor electrical leakage affecting fish behavior.

[0109] During the low-load period in the early morning each day, the system automatically performs self-checks: the feeder completes positioning accuracy calibration, the aeration unit performs valve response tests, and the sensor network performs cross-validation. Minor deviations detected (such as a 2mm deviation in the feeder track) are automatically compensated for, without affecting normal operation for the day. When the cumulative wear exceeds the maintenance threshold (such as 80% wear on the aeration valve), the system generates a preventative maintenance plan 72 hours in advance.

[0110] Example 5: The data backtracking module's operating mechanism is based on a dynamic storage management strategy. The system continuously monitors the changing characteristics of the water quality deviation vector, including fluctuation amplitude, duration, and spatial distribution pattern. The initial backtracking period is set to 7 days, a baseline value determined based on the typical change cycle of aquaculture water environmental parameters. The module analyzes the frequency of deviation vector changes to identify the system's stable state or abnormal fluctuation patterns, and adjusts the data retention period accordingly. When high-frequency abnormal fluctuations are detected, the backtracking period is automatically shortened to retain more detailed recent data; when the system is operating smoothly, the period is appropriately extended to optimize storage resources.

[0111] The system calculates the modulus range within a 60-minute window every 15 minutes, recording the number of times it exceeds a set threshold. The warning frequency is calculated based on a moving average, taking the weighted sum of the number of exceedances within the most recent 24 hours, with the weight of data from the last 6 hours at 0.6, and earlier data weighted exponentially. The cumulative frequency is determined using a dual condition: both the absolute value threshold must be reached (e.g., 8 times per day during the adult fish stage) and a relative growth requirement must be met (an increase of more than 50% compared to the average of the previous three days). This design avoids false triggers caused by seasonal or periodic changes.

[0112] The system assigns multiple tags to each data record in the historical status profile, including data type (water quality / biological behavior), collection period (peak / valley), and associated events (stress / normal). During the cleanup decision-making process, firstly, basic monitoring data exceeding the storage timeframe is removed, while event-related data from the same period is retained; secondly, duplicate or highly similar aggregated feature data is cleaned up, retaining the original key node records; finally, the data value density is assessed, and transitional status records with low information content are deleted. The entire cleanup process maintains a complete metadata index, ensuring that the data source can be traced when needed later.

[0113] In a specific implementation case, a fishpond experienced five consecutive days of rainy weather, and the system detected frequent exceedances of the water quality deviation vector magnitude. The cumulative frequency reached the warning value on the third day, triggering a shortened backtracking period mechanism: gradually reducing it from seven days to three days. Detailed data retained during this period helped the system identify the coupling change pattern between influent turbidity and dissolved oxygen, updating the stress factor identification rules. After the weather cleared, the system showed no significant deviations for seven consecutive cycles, automatically extending the backtracking period to ten days and cleaning up some intermediate calculation data from the rainfall period, while retaining all key event node records.

[0114] The basic timeliness settings vary according to the aquaculture stage: 15 days for the fry stage (0-3 months), 30 days for the rapid growth stage (3-12 months), and 45 days for the adult stage (over 12 months). Based on this, the system monitors the remaining storage capacity in real time and initiates automatic cleanup when available space falls below 20%, prioritizing the retention of core data from the most recent three aquaculture cycles. Data marked with special events (such as equipment failure or extreme weather) is not subject to timeliness restrictions and is permanently stored in a separate storage area.

[0115] The system projects feature vectors from historical state profiles into a high-dimensional space and uses density clustering to identify highly clustered data points. For each cluster, the most representative center point record is retained, while edge similar points are removed. For biological behavior data, the system pays special attention to time series continuity, avoiding the deletion of key turning points. In one case, the system clustered 5760 water quality monitoring records collected over 24 consecutive hours into 48 groups, retaining the record with the largest standard deviation within each group as the representative sample.

[0116] Each periodic change is limited to no more than 30% of the current value to avoid drastic changes in storage strategy that could impact system stability. When the period needs to be shortened, the system initiates data pre-screening 6 hours in advance, marking candidate data to be cleaned; when the period needs to be extended, the data retention criteria are gradually relaxed, and adjustments are confirmed only after observing the system's operational status. This gradual adjustment approach effectively prevents the accidental loss of important data.

[0117] When the water quality deviation vector magnitude reaches an extreme peak (exceeding 80% of the historical maximum), the system automatically creates an event snapshot, containing complete raw data for two hours before and after the event, all intermediate calculation results, and relevant equipment status logs. This snapshot data is stored in an independently compressed format and generates a detailed event analysis report. In subsequent operation, the system will periodically retrieve similar event patterns to optimize early warning threshold settings.

[0118] The system performs deep data mining monthly to extract representative patterns from massive historical records. For example, by analyzing three years of winter operating data, it was found that fish populations become 20% more sensitive to dissolved oxygen fluctuations when the water temperature is below 8°C, and the monitoring frequency during low-temperature periods was adjusted accordingly. This knowledge is condensed into configuration parameter updates, and the raw data is appropriately cleaned after retaining key features.

[0119] When the stress response module flags new anomalous factors, the data backtracking module automatically retrieves historical records to find handling experiences for similar events. The feeding optimization module can query feeding response curves for the same area over the past 30 days to improve prediction accuracy. This cross-module data linkage is achieved through a unified timestamp index, ensuring consistency of spatiotemporal information.

[0120] Breeding administrators can manually mark important data segments to exclude them from automatic cleanup; they can also temporarily adjust the backtracking cycle to meet special needs, such as setting it to full recording during scientific research observations. All manual interventions are recorded in the audit log to prevent accidental operations from affecting the system's automatic management logic. During critical periods such as the breeding season, the system will automatically upgrade the data retention level, which is equivalent to setting virtual markers.

[0121] Raw data collected at high frequencies is stored in a high-speed cache with a short retention period; aggregated data after feature extraction is stored in the main database with a medium retention period; and pattern data after knowledge extraction is written to long-term storage for permanent preservation. Data pipelines are set up between each layer to periodically pass important information to the next lower level. This structure optimizes storage efficiency while ensuring data availability.

[0122] The system performs incremental backups of core data every day at midnight, with the backup retention period being twice that of the primary storage. When a hardware failure risk is detected, an emergency full backup is automatically triggered. All backup data is also subject to a rollback period; backups that have exceeded their expiration date are automatically deleted. In one server failure case, the system fully restored its operational status using incremental backups from the most recent three days and snapshots of critical events.

[0123] The timeline browser can display data change curves for any time period and supports multi-parameter overlay comparison. The heatmap tool marks high-frequency abnormal areas to help locate potential problem points. Administrators can use these tools to intuitively understand the system's operational history, verify the rationality of automatic cleanup decisions, and make manual corrections when necessary.

[0124] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0125] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An Atlantic salmon farming system under an LNG-IFV system, characterized in that, The Atlantic salmon farming system includes: The water monitoring module is used to collect seawater parameters after LNG-IFV heat exchange in real time. The seawater parameters include temperature gradient, dissolved oxygen concentration, and flow velocity distribution. It also monitors the biological behavior data of Atlantic salmon, including feeding activity and population density. The aquaculture zoning module is used to divide the fluid dynamic zone of the aquaculture pond based on the temperature gradient and the flow velocity distribution, and to set the aquaculture density threshold for each dynamic zone in combination with the population distribution density. The status profile module is used to extract the seawater parameters and biological behavior data in the historical period, construct the historical status profile, and obtain the real-time status profile of the current period. The stress response module is used to calculate water quality deviation vector and biological behavior deviation vector based on the historical state profile and the real-time state profile, and to identify abnormal stress factors based on the water quality deviation vector. The feeding optimization module is used to associate the abnormal stress factors with the feeding activity level and generate feeding strategy adjustment instructions. The parameter decision module is used to integrate the fluid dynamic region, the breeding density threshold and the feeding strategy adjustment command, and output LNG cold energy input regulation parameters and feeder control parameters. The parameter decision module outputs the following LNG cold energy input control parameters: Based on the temperature distribution of the fluid dynamic region, the target heat transfer efficiency of each region is calculated. Based on the aforementioned stocking density threshold, temperature control priority is set; By combining the target heat exchange efficiency and the temperature control priority, the LNG vaporization rate increment is calculated. The parameter decision module outputs the following control parameters for the feeder: Decompose the area feeding amount correction value in the feeding strategy adjustment instruction; Calculate the diffusion radius of the feeding point based on the coordinates of the eddy generation zone in the fluid dynamic region; Based on the diffusion radius and the regional feeding amount correction value, the feeder movement trajectory and feeding sequence are generated.

2. The Atlantic salmon farming system under an LNG-IFV system as described in claim 1, characterized in that, The aquaculture zoning module divides the fluid dynamic region as follows: Based on the temperature gradient, the high and low temperature transition zone of the aquaculture pond is located, and combined with the flow velocity distribution, the eddy generation zone and the steady flow zone are marked. Calculate the density variance of each region based on the population distribution density; The aquaculture density threshold is dynamically adjusted based on the density variance and the preset density tolerance.

3. The Atlantic salmon farming system under an LNG-IFV system as described in claim 1, characterized in that, The status profiling module constructs the historical status profile by including: Extract the mean sequence of the seawater parameters and the peak sequence of the biological behavior data for N consecutive aquaculture cycles; Based on the mean and peak sequences, the water quality stability coefficient and biological activity coefficient for each period are fitted. By associating the water quality stability coefficient with the biological activity coefficient, the historical state profile is generated.

4. The Atlantic salmon farming system under an LNG-IFV system as described in claim 1, characterized in that, The stress response module calculates the water quality deviation vector including: Decompose the baseline dissolved oxygen concentration and baseline temperature fluctuation range in the historical state profile; The dissolved oxygen concentration is compared with the baseline dissolved oxygen concentration in the real-time status image to obtain the dissolved oxygen deviation value; Analyze the deviation of the current temperature fluctuation range from the reference temperature fluctuation range; The dissolved oxygen deviation value and the deviation degree are combined to construct the water quality deviation vector.

5. The Atlantic salmon farming system under an LNG-IFV system as described in claim 4, characterized in that, The stress response module identifies the abnormal stress factors including: When the magnitude of the water quality deviation vector exceeds the first threshold, the dissolved oxygen compensation mechanism is activated; When the rate of change of the biological behavior deviation vector exceeds the second threshold, the group behavior correction mechanism is activated. The dissolved oxygen compensation mechanism and the group behavior correction mechanism are correlated, and the abnormal stress factors are labeled.

6. The Atlantic salmon farming system under an LNG-IFV system as described in claim 1, characterized in that, The feeding optimization module generates the feeding strategy adjustment instructions including: Map the abnormal stress factors to a feeding response curve library and output the feeding attenuation coefficient; The metabolic rate weights of each region are calculated based on the temperature distribution of the fluid dynamic region. The feeding amount per unit time is dynamically adjusted by combining the feed attenuation coefficient and the metabolic rate weight.

7. The Atlantic salmon farming system under an LNG-IFV system as described in claim 1, characterized in that, The system also includes a data backtracking module; The data backtracking module is used to set the data backtracking period based on the change frequency of the water quality deviation vector; Based on the data backtracking cycle, redundant data in the historical status profile that has exceeded the storage time limit is cleaned up.

8. An Atlantic salmon farming system under an LNG-IFV system as described in claim 7, characterized in that, The data backtracking module sets the data backtracking period as follows: The cumulative frequency of the magnitude of the water quality deviation vector exceeding the third threshold is counted. When the cumulative frequency reaches the warning frequency, the data backtracking cycle is shortened. If the biological behavior deviation vector fails to trigger the correction mechanism for M consecutive cycles, the data backtracking cycle is extended.

Images