Intelligent hierarchical precision feeding method based on physiological metabolism energy requirement model of sea bass

Abstract

The application discloses an intelligent layered precision feeding method based on a physiological metabolism energy demand model of sea bass, and comprises the following steps: step one, online estimation of instantaneous total metabolism energy demand of a sea bass group; step two, dynamic partition of a pond based on feeding competition intensity; step three, generation and execution of a layered dynamic feeding strategy; and step four, optimization of model parameters based on individual growth homogeneity feedback. Through direct and real-time online estimation of instantaneous total metabolism energy demand of the fish group, real-time perception and layered satisfaction of real physiological demands of the fish group are realized. The precise perception can effectively reduce the feed conversion ratio, alleviate uneven growth caused by feeding competition, and significantly improve the breeding benefit.

Description

Technical Field

[0001] This invention belongs to the field of smart aquaculture, specifically involving a pond-based precise feeding decision-making and execution method based on the real-time physiological metabolic energy demand calculation of perch and coupled with water quality micro-zone dynamics and feeding competition behavior. Background Technology

[0002] The core challenge in precision feeding for sea bass lies in the static nature of supply versus the dynamic nature of demand. Existing technologies largely rely on "historical growth models + environmental factor correction" to predict current demand, essentially an extrapolation method that cannot detect in real-time metabolic fluctuations caused by stress, disease, or group competition. Furthermore, existing feeding logic is often unidirectional, with adjustments typically based on post-event feedback of uneaten feed levels, resulting in long feedback cycles and an inability to distinguish the causes of uneaten feed. Simultaneously, existing technologies treat the fish population as uniformly distributed, ignoring the sea bass's strong competitive feeding behavior, leading to chronic underfeeding in highly competitive areas for weaker individuals and causing severe growth differentiation. This invention aims to overturn the "feed-centric" feeding logic and establish a new paradigm of "fish-centric, real-time sensing, and tiered satisfaction," fundamentally solving the aforementioned problems. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies, this invention emphasizes a shift in logic from "feed-centric" to "fish-centric." Existing feeding techniques suffer from drawbacks such as delayed demand estimation, unidirectional feeding logic, and neglect of competitive behavior within fish schools. This results in an inability to respond in real-time to instantaneous metabolic changes in fish caused by stress or competition, and easily leads to insufficient feeding in weaker individuals. This invention provides an intelligent stratified precision feeding method based on a sea bass physiological metabolic energy demand model. By directly and in real-time estimating the instantaneous total metabolic energy demand of the fish school online, it achieves real-time perception and stratified fulfillment of the fish's true physiological needs. This precise perception can effectively reduce the feed conversion ratio, alleviate uneven growth caused by feeding competition, and significantly improve aquaculture efficiency.

[0004] The technical solution of this invention is: an intelligent stratified precision feeding method based on a bass physiological metabolic energy demand model, comprising the following steps:

[0005] Step 1: Online estimation of instantaneous total metabolic energy demand of the bass population;

[0006] (1) Real-time acquisition of dissolved oxygen consumption rate and fish movement data in the pond to be fed, establishment of physiological metabolic model, real-time estimation of instantaneous metabolic energy demand of fish in the current environment, and calculation of basal respiratory metabolic energy demand. :

[0007] ;

[0008] In the formula: The oxygen consumption rate of pure fish after removing the losses in the blank control group; The effective water volume of the pond; The oxygen thermal equivalent coefficient; Based on real-time water temperature Dynamic correction is performed, and the correction formula is as follows: ;

[0009] (2) Estimating the metabolic energy demand of activities by linking acoustic Doppler current profiler and camera network image recognition :

[0010] ;

[0011] In the formula: , It is the coefficient of variation of local water flow velocity monitored using an acoustic Doppler current profiler. For calibration coefficients; The average swimming speed of the fish; and To correct the parameters; The number of fish;

[0012] (3) The sum of the two gives the instantaneous total metabolic energy requirement. :

[0013] ;

[0014] Step 2: Identify dynamic pond zoning based on feeding competition intensity;

[0015] Before feeding, a feeding detection exercise was performed to track the feeding behavior distribution of the fish in real time and to construct a feeding competition intensity model. Based on this, the pond is divided into several hierarchical areas with differentiated competitive capabilities;

[0016] Step 3: Generation and execution of hierarchical dynamic feeding strategy;

[0017] Based on the instantaneous total metabolic energy requirement Growth energy required for the target daily growth rate Determine the total amount of feed The system implements a differentiated feeding mode based on the distribution weight of the hierarchical regions and makes real-time dynamic adjustments based on the feeding images during the feeding process.

[0018] Step 4: Optimize model parameters based on feedback of uniform individual growth;

[0019] Monitor feeding effects and calculate the coefficient of variation of population body weight. ,based on Closed-loop correction of the feeding weight allocation coefficient and competition level division threshold in subsequent cycles.

[0020] Furthermore, in step two, the competition intensity model... The construction logic is as follows: use object detection algorithms to identify the center coordinates of the fish school and calculate the local fish density. Compared with the average individual distance Combined with the frequency of food-grabbing behavior per unit time Overall assessment.

[0021] Furthermore, in step two, the partitioning level adopts an adaptive threshold division: high-competition region threshold. =Mean CI of the entire pool + 1.2 standard deviations; Threshold for non-feeding areas =Mean CI of the entire pool - 0.8 times the standard deviation.

[0022] Furthermore, in step three, the total amount of feed... The calculation formula is: In the formula: To determine the growth energy required based on the target daily growth rate, For feed energy conversion efficiency, This refers to the energy value per unit weight of the feed used.

[0023] Furthermore, in step three, the differentiated feeding model includes: adopting a "high single-volume, precise point-to-point feeding" model in high-competition areas, and introducing a compensation coefficient. Adjust quotas and feed volume in highly competitive areas The calculation formula is: , Weights are assigned to high-competition areas; a standard uniform surface projection mode is used for medium / low-competition areas.

[0024] Furthermore, in step three, the adaptive adjustment logic of the feeding process is as follows: real-time monitoring of the feeding competition images in each zone; if the growth rate of the clustered area in a certain zone decreases... If the quota for that zone is reduced, the saved bait will be added in real time to the zones where the feeding intensity is still high.

[0025] Furthermore, in step four, the coefficient of variation of population body weight The formula for estimating the linkage is: In the formula: For correction factor, and These represent the average competition intensity in the high-competition and low-competition zones, respectively.

[0026] Furthermore, it also includes a parameter closed-loop optimization process: the system automatically updates once every set threshold number of days (3 days). If the value shows an increasing trend, the weights allocated to high-competition regions will be automatically increased using a reinforcement learning algorithm. and compensation coefficient This enables precise allocation of resources.

[0027] Furthermore, it also includes emergency treatment for stress responses in bass under extreme conditions: if a sudden increase in water flow velocity is detected... or water body If the value deviates from the range of 7.0-8.5, immediately suspend feeding and resume feeding at 50% of the initial quota after a set threshold time (5 minutes).

[0028] The beneficial effects of this invention are: it provides an intelligent stratified and precise feeding method based on a model of the physiological metabolic energy requirements of sea bass.

[0029] Demand perception: The system shifts from indirect model prediction to direct online measurement of fish metabolism (oxygen consumption rate + activity), enabling a "direct reading" of the fish's real and instantaneous physiological needs with extremely fast response speed.

[0030] Feeding logic: It pioneers the "explore before feeding, tiered treatment" model. By pre-feeding attractants, it dynamically detects the distribution and competitive landscape of fish schools, and then implements differentiated feeding strategies accordingly, fundamentally alleviating uneven growth caused by feeding competition.

[0031] Adaptive process control: During the feeding process, the feeding amount in each zone is dynamically adjusted in real time based on the images of food competition, achieving "second-level" feedback and optimization, which surpasses the traditional long cycle mode of "feeding-feeding platform observation-next adjustment".

[0032] Optimization goal: To achieve uniform population growth ( ) Instead of simply using total output or feed conversion ratio as long-term optimization goals, it better meets the comprehensive needs of healthy farming and improved economic benefits, and continuously improves farming results through closed-loop optimization. Attached Figure Description

[0033] Figure 1 : A schematic diagram of the system architecture and data flow of this invention;

[0034] Figure 2 Line chart showing the dynamic changes in the competition intensity index and partition weights;

[0035] Figure 3 Flowchart of a stratified precision feeding method based on a physiological and metabolic demand model of bass;

[0036] Figure 4 : Schematic diagram of parameter closed-loop optimization based on uniform population growth. Detailed Implementation

[0037] The present invention will now be further described with reference to the accompanying drawings.

[0038] like Figure 1 , 3As shown, an intelligent stratified precision feeding method based on a sea bass physiological metabolic energy demand model includes the following steps:

[0039] Step 1: Online estimation of instantaneous metabolic energy demand of bass population

[0040] Respiratory metabolic energy estimation: Three fixed-point high-precision dissolved oxygen sensors were deployed in representative waters of the pond (located in the upwind, midwind, and downwind areas, respectively). A blank control group (enclosed by a net, with a volume of 1% of the total pond volume) was also established. By measuring the dissolved oxygen consumption rate in the water, subtracting the dissolved oxygen consumption rate of the blank control group (to offset interference from plankton, uneaten food, etc.), and combining this with the pond water volume, the instantaneous oxygen consumption rate of the entire fish population was inferred. Based on the oxygen calorific value of perch (adapted to dynamic water temperature correction), the oxygen consumption rate was converted into the basal respiratory metabolic energy requirement of the entire population. The calculation formula is:

[0041]

[0042] Wherein, ΔO2 is the decrease in dissolved oxygen concentration per unit time. (Take the average value of 3 fixed-point sensors) The effective water volume of the pond (m³) The oxygen-heat equivalent coefficient for sea bass was derived from respiratory metabolism experimental data and verified under multiple water temperature conditions. Thirty sea bass of three sizes (50g, 200g, and 500g) were selected for the experiment. Under five water temperature gradients (18℃, 22℃, 25℃, 28℃, and 32℃), the oxygen consumption and heat production of the sea bass in a closed breathing chamber were measured. A linear regression analysis was used to establish a model relating water temperature and oxygen-heat equivalent coefficient. (Where, the coefficient of determination R² = 0.97, and the sample size n = 450). The estimation error of this formula is ≤3% within the range of 18-32℃ (the suitable water temperature range for sea bass growth), where T is the real-time water temperature (℃), which is synchronously collected by a dissolved oxygen sensor.

[0043] Activity-based metabolic energy correction: Coefficient of variation of local water flow velocity monitored using acoustic Doppler current profiler (ADCP) Indirectly derive the average swimming speed of the bass population The formula is (k is the calibration coefficient, with a default value of 0.03, which can be calibrated through one initial sampling). The fish population N is estimated by linking ADCP flow velocity data with camera network image recognition: ( The average fish density of the entire pond is output in real time by network image recognition from surface and underwater cameras. The average volume of a single fish is used, dynamically updated based on the initial stocking size and growth model. The metabolic energy requirement due to increased activity is calculated using a fish swimming metabolic model. :

[0044]

[0045] in, and The default parameters for the model related to sea bass weight are (a=0.08, suitable for sea bass of 50-500g size; b=2.3). Users can automatically correct these parameters by inputting the current average weight of the sea bass, without the need for additional experimental calibration.

[0046] Total metabolizable energy requirement: The instantaneous total metabolizable energy requirement of the fish population under the current environment. for:

[0047]

[0048] This requirement is the lower limit of energy that needs to be obtained from food to maintain current life activities.

[0049] Step 2: Dynamic pond zoning based on feeding competition intensity identification

[0050] Feeding Hotspot Detection: Ten minutes before planned feeding, 0.5g of ordinary attractant bait (no marking required) is automatically added to each of 5-8 pre-set fixed detection points (evenly distributed throughout the pond) using built-in mini feeders. Utilizing surface and underwater cameras at each detection point, image recognition technology tracks the spatiotemporal distribution of the bait as the fish discover and feed, quickly delineating the fish's "initial feeding hotspot" area at that moment. If the water transparency is <30cm (as determined by the camera), the system switches to a combined "water disturbance sensor + sound sensor" approach: the frequency of water vibrations caused by fish swimming (≥5Hz) and the intensity of feeding sounds (≥40dB) replace visual identification of feeding behavior.

[0051] Competition intensity modeling: In the "initial feeding hotspot" area, by analyzing images of fish school aggregation, we identify individual spacing and the frequency of feeding behavior, and establish a micro-feeding competition intensity index. :

[0052]

[0053] in, Local fish density The average inter-individual distance, The growth rate of the area where fish gather within 5 minutes ( , This represents the increase in the area of ​​clustering. (This refers to the initial aggregation area). The higher the value, the more intense the competition for food in the area, and the more difficult it is for weaker individuals to obtain food. The parameters are obtained as follows:

[0054] Local fish density Real-time images of the "initial feeding hotspot" area are collected using a network of surface and underwater cameras. The YOLO object detection algorithm is used to identify the outline coordinates of each bass in the images. The actual area S of the detection area (calculated from the camera installation position, angle, and calibration parameters, unit: m²) is used as the denominator to identify the number of fish in the school. For the molecule, the calculation yields (Unit: tails / m²). If there is an error in the recognition of a single frame, the average value of 10 consecutive frames is taken to improve stability.

[0055] Average interpersonal distance The center coordinates of each fish identified based on the YOLO algorithm described above. The distance between any two fish can be calculated using the Euclidean distance formula. Calculate the sum of all pairwise distances within the region. Then divide by the logarithm of the effective spacing ( ), thus obtaining the average inter-individual distance. (Unit: m). If the number of fish in the area... The default value is 1.5 times the body length of a single fish (the body length is dynamically updated by the initial stocking size and growth model, without the need for manual measurement).

[0056] The number of feeding behaviors observed per unit time (statistically counted in real time by image recognition algorithm: when the fish body outline overlaps with the bait outline and is accompanied by a tail swing amplitude ≥30°, it is judged as 1 feeding behavior, and the total number within 1 minute is counted).

[0057] Layered feeding partition generation: Based on the statistical characteristics of the entire pool's CI values, an adaptive threshold is used to divide the layers (avoiding the limitations of a fixed threshold): =Mean CI of the entire pool + 1.2 standard deviations =The total pool CI mean minus 0.8 standard deviations divides the pond into at least three tiers:

[0058] High-competition zone (H zone): They need to be fed sufficient amounts of bait quickly and promptly to alleviate competition.

[0059] Medium / Low Competition Zone (M / L Zone): Feed them according to the usual strategy.

[0060] Non-feeding zone (N zone): No effective feeding behavior was observed, so feeding was suspended or only very small amounts were given as a trial.

[0061] Step 3: Generation and Execution of Layered Dynamic Feeding Strategy

[0062] Total feeding energy determined: based on the results obtained in step one. The growth energy required to achieve the target daily growth rate (1.2% / day) and feed energy conversion efficiency (0.40), determine the total energy of the feed required for this feeding. :

[0063] ;

[0064] Among them, the feed energy conversion efficiency This refers to the portion of the total energy consumed by fish that is used for metabolic maintenance. With growth energy The effective utilization rate.

[0065] Based on the energy value per unit weight of the feed used Converted to total feeding amount :

[0066] ;

[0067] Hierarchical dynamic allocation: Dynamically allocate partitions according to the following principles:

[0068] High-competition zone (H zone): Obtain allocation weight Furthermore, it employs a feeding pattern of "high single-batch quantity, low frequency, and precise spot feeding" to maximize bait accessibility. The calculation formula is as follows: ,in The compensation coefficient for area H based on historical data ( ).

[0069] Medium / Low Competition Zone (M / L Zone): Obtaining Allocation Weight =0.45, using the conventional feeding mode.

[0070] Non-feeding zone (N zone): assigned minimal weight =0.05, only exploratory feeding was conducted to monitor changes in the fish population's condition.

[0071] The weights satisfy: =1, and the weights can be dynamically initialized based on the area of ​​each partition and the initial feeding hotspot detection results.

[0072] Adaptive adjustment during feeding: During the feeding process, the feeding intensity of each zone is analyzed in real time. If the feeding intensity of a certain zone drops rapidly after the start of feeding, the subsequent feeding amount for that zone is dynamically reduced, and the saved feed quota is added in real time to the zones where the feeding intensity remains high.

[0073] Step 4: Optimization of model parameters based on feedback of uniform individual growth

[0074] Without the need for manual sampling and harvesting, the coefficient of variation in population body weight can be estimated by linking regional feeding intensity with growth models. The formula is ( CI value for high-competition regions CI value for low competition region; The correction coefficient (default 0.85) is automatically updated every 3 days and serves as the core evaluation indicator for feeding uniformity and effectiveness.

[0075] like Figure 2 , 4 As shown, parameter closed-loop optimization: As the optimization objective, a feedback loop is established. If An increase indicates that uneven growth within the population is intensifying, thus automatically adjusting the allocation weights in step three. , , and H-zone compensation coefficient In subsequent feeding, resources will be further allocated to high-competition areas, and the competition threshold in step two will be optimized. and This process can be achieved using reinforcement learning algorithms.

[0076] Implementation Case 1

[0077] The stratified precision feeding method based on the physiological metabolic energy demand model of sea bass proposed in this invention is based on real-time perception of the metabolic demand and feeding competition pattern of sea bass populations through multi-source sensing, generating dynamic differentiated feeding strategies to solve the problem of mismatch between static supply and dynamic demand in traditional feeding. The following details the equipment deployment, parameter calibration, full-process execution, and effect verification in a practical aquaculture scenario.

[0078] I. Implementation Scenarios and Basic Configuration

[0079] This embodiment selects a typical California bass intensive culture pond in the Pearl River Delta region of Guangdong Province. The specific basic conditions are as follows:

[0080] Pond specifications: Area 5 mu (approximately 3335㎡), rectangular layout (length to width ratio 3:1), average water depth 2m, effective water volume The pool embankment is 1.2m wide and equipped with conventional aeration equipment (4 1.5kW impeller aerators).

[0081] Aquaculture parameters: The species stocked is California bass (largemouth bass), the initial stocking size is 50g / fish, the stocking density is 8000 fish / acre, and the initial total stocking quantity is 40000 fish; the target size for growth is 500g / fish, the aquaculture cycle is 150 days, and the target daily growth rate is 1.2% / day.

[0082] Feed selection: Use commercial compound feed with a crude protein content of 45% and energy per unit weight. =22kJ / g, feed energy conversion efficiency =0.40, feed particle size is suitable for the growth stage of sea bass (initially 2mm, gradually increase to 6mm).

[0083] II. Equipment Deployment and Initial Calibration

[0084] Three high-precision dissolved oxygen sensors were installed at the upwind, middle and downwind points of the pond, and a blank control group (1% by volume) without fish was set up to offset the interference of non-fish oxygen consumption.

[0085] An ADCP (Acoustic Doppler Velocity Profiler) was deployed in the center of the pond to monitor the coefficient of variation of the flow velocity. .

[0086] Eight fixed detection points are evenly distributed throughout the pool, integrating miniature automatic feeders and a camera network.

[0087] Feeding equipment: Equipped with 3 intelligent zone feeders with GPS positioning function, supporting the switching between "precise spot feeding" and "uniform area feeding" modes.

[0088] Parameter calibration: Initialize the average volume of a single fish =0.0005m 3 / tail.

[0089] III. Full Process Implementation Steps

[0090] 1. Online calculation of instantaneous metabolizable energy demand on day 90 of aquaculture.

[0091] The system updates physiological data every 10 seconds. It measures the decrease in dissolved oxygen in real time. =0.22mg / L / h (loss of blank group has been removed), real-time water temperature T=28℃.

[0092] According to the formula The oxygen thermal equivalence coefficient is calculated as follows: .

[0093] Substitute into the formula The basal respiratory metabolic energy was obtained. .

[0094] Simultaneous monitoring of flow velocity variability Derivation of average swimming speed The number of fish can be determined by combining image recognition. Tail. According to the formula Calculated metabolic energy .

[0095] Ultimately, the instantaneous total metabolic energy requirement is obtained. .

[0096] 2. Dynamic zoning of ponds

[0097] Add 0.5g of attractant bait 10 minutes before the actual feeding, and use a camera to identify the initial feeding hotspots.

[0098] Competition intensity calculation: Based on image recognition of local density Interpersonal spacing and frequency of food grabbing calculate .

[0099] Partition generation: Based on adaptive thresholds, the area is divided into a high-competition zone (H zone), a medium / low-competition zone (M / L zone), and a non-feeding zone (N zone).

[0100] 3. Strategy Execution and Adaptive Adjustment

[0101] Calculate the total amount of feed. :based on With target growth energy The calculated feed intake is 33.4 kg. H zone: weight 0.50, compensation coefficient... It adopts a "high single-volume, low-frequency, precise point-to-point" mode.

[0102] M / L region: weight 0.45, using conventional uniform surface projection.

[0103] Real-time feedback: If the growth rate of the clustered area in a certain zone decreases by ≥50%, the quota for that zone will be reduced and added to a highly competitive zone.

[0104] 4. Parameter closed-loop optimization

[0105] The population weight variation coefficient is automatically updated every 3 days. .like The system automatically adjusts the allocation weight and compensation coefficient of the H zone to achieve precise resource allocation.

[0106] 5. To objectively evaluate the aquaculture benefits of the precision feeding method described in this invention, the feed conversion ratio (FCR) is introduced as an evaluation index. FCR is the ratio of the cumulative feed amount to the total weight gain of the fish population during the aquaculture cycle, and its calculation method follows the formula:

[0107] FCR=

[0108] Among them: Total feed intake (total feeding amount): The cumulative value of the total amount of feed fed in each partition (H, M / L, N) during the experiment; Accumulated value;

[0109] Overall weight gain (total weight gain): The total weight at harvest minus the initial total weight at stocking.

[0110] IV. Verification of implementation effect

[0111] After the end of the culture cycle, the experimental group showed significant advantages:

[0112] Table 1: Comparison table of the effects of the intelligent stratified feeding method and the traditional feeding method of the present invention

[0113]

[0114] Chart analysis: The chart shows that the final weight of growth performance increased by 9.8%, the uniformity of the population increased by 55.7%, the feed coefficient decreased by 17.0%, the survival rate increased by 7.7 percentage points, and the culture cycle was shortened by 9.1%.

[0115] Implementation case 2

[0116] I. Implementation scenario and basic configuration

[0117] This embodiment selects a precision farming base for Micropterus salmonides in Liyang City, Jiangsu Province, and the basic conditions are as follows:

[0118] Pond specifications: The area is 10 mu (about 6677), in a rectangular layout (the length-width ratio is about 4:1), the average water depth is 2.5 m, and the effective water volume is calculated. . The pond is equipped with 6 2.2 kW impeller aerators and 2 bottom aeration devices.

[0119] Culture parameters: The stocked variety is "Youlu No. 3", the initial stocking size is 60 g / tail, the stocking density is 10,000 tails / mu, and the initial total stocking amount is 100,000 tails. The target growth size is 550 g / tail, the culture cycle is set to 150 days, and the target daily growth rate is set to 1.15% / day.

[0120] Feed selection: High-protein expanded floating feed is used, and its unit weight energy , feed energy conversion efficiency .

[0121] II. Equipment deployment and initial calibration

[0122] Multi-source sensor network: Five high-precision dissolved oxygen and temperature sensors are deployed at the upwind, midwind, and leeward sides of the pond, as well as at the geometric center. Twelve fixed detection points are evenly distributed throughout the pond, each integrating a miniature automatic feeder, surface and underwater cameras. An ADCP current meter is deployed at the centerline of the pond's long axis.

[0123] Initial parameter calibration: Initialize the average volume of a single fish The average oxygen consumption rate of the blank control group was determined by continuous data collection using sensors. .

[0124] III. Implementation Process of the Entire Process (Taking Day 60 of Breeding as an Example)

[0125] Online estimation of population instantaneous metabolic energy: Actual measurement at the main control terminal Real-time water temperature Calculate according to the formula. The basal respiratory metabolic energy of the population was calculated. Combined with the swimming velocity variation monitored by ADCP And the number of fish N identified by image recognition, to derive the activity metabolic energy. Ultimately, the instantaneous total metabolic energy is obtained. .

[0126] Pond dynamic zoning based on competition intensity identification: Hotspot detection is performed before feeding, and the system calculates the statistical characteristics of the competition intensity (CI) values ​​for the entire pond. The calculated mean CI for the entire pond is 45.2, and the standard deviation is 12.8. Zoning is then based on an adaptive threshold.

[0127] High-competition zone (H zone): CI > 60.56;

[0128] Medium / Low Competition Zone (M / L Zone): 34.96 <CI<60.56;

[0129] Non-feeding zone (N zone): CI < 34.96;

[0130] Layered dynamic feeding strategy generation and execution: Calculate the total amount of feed given in a single feeding. H area weight allocation W H =0.50, adopting the "high single-volume, precise point-to-point" mode. During the execution, if the food competition intensity of probe point 3 decreases by ≥55%, the system will automatically add its subsequent quota to probe point 5, where the competition intensity is still high, in real time.

[0131] IV. Implementation Results Verification

[0132] Table 2 Comparison of Application Effects in Large-Scale Intensive Aquaculture Ponds in Liyang, Jiangsu

[0133]

[0134] Chart Analysis: The chart shows that the final weight increased by 10.8%, the breeding cycle was shortened by 10.7% to 150 days, the average weight of the group increased by 48.2%, the feed conversion ratio decreased by 23.4%, and the survival rate increased by 6.5 percentage points.

[0135] V. Precautions

[0136] Equipment maintenance: After each feeding, clean the camera lens and sensor probe to avoid feed residue and algae adhesion affecting data accuracy; calibrate the feeding accuracy of the mini feeder and smart feeder once a week, and adjust in time if the error exceeds ±2%; check the cleanliness of the ADCP sensor probe monthly to ensure the accuracy of flow rate monitoring.

[0137] Emergency Handling: If a sensor malfunctions (automatically detected by the system), immediately activate the backup calculation model (e.g., if the dissolved oxygen sensor malfunctions, use a regression model of "temperature-pH value-fish activity intensity" to estimate). At the same time, an alarm will be issued; during extreme weather such as heavy rain and high temperature, the system will automatically reduce the target daily growth rate to 1.0% / day and reduce the amount of feed; when the pH value of the water body deviates from the range of 7.0-8.5, feeding will be suspended and trial feeding will be carried out after the water quality recovers.

[0138] Data Management: The main control terminal automatically stores daily key data (metabolic energy requirements, feeding amount in different zones, CI value, water quality parameters, etc.) for a period of no less than one year. It supports data export and trend analysis to provide a basis for subsequent aquaculture optimization.

[0139] In summary, this invention discloses an intelligent stratified precision feeding method based on a physiological metabolic energy demand model for sea bass. This method utilizes online monitoring... Real-time estimation of physiological metabolic energy requirements based on fish movement data A competition intensity model was constructed using pre-feeding entrapment detection. The pond is adaptively divided into hierarchical regions such as high-competition areas. Under parameter control, the system combines growth energy Determine the total amount of feed It implements tiered, differentiated feeding and adjusts based on real-time image feedback. Relying on... The closed-loop system enables autonomous parameter optimization, ultimately reducing the free flow rate (FCR) to 1.11 (an optimization of 23.4%) in practical applications, increasing the average body weight by 10.8%, optimizing the population's average weight by 55.7%, and achieving a stable survival rate of 96.5% even in low-transparency environments. This method constructs a standardized production paradigm of "direct metabolic reading - hierarchical compensation - closed-loop evaluation," providing high-yield, high-quality, and low-cost technical support for intensive bass farming.

[0140] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A smart, stratified, and precise feeding method based on a physiological metabolic energy demand model for sea bass, characterized in that, Includes the following steps: Step 1: Online estimation of instantaneous total metabolic energy demand of the bass population; (1) Real-time acquisition of dissolved oxygen consumption rate and fish movement data in the pond to be fed, establishment of physiological metabolic model, real-time estimation of instantaneous metabolic energy demand of fish in the current environment, and calculation of basal respiratory metabolic energy demand. : ； In the formula: The oxygen consumption rate of pure fish after removing the losses in the blank control group; The effective water volume of the pond; The oxygen thermal equivalent coefficient; Based on real-time water temperature Dynamic correction is performed, and the correction formula is as follows: ; (2) Estimating the metabolic energy demand of activities by linking acoustic Doppler current profiler and camera network image recognition : ； In the formula: , It is the coefficient of variation of local water flow velocity monitored using an acoustic Doppler current profiler. For calibration coefficients; The average swimming speed of the fish; and To correct the parameters; The number of fish; (3) The sum of the two gives the instantaneous total metabolic energy requirement. : ； Step 2: Identify dynamic pond zoning based on feeding competition intensity; Before feeding, a feeding detection exercise was performed to track the feeding behavior distribution of the fish in real time and to construct a feeding competition intensity model. Based on this, the pond is divided into several hierarchical areas with differentiated competitive capabilities; Step 3: Generation and execution of hierarchical dynamic feeding strategy; Based on the instantaneous total metabolic energy requirement Growth energy required for the target daily growth rate Determine the total amount of feed The system implements a differentiated feeding mode based on the distribution weight of the hierarchical regions and makes real-time dynamic adjustments based on the feeding images during the feeding process. Step 4: Optimize model parameters based on feedback of uniform individual growth; Monitor feeding effects and calculate the coefficient of variation of population body weight. ,based on Closed-loop correction of the feeding weight allocation coefficient and competition level division threshold in subsequent cycles.

2. The intelligent stratified precision feeding method based on the physiological metabolic energy demand model of sea bass according to claim 1, characterized in that, In step two, the competition intensity model The construction logic is as follows: use object detection algorithms to identify the center coordinates of the fish school and calculate the local fish density. Compared with the average individual distance Combined with the frequency of food-grabbing behavior per unit time Overall assessment.

3. The intelligent stratified precision feeding method based on the physiological metabolic energy demand model of sea bass according to claim 1, characterized in that, In step two, the partitioning hierarchy uses an adaptive threshold: high-competition region threshold. =Mean CI of the entire pool + 1.2 standard deviations; Threshold for non-feeding areas =Mean CI of the entire pool - 0.8 times the standard deviation.

4. The intelligent stratified precision feeding method based on the physiological metabolic energy demand model of sea bass according to claim 1, characterized in that, In step three, the total amount of feed... The calculation formula is: In the formula: To determine the growth energy required based on the target daily growth rate, For feed energy conversion efficiency, This refers to the energy value per unit weight of the feed used.

5. The intelligent stratified precision feeding method based on the physiological metabolic energy demand model of sea bass according to claim 1, characterized in that, In step three, the differentiated feeding mode includes: in high-competition areas, a "high single-volume, precise spot feeding" mode is adopted, and a compensation coefficient is introduced. Adjust quotas and feed volume in highly competitive areas The calculation formula is: , Weights are assigned to high-competition areas; a standard uniform surface projection mode is used for medium / low-competition areas.

6. The intelligent stratified precision feeding method based on the physiological metabolic energy demand model of sea bass according to claim 1, characterized in that, In step three, the adaptive adjustment logic of the feeding process is as follows: real-time monitoring of the feeding images of each zone; if the growth rate of the clustered area in a certain zone decreases by ≥50%, the subsequent quota for that zone is reduced, and the saved feed is added in real time to the zone where the feeding intensity is still high.

7. The intelligent stratified precision feeding method based on the physiological metabolic energy demand model of sea bass according to claim 1, characterized in that, In step four, the coefficient of variation of population body weight The formula for estimating the linkage is: In the formula: For correction factor, and These represent the average competition intensity in the high-competition and low-competition zones, respectively.

8. The intelligent stratified precision feeding method based on the physiological metabolic energy demand model of sea bass according to claim 1, characterized in that, It also includes a parameter closed-loop optimization process: the system automatically updates once every set threshold number of days. If the value shows an increasing trend, the weights allocated to high-competition regions will be automatically increased using a reinforcement learning algorithm. and compensation coefficient This enables precise allocation of resources.

9. The intelligent stratified precision feeding method based on the physiological metabolic energy demand model of sea bass according to claim 1, characterized in that, It also includes emergency treatment for stress responses in bass under extreme conditions: if a sudden increase in water flow velocity is detected. or water body If the value deviates from the range of 7.0-8.5, immediately suspend feeding and resume feeding at 50% of the initial quota after a set threshold time.

Images