Industrial aquaculture intelligent feeding method and system based on feeding plan

Abstract

The invention relates to an industrial aquaculture intelligent feeding method and system based on a feeding plan, and the method comprises the following steps: (1) making an initial feeding plan which comprises feeding time, feeding frequency and feeding amount; (2) performing pre-feeding evaluation before feeding every time, wherein the pre-feeding evaluation comprises evaluation of water quality conditions and disease states; (3) dynamically adjusting the feeding strategy according to the evaluation result before feeding; (4) after the feeding is finished, carrying out evaluation after feeding; and (5) generating a feeding table according to an evaluation result after feeding, wherein the feeding table is used for optimizing a subsequent feeding plan. The method has the advantages that by accurately controlling the feeding amount and time every time, it is ensured that animals obtain the most appropriate nutrition according to the growth stages and health conditions of the animals, and rapid and healthy growth is promoted.

Description

Technical Field

[0001] This invention relates to an intelligent feeding method for factory farming based on a feeding plan, belonging to the field of computer technology. Background Technology

[0002] Factory farming is a high-density, intensive aquaculture model that utilizes modern engineering technologies and facilities to create a controlled environment. Through technologies such as recirculating aquaculture systems (RAS), this model can effectively control key environmental parameters such as water quality, temperature, and dissolved oxygen, significantly improving resource utilization and yield per unit area, and is an important development direction for the modernization of aquaculture.

[0003] In factory farming, feeding management is a core element affecting both farming efficiency and the ecological environment. Insufficient feeding restricts the growth rate of farmed organisms and prolongs the farming cycle; while overfeeding leads to feed waste, increases the content of harmful substances such as ammonia nitrogen in the water, deteriorates water quality, triggers diseases, and ultimately creates a vicious cycle. Traditional feeding methods mainly rely on the experience of farmers to feed at fixed times and in fixed quantities, which has problems such as strong subjectivity, poor consistency, and inability to respond to real-time changes in the dynamics of farmed organisms.

[0004] With the development of automation technology, automated feeders have been introduced into the aquaculture process, achieving preliminary automated feeding at fixed times and in fixed quantities. However, these systems still require preset fixed feeding programs and lack the ability to perceive and respond to the real-time feeding needs, health status, and environmental changes of farmed organisms. Essentially, they are still "open-loop" controls with limited levels of intelligence.

[0005] In recent years, intelligent feeding technology, based on sensors, machine vision, and artificial intelligence algorithms, has become a research hotspot, attempting to solve the aforementioned problems. Existing intelligent feeding solutions mainly fall into the following categories: 1. Feeding Model Based on Environmental Sensors: This type of model monitors the aquatic environment in real time by deploying water quality sensors such as dissolved oxygen, temperature, pH, and ammonia nitrogen. The system adjusts or triggers feeding behavior based on preset water quality parameter thresholds. However, this method only makes indirect judgments from an environmental perspective and cannot directly obtain the feeding status (such as hunger level and appetite) and health status of the cultured organisms themselves. The decision-making basis is singular, and the accuracy is limited.

[0006] 2. Acoustic-based feeding model: This technology determines the feeding intensity of fish by collecting and analyzing the sound signals generated by their feeding activities, thereby controlling the feeding. It has certain advantages in turbid water, but is easily affected by noise from equipment such as water pumps and aerators, and struggles to distinguish feeding behaviors of different species or individual sizes, posing challenges to its accuracy and universality.

[0007] 3. Computer Vision-Based Feeding Models: These models utilize underwater cameras to capture images or videos, employing image processing or deep learning algorithms to identify fish aggregation behavior, feed residue levels, etc., thereby achieving feeding control. This is currently the most promising direction. However, existing applications mostly focus on real-time feedback control of a single feeding process, such as monitoring whether the feed has been consumed to determine when to stop feeding. Its limitations are: Lack of integration with macro-production planning: Most of the focus is on "real-time control" on a case-by-case basis, failing to dynamically link the effect evaluation of each feeding with long-term, phased feeding plans.

[0008] The decision-making dimension is too narrow: it usually relies on only one or two dimensions such as feeding behavior or food residue, and fails to take into account equally important influencing factors such as water quality and disease characteristics, resulting in insufficient comprehensiveness and scientific rigor in the model's decision-making.

[0009] Insufficient utilization of historical data: Failure to systematically build feeding history archives, inability to continuously learn and iterate to optimize subsequent feeding strategies, and difficulty in achieving true "plan-execution-evaluation-optimization" closed-loop management.

[0010] Therefore, existing technologies have not yet provided an intelligent feeding solution that can closely integrate feeding plans, synthesize multi-source information, and perform pre- and post-feeding evaluations to achieve continuous self-optimization. This invention aims to overcome the shortcomings of the existing technologies and provide a more refined, intelligent, and systematic feeding model and method. Summary of the Invention

[0011] To overcome the shortcomings of existing technologies, this invention provides an intelligent feeding method for factory farming based on a feeding plan. The technical solution of this invention is as follows: A smart feeding method for factory farming based on a feeding plan includes the following steps: (1) Develop an initial feeding plan, including feeding time, feeding frequency and feeding amount; (2) Conduct a pre-feeding evaluation before each feeding, including an assessment of water quality and disease status; (3) Adjust the feeding strategy dynamically based on the pre-feeding evaluation results; (4) Conduct a post-feeding evaluation after the feeding is completed; (5) Generate a feeding table based on the evaluation results after feeding, which is used to optimize the subsequent feeding plan.

[0012] In step (2), the water quality assessment in the pre-feeding evaluation uses a parabolic vertex function to quantify and score water quality parameters, specifically including: Set standard upper and lower limits for water quality parameters, and calculate the water quality score corresponding to the real-time monitoring value using a parabolic function; Assuming the upper and lower limits of the water quality parameter standard values ​​are X1 and X2, the water quality score is set to 90 points when the real-time monitoring value is X1 or X2, and 90 points when the monitoring value is X1. At that time, the water quality score was 100 points. Substituting the parameters into the parabolic vertex equation, as shown in Formula 1: (1); Where (h,k) are the coordinates of the vertex of the parabola, the function for water quality scoring is obtained, as shown in Formula 2: (2); In step (2), the disease status assessment in the pre-feeding evaluation uses the YOLO target detection method to identify underwater video images. The identification content includes the outline of the shrimp body, the health status of the intestines, stomach, and eyes. For each disease or molting phenomenon identified, the disease coefficient decreases by 0.01.

[0013] In step (4), the post-feeding evaluation includes: monitoring the bait coverage area and the number of shrimp through an underwater camera, calculating the ratio of the actual feeding time to the expected feeding time, and judging whether the feeding amount is appropriate based on the shrimp aggregation behavior; The area covered by the feed at the moment of completion of feeding is recorded as 1. When the area covered is less than 0.1, it is considered that the feed has been eaten. The expected feeding time is recorded as Y, and the actual feeding time is recorded as y. When the number of shrimp appearing in the video exceeds X, it is considered to indicate aggregation behavior. X needs to be determined based on the stocking density, the camera's field of view, and the location of the feeding platform. Feeding is evaluated by comparing the relationship between y and Y, and the number of shrimp X in the video at time Y.

[0014] An intelligent feeding system for implementing the intelligent feeding method for factory farming based on feeding plans includes: The feeding plan creation module is used to create an initial feeding plan; The pre-feeding evaluation module is used to assess water quality and disease status; The feeding control module is used to dynamically adjust the feeding strategy based on the evaluation results before feeding. The post-feeding evaluation module is used to analyze feeding behavior and generate feeding tables; The data storage module is used to store the feeding table and historical data.

[0015] The advantages of this invention are: (1) Improve breeding efficiency: By precisely controlling the amount and time of each feeding, ensure that animals receive the most suitable nutrition according to their growth stage and health status, and promote rapid and healthy growth.

[0016] (2) Reduce feed waste: Intelligent management of feeding plans significantly reduces feed waste caused by overfeeding or underfeeding, while also reducing health problems caused by indigestion.

[0017] (3) Data-driven decision support: real-time collection of animal growth data and environmental data, and scientific feeding strategies through data analysis to achieve data-driven breeding management. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the present invention.

[0019] Figure 2 This is a schematic diagram of the intelligent feeding system of the present invention. Detailed Implementation

[0020] The present invention will be further described below with reference to specific embodiments, and the advantages and features of the present invention will become clearer as a result. However, these embodiments are merely exemplary and do not constitute any limitation on the scope of the present invention. Those skilled in the art should understand that modifications or substitutions can be made to the details and form of the technical solutions of the present invention without departing from the spirit and scope of the present invention, but all such modifications and substitutions fall within the protection scope of the present invention.

[0021] See Figure 1 and Figure 2 This invention relates to an intelligent feeding method for factory farming based on a feeding plan, comprising the following steps: (1) Develop an initial feeding plan, including feeding time, feeding frequency and feeding amount; (2) Conduct a pre-feeding evaluation before each feeding, including an assessment of water quality and disease status; (3) Adjust the feeding strategy dynamically based on the pre-feeding evaluation results; (4) Conduct a post-feeding evaluation after the feeding is completed; (5) Generate a feeding table based on the evaluation results after feeding, which is used to optimize the subsequent feeding plan.

[0022] In this embodiment, when formulating the feeding plan, the initial feeding plan is based on days. The daily feeding amount and frequency are shown in the table. The feeding times for the four meals a day are 6:00, 12:00, 18:00 and 21:00, and the feeding times for the six meals a day are 3:00, 7:00, 11:00, 15:00, 19:00 and 23:00. The feeding during the peak feeding period accounts for 70%, and the feeding during other times accounts for 50%.

[0023] In step (2), the water quality assessment in the pre-feeding evaluation uses a parabolic vertex function to quantify and score water quality parameters, specifically including: Set standard upper and lower limits for water quality parameters, and calculate the water quality score corresponding to the real-time monitoring value using a parabolic function; Assuming the upper and lower limits of the water quality parameter standard values ​​are X1 and X2, the water quality score is set to 90 points when the real-time monitoring value is X1 or X2, and 90 points when the monitoring value is X1. At that time, the water quality score was 100 points. Substituting the parameters into the parabolic vertex equation, as shown in Formula 1: (1); Where (h,k) are the coordinates of the vertex of the parabola, the function for water quality scoring is obtained, as shown in Formula 2: (2); Taking dissolved oxygen as an example, assuming the standard value is 3.5 mg / L to 4.5 mg / L, it can be calculated that... If the real-time dissolved oxygen is 3 mg / L, the water quality score is 60 points.

[0024] In step (2), the disease status assessment in the pre-feeding evaluation uses the YOLO target detection method to identify underwater video images. The identification content includes the outline of the shrimp body, the health status of the intestines, stomach, and eyes. For each disease or molting phenomenon identified, the disease coefficient decreases by 0.01.

[0025] The disease coefficient model is based on an underwater camera and the YOLO object detection method. It identifies real-time underwater video footage. The underwater camera is installed near the feeding platform to observe leftover food and shrimp feeding on it. The underwater video is output as real-time video frames, and the LabelImg tool is used to annotate key parts of the shrimp, such as the outline, intestines, stomach, and eyes. After annotation, the annotated images are divided into training, validation, and test sets. This reasonable dataset partitioning ensures the model's generalization ability and robustness. Then, YOLOv7-Tiny is used to identify diseases in the underwater images. The identification results are stored in the database at 2-minute intervals, including the number of organisms identified within that time period, average body length, average weight, and health coefficient. The initial disease coefficient is set to 1. During the identification process, for each disease or molting phenomenon detected, the disease coefficient is decreased by 0.01, thus quantifying the health status.

[0026] In step (4), the post-feeding evaluation includes: monitoring the bait coverage area and the number of shrimp through an underwater camera, calculating the ratio of the actual feeding time to the expected feeding time, and judging whether the feeding amount is appropriate based on the shrimp aggregation behavior; The area covered by the feed at the moment of completion of feeding is recorded as 1. When the area covered is less than 0.1, it is considered that the feed has been eaten. The expected feeding time is recorded as Y, and the actual feeding time is recorded as y. When the number of shrimp appearing in the video exceeds X, it is considered to indicate aggregation behavior. X needs to be determined based on the stocking density, the camera's field of view, and the location of the feeding platform. Feeding is evaluated by comparing the relationship between y and Y, and the number of shrimp X in the video at time Y.

[0027] The feeding evaluation criteria are shown in the table below: The feeding log summarizes the evaluation results of each feeding session. It records detailed information about each feeding, including water quality and disease information at the time of feeding, and serves as an important basis for developing subsequent feeding plans. The feeding log data is shown in the table below.

[0028] The present invention also relates to an intelligent feeding system for implementing the intelligent feeding method for factory farming based on feeding plans, comprising: Feeding plan formulation module 1 is used to formulate the initial feeding plan; Pre-feeding evaluation module 2 is used to assess water quality and disease status; Feeding control module 3 is used to dynamically adjust the feeding strategy based on the evaluation results before feeding. Post-feeding evaluation module 4 is used to analyze feeding behavior and generate feeding tables; Data storage module 5 is used to store the feeding table and historical data.

[0029] This invention constructs a dynamic closed-loop feedback system of "planning-execution-evaluation-optimization" based on a feeding plan-based intelligent feeding method for factory farming. It combines long-term feeding plans with short-term real-time adjustments, and achieves continuous self-optimization of the feeding strategy through two evaluation mechanisms. Its working principle is as follows: 1. Planning Level: Developing the initial feeding plan. First, based on factors such as the farmed species (e.g., Litopenaeus vannamei), growth stage (early stage, early stage, middle stage, late stage, and late stage), and historical data, an initial feeding plan is developed on a daily basis. This plan specifies the daily feeding time, frequency, and amount, providing a scientific and controllable baseline framework for the entire farming process.

[0030] 2. Feedforward control: Pre-feeding evaluation and real-time adjustment Before each planned feeding, the system initiates a pre-feeding evaluation mechanism, a feedforward control step designed to predict the suitability of the feeding. It comprehensively assesses two key factors: Water quality status: Water parameters (such as dissolved oxygen, pH, and temperature) are acquired through sensors, and an innovative parabolic vertex function model is used for quantitative scoring. This model can more scientifically reflect the degree to which parameter values ​​deviate from the ideal range, rather than simply judging "qualified / unqualified".

[0031] Disease Status: Using underwater cameras deployed near the feeding platform, the YOLO target detection algorithm analyzes video footage in real time to automatically identify the health status of shrimp (such as disease symptoms and molting behavior). For each negative feature identified, the disease coefficient decreases accordingly, thus quantifying the health risk.

[0032] Based on the results of the previous evaluation (water quality score, disease coefficient), the feeding strategy is dynamically adjusted (such as reducing the amount of feed or even suspending feeding as appropriate) to adapt to the real-time environment and biological conditions, and to avoid waste or pollution caused by feeding under unfavorable conditions.

[0033] 3. Execution layer: Performs the feeding operation. The feeding control module controls the automatic feeder to perform precise feeding according to the adjusted strategy.

[0034] 4. Feedback Control: Post-feeding evaluation and effect assessment After feeding, a post-feeding evaluation mechanism is activated. This feedback control step aims to assess the actual effectiveness of the feeding. Analysis is also performed using an underwater camera. Monitoring feeding duration: The system tracks changes in the bait coverage area, records the actual feeding duration (y), and compares it with the expected feeding duration (Y).

[0035] Analyze feeding behavior: At the same time, the system counts the number of shrimp around the feeding platform to determine whether there is aggregation behavior (whether the number exceeds the threshold X).

[0036] By combining the feeding time and the number of shrimp, the system uses a preset evaluation standard table to determine the feeding coefficient for this feeding, thus making a precise and quantitative evaluation of whether the feeding amount is "too much", "too little" or "appropriate".

[0037] 5. Optimization and Learning: Generating Feeding Tables and Planning Iterations Detailed information for each feeding (time, amount of feed, environmental data, before-and-after evaluation results, etc.) is generated into a structured feeding table and stored in a database. This table has become valuable big data for aquaculture.

[0038] When formulating subsequent feeding plans (such as the plan for the next day), the system calls upon and analyzes this historical data. For example, if evaluations after multiple feedings show that the feeding coefficient remains at 1.2 (indicating a high feeding amount), the system will automatically adjust the future feeding amount benchmark based on the initial plan. This achieves closed-loop management based on data-driven, continuously iterative optimization.

[0039] By deeply integrating computer vision, sensor technology, mathematical models, and aquaculture management processes, this invention achieves a leap from traditional "fixed procedure" or "single-response" feeding to intelligent feeding that is globally optimized, adaptive, and data-driven, ultimately improving efficiency, reducing waste, and lowering risks.

[0040] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A smart feeding method for factory farming based on a feeding plan, characterized in that, Includes the following steps: (1) Develop an initial feeding plan, including feeding time, feeding frequency and feeding amount; (2) Conduct a pre-feeding evaluation before each feeding, including an assessment of water quality and disease status; (3) Adjust the feeding strategy dynamically based on the pre-feeding evaluation results; (4) Conduct a post-feeding evaluation after the feeding is completed; (5) Generate a feeding table based on the evaluation results after feeding, which is used to optimize the subsequent feeding plan.

2. The intelligent feeding method for factory farming based on a feeding plan according to claim 1, characterized in that, In step (2), the water quality assessment in the pre-feeding evaluation uses a parabolic vertex function to quantify and score water quality parameters, specifically including: Set standard upper and lower limits for water quality parameters, and calculate the water quality score corresponding to the real-time monitoring value using a parabolic function; Assuming the upper and lower limits of the water quality parameter standard values ​​are X1 and X2, the water quality score is set to 90 points when the real-time monitoring value is X1 or X2, and 90 points when the monitoring value is X1. At that time, the water quality score was 100 points. Substituting the parameters into the parabolic vertex equation, as shown in Formula 1: (1)； Where (h,k) are the coordinates of the vertex of the parabola, the function for water quality scoring is obtained, as shown in Formula 2: (2)。 3. The intelligent feeding method for factory farming based on a feeding plan according to claim 1 or 2, characterized in that, In step (2), the disease status assessment in the pre-feeding evaluation uses the YOLO target detection method to identify underwater video images. The identification content includes the outline of the shrimp body, the health status of the intestines, stomach, and eyes. For each disease or molting phenomenon identified, the disease coefficient decreases by 0.

01.

4. The intelligent feeding method for factory farming based on a feeding plan according to claim 3, characterized in that, In step (4), the post-feeding evaluation includes: By monitoring the bait coverage area and the number of shrimp using underwater cameras, the ratio of actual feeding time to expected feeding time is calculated, and the appropriateness of the feeding amount is determined by combining the shrimp aggregation behavior. The area covered by the feed at the moment of completion of feeding is recorded as 1. When the area covered is less than 0.1, it is considered that the feed has been eaten. The expected feeding time is recorded as Y, and the actual feeding time is recorded as y. When the number of shrimp appearing in the video exceeds X, it is considered to indicate aggregation behavior. X needs to be determined based on the stocking density, the camera's field of view, and the location of the feeding platform. Feeding is evaluated by comparing the relationship between y and Y, and the number of shrimp X in the video at time Y.

5. An intelligent feeding system for implementing the intelligent feeding method for factory farming based on a feeding plan as described in any one of claims 1 to 4, characterized in that, include: The feeding plan creation module is used to create an initial feeding plan; The pre-feeding evaluation module is used to assess water quality and disease status; The feeding control module is used to dynamically adjust the feeding strategy based on the evaluation results before feeding. The post-feeding evaluation module is used to analyze feeding behavior and generate feeding tables; The data storage module is used to store the feeding table and historical data.

Images