KR102570791B1 - System and Method for Controlling of Intelligent Fish Feeder - Google Patents

Abstract

The present invention is an intelligent farm feeding system for supplying food in conjunction with the fish feedability, by measuring the behavior of the fish at specific times - the specific times are the measurement start time and measurement end time of the fish trajectory for a plurality of periods Divided into and classified -, extracting the trajectory image of the fish, and classifying the extracted trajectory image of the fish using a learning classifier - the learning classifier is the horizontal distance and vertical distance that the fish moved - a plurality of steps Feed amount to be supplied to the tank by inputting the feeding desire of the fish in a plurality of stages calculated by the feeding degree extraction unit that calculates the feeding desire of the fish and the tank data measured by the plurality of sensors installed in the farm It includes a feed supply amount management system that calculates data and a feed supply device that receives the calculated feed amount data, measures the weight of feed to be supplied, and supplies it to the water tank.

Description

Intelligent feeder control system and method {System and Method for Controlling of Intelligent Fish Feeder}

The present invention relates to an apparatus and method for supplying feed to fish in a fish farm, and more particularly, to an intelligent feed feeder interlocked with the degree of subsidization of fish.

Recently, climate change such as typhoons, red tides, high water temperature, and large influx of freshwater and negative impacts from pollution of marine landscapes have led to the shift from marine cage farming to land farming, as well as the farming of expensive, highly functional fish species through large-scale land farming and related businesses. It's going on.

For the success of onshore aquaculture, large-scale facility investment must be made, and steady profits must be achieved to sustain the business. To this end, it is necessary to maximize profit realization by reducing waste and mortality of feed and controlling the timing of shipment of superior fish. Therefore, the system of the land farm should be intelligent to achieve the above purpose, and should have an interface that is easy for an actual operator to understand and operate.

In order to reduce the waste and death of feed and to control the timing of the release of superior fish, it is necessary to supply the optimal feed in an environment where the growth of fish is optimal. Here, the meaning of optimal feed supply is to supply the optimal amount at the optimal time to speed up the growth of fish and to ensure that no feed is discarded without being eaten.

As a method for optimal feed supply, research on how to measure the movement of fish using artificial intelligence and convert it into a need for feeding to feed it when the need for feeding is active has been actively conducted. The relationship with the island is not specifically disclosed, and there is no system that provides a neural network model that can learn fish activity by applying it in the actual field.

If, in the case of large-scale aquaculture, water pollution may occur due to oversupply of feed due to incorrect judgment, and death may occur along with disease, or, due to insufficient supply of feed, it is difficult to ship fish with the maximum weight. can have

The reason for this problem is that there is no accurate criterion for determining the feed desire of fish, and the amount of feed is controlled without reflecting the tank environment that affects the growth of fish. For example, when dissolved oxygen is low, metabolism of feed is low, high water temperature can induce growth rate of fish, and sunlight affects the maturation rate of fish. In other words, the most active water temperature for fish growth is between 18 and 22 degrees. Even if the current water temperature of the tank is 15, even if the fish feed rate is recognized as the maximum, if the feed amount is given to the maximum, there is a high probability that the feed will remain, and the metabolism will be lowered, which can cause disease. Therefore, when the water temperature is between 18 and 22 degrees and the fish feed rate is maximum, the maximum amount of feed must be supplied to promote the maximum growth of fish. That is, a model and system for controlling the optimal feed supply amount reflecting the environment of the tank as described above are required.

In general, extract results using multi-task neural network models using multiple inputs (e.g., movement levels, tank environment data, weight, etc.) and higher-order hidden layers to derive optimal feed rates. do. However, it has a disadvantage that it is difficult to apply it to the actual field because it is difficult for the end user to fully understand the extracted results.

In addition, in actual aquaculture sites, the metabolism of fish differs significantly depending on the water temperature in the tank, dissolved oxygen, and the weight of the fish. Therefore, there is a need for a decision-making model of feeding using empirical data in the actual field.
Prior art literature
Korean Registered Patent Publication No. 10-2099447 (2020.04.09.)

The technical problem of the present invention for solving the above problems is to provide an intelligent feed supply system and method capable of preventing waste of feed and contamination of water quality.

Another object of the present invention is to provide an intelligent feeding system and method capable of preventing disease occurrence in fish.

Another object of the present invention is to provide an intelligent feeding system and method capable of accurately supplying feed to a corresponding fish farm tank.

According to one aspect of the present invention for solving the above problems, in an intelligent farm feeding system for supplying food in conjunction with fish feedability, the behavior of fish is measured at specific times - the specific time is the trajectory of the fish The measurement start time and measurement end time of are divided into a plurality of periods and classified -, the fish trajectory image is extracted and the extracted fish trajectory image is a learning classifier - the learning classifier determines the horizontal distance the fish has moved and The vertical distance is - from the feeding degree extraction unit that calculates the feeding desire of the fish in a plurality of stages by using -, and the feeding desire of the fish in the plurality of stages calculated in the feeding degree extraction unit and the plurality of sensors installed in the farm It includes a feed supply management system that calculates feed amount data to be supplied to the tank by inputting the measured tank data, and a feed supply device that receives the calculated feed amount data, measures the weight of feed to be supplied, and supplies it to the tank.

In addition, the susceptibility extraction unit classifies the extracted trajectory image of the fish using the learning classifier using machine learning to calculate the craving for feeding of the fish in the plurality of steps, and as the plurality of inputs of the machine learning It is characterized in that the trajectory image of the fish measured for each specific time.

In addition, among the plurality of inputs of the machine learning, the n-th input may be a trajectory image of a fish measured at an n-th time among the predetermined specific times.

The prey extractor may further include a trajectory tracking unit that tracks a specific fish, assigns an object identification number, and generates a trajectory image of the fish using a plurality of algorithms.

Also, the plurality of algorithms include at least two of YOLOv3, YOLOv2, Farneback, Pyflow, EpicFlow, and FlowNet algorithms.

In addition, the plurality of cravings are determined based on labeling of prey classified by an expert in advance.

In addition, in the type of the feeding degree labeling, the trajectories of the fish for each of the plurality of times are classified into a plurality of types of patterns and classified into active states of low-speed famous, high-speed swimming, stationary state, and abnormal swimming, and if the low-speed swimming, In the case of the high-speed swimming, the need is defined as hunger, in the case of the stationary swimming, the need is defined as very full, and in the case of the abnormal swimming, the state of the fish is defined as disease or stress.

In addition, the feed supply amount management system calculates feed amount data to be supplied to the tank using a fuzzy reasoning engine, and the fuzzy reasoning engine uses a neuro-fuzzy reasoning system to calculate the feed desire of the fish and the measured data from the plurality of sensors. The first feed amount data is derived using the tank data and expert knowledge as input.

In addition, the fuzzy inference engine includes a crisp input layer, an input membership function, a fuzzy rule layer, an input membership function, and a defuzzification layer. and an output layer.

In addition, inputs to the fuzzy reasoning engine include water temperature, dissolved oxygen amount, fish weight, and (?) fish activity status.

In addition, the feeding amount management system further includes a correction calculator that calculates second feed amount data by calculating the first feed amount data and the feeding schedule data input by the farm manager.

In addition, the first feed amount may have a positive feed weight or a negative feed weight according to the feed desire of the fish, the measured tank data from the plurality of sensors, and the expert's knowledge, and the correction calculator The second feed amount is derived by calculating 1 feed amount and feed weight included in the feed supply schedule data.

Also, the second amount of feed may be equal to, greater than, or less than the weight of the feed included in the feed supply schedule.

In addition, the feed supply device is connected to a manager terminal, and the manager terminal receives the feed amount data received through the feed amount management system and controls the feed supply device using an MQTT protocol.

In addition, the feed supply device accumulates the second amount of feed in the feed container by using the opening and closing operation of the slide, and then supplies it to the water tank by using a high-pressure blower while making it fall.

In addition, the feed supply device includes a slide that moves the feed container in which the second amount of feed is accumulated to be coupled to a feed supply pipe connected to a water tank to be fed with the second amount of feed.

In addition, the tank environment data includes at least one of temperature data in the tank, dissolved oxygen data, and fish weight data.

The disclosed technology may have the following effects. However, it does not mean that a specific embodiment must include all of the following effects or only the following effects, so it should not be understood that the scope of rights of the disclosed technology is limited thereby.

According to the intelligent feeding system and method according to the embodiments of the present invention, in the operation of a large fish farm, the feeding point of the fish is reflected by the fish feed rate and the aquatic environment that affects the metabolism recognized by experts in the actual field. By determining the oversupply amount, it is possible to provide a method for preventing the actual waste of feed and the occurrence of fish by quality.

According to the intelligent feeding system and method according to the embodiments of the present invention, by correcting the inferred feeding time and feed amount using the weight of the fish and a pre-determined feed amount schedule, growth promotion of fish and high-quality It can provide a way to ship fish.

1 is a diagram showing a configuration diagram and data flow constituting an intelligent feeding device according to an embodiment of the present invention.
2 is a diagram showing a fish activity data set for labeling and fish behavior learning for completing a trajectory tracking model according to an embodiment of the present invention.
3 is a diagram for explaining tracking and displaying a trajectory of a fish using a fish behavior state trajectory tracking algorithm according to an embodiment of the present invention.
4 is a diagram showing an intelligent fuzzy inference engine of a feed supply amount management system according to an embodiment of the present invention.
5 is a diagram showing a membership function editor for fuzzing input into membership functions according to an embodiment of the present invention.
6 is a diagram showing a fuzzy rule viewer showing input and output fuzzy rules according to an embodiment of the present invention.
7 is a diagram showing the configuration of an intelligent feeder according to an embodiment of the present invention.

1 is a diagram showing a configuration diagram and data flow constituting an intelligent feeding device according to an embodiment of the present invention.

Referring to FIG. 1, the intelligent feed supply system 1 consists of a feed rate extraction unit 100, a feed supply amount management system 200, a feed supply device 300, a sensor 400, and a manager input device 500 do.

The disparity extractor 100 may include a trajectory tracking unit 110 and a trajectory tracking model 120 .

The trajectory tracking unit 110 generates trajectory data (1) from a fish video and transmits it to the trajectory tracking model 120, and the sensor 400 measures the water temperature, oxygen, and fish weight in the tank to obtain temperature data (3). ), dissolved oxygen data (4) and weight data (5) are generated and the generated data is transmitted to the feed supply amount management system 200, and the manager input device 500 inputs the feed supply schedule data (7) from the manager and transmits the data to the feed supply management system 200.

The trajectory tracking model 120 derives the final degree of sedation by classifying the behavior patterns of fish, and trajectory data input using at least one neural network model learned by labeling the degree of sedation defined for each behavior type (9) ) to calculate the degree of subordination data (2).

The feeding amount management system 200 may include a fuzzy reasoning engine 210 and a correction operator 220 . The fuzzy inference engine 200 takes the temperature data (3), dissolved oxygen data (4), and weight data (5) from the sensor 400 as well as the subsurface data (2), which is the result of the trajectory tracking model 120, as inputs. to derive the first feed amount data (6), and the correction operator 220 corrects the feed supply schedule data (7) using the first feed amount data (6) to calculate the second feed amount data (8) and transmitted to the feed supply device 300.

The feeding device 300 controls the internal device to achieve the second feed amount and feed amount data 8 derived from the feed amount management system 200 . For example, it is controlled by combining the limit of the instrument box and the motor speed of the high-pressure blower. The feed supply device 300 can be increased or decreased by 0 to 2 times in units of 1/100 of the target feed amount.

A more accurate model can be completed by updating the trajectory tracking model 120 every feeding cycle of a specific period. In particular, reinforcement learning does not require an exact pair of inputs and outputs, and a balance is formed between the use of known data and the exploration of unknown information, making it suitable for solving problems involving short-term compensation versus long-term. A more accurate model is created through learning through a plurality of images of .

By deriving the feed supply amount using the trajectory tracking model 120, which is a neural network model learned by reflecting the expert's knowledge, and the fuzzy inference engine 210, the derivation process can be easily understood by the user and reliability can be secured.

The overall feed supply frame is determined by the feed supply schedule data 7, which is directly input by the manager or calculated through a separate program, and the feed supply amount management system 200 sets the first feed amount data in the feed supply schedule data 7 (6) is added or subtracted to calculate the second feed amount data (8). For example, according to the feed supply schedule data (7), the feed supply amount is 5 kg at a specific point in time, but if a reduction instruction is derived from the first feed amount data (6), the feed reduced by the indicated reduction in 5 kg is supplied. . This process may be performed automatically, or may further go through a process of receiving confirmation from an administrator.

2 is a diagram showing a fish activity data set for labeling and fish behavior learning for completing a trajectory tracking model according to an embodiment of the present invention.

Referring to FIG. 2 , labeling 610 for completing the fish behavior state trajectory tracking model 120 of the present invention is classified into four types. Here, labeling 610 means a pair of input and output that is the correct answer for learning in machine learning. That is, behavior TYPE A (611) is slow swimming (input) and is defined as normally full (output), behavior TYPE B (612) is fast swimming (input) and is defined as very hungry (output), and behavior TYPE C (613) is static swimming (mouth) and is defined as very full (output), and behavior TYPE D (614) is abnormal swimming (input) and is defined as being sick or stressed (output).

The fish behavior image learning data set 630 may include Type A (631), Type B (632), Type C (633), and Type D (634).

The fish behavior image learning data set 630 includes an Aim step (step 1) that measures the fish trajectory 620 for the first time, a creep step (step 2) that measures the second time, and an attack step that measures it during the third cycle. (step 3), and the return step (step 4), which is measured during the fourth cycle, and the learning classifier of the neural network model classifies the trajectory of the fish measured at each step using the horizontal distance and vertical distance traveled.

For example, in the case where the swimming direction of fish from the Aim stage (Step 1) to the Atack stage (Step 3) increases in the vertical direction, and in the Return stage (Step 4), it shows a descending trajectory after long horizontal movements. Labeled as Behavior Type B (612). In the case of behavior TYPE D (614), it is a case in which a new ATTACK step (step 3) is displayed after continuing to swim horizontally without showing the complete descent shown in the RETURN step (step 4) after the ATTACK step (step 3).

3 is a picture showing the trajectory of a fish tracked and displayed using a fish behavior state trajectory tracking algorithm according to an embodiment of the present invention.

Referring to FIG. 3 , the trajectory tracking unit 110 may track fish from image data 700 and assign object IDs (eg, object ID 1 and object ID 2). The trajectory tracking unit 110 may track fish using various algorithms, for example, a YOLOv3 or YOLOv2 algorithm. In the illustrated example, the YOLOv3 algorithm was used.

The trajectory tracking unit 110 may generate the trajectory data 1 by extracting the optical flow for each object ID from the image data 700 . Optical flow extraction may use various algorithms, such as the Farneback algorithm, the Pyflow algorithm, the EpicFlow algorithm, or the FlowNet algorithm to extract the optical flow. Also, the trajectory tracking unit 110 may measure the size and swimming speed of fish.

4 is a diagram showing an intelligent fuzzy inference engine of a feed supply amount management system according to an embodiment of the present invention.

Intelligent fuzzy inference is an operation method that combines the parallel operation and learning ability of artificial neural networks and the human knowledge expression and explanatory ability of fuzzy systems. The neural network model is an operation method applied when dealing with raw data. can be learned, but it is difficult for users to understand. On the other hand, intelligent fuzzy inference compensated for the disadvantages of neural network models to make it easier for users to understand through expert information. Therefore, by using the intelligent fuzzy reasoning engine, it is possible to learn the expert's high-level data and output a judgment embodying the field expert's experience.

Referring to FIG. 4, the fuzzy inference engine 210 includes a crisp input layer 801, an input membership function 802, a fuzzy rule layer 803, and an output membership. It may include an input membership function (804), a defuzzification layer (805) and an output layer (806). According to an embodiment, the temperature data (3), dissolved oxygen data (4), weight data (5), and permeability data (2) are applied to x1 and x2 of the crisp input layer 801, and the input membership function ( 802) are membership functions associated with inputs (x1, x2) and outputs (y). Outputs of the membership functions may be applied to nodes R (R1, R2, R3, R4, R5, and R6) of the fuzzy rule layer 803. Nodes R of the fuzzy rule layer 803 may apply outputs according to the learned rules to nodes C of the output membership function layer 804, and nodes C(C1, C2) of the output membership function layer 804. ) may be applied to the node of the defuzzification layer 805. A node of the output layer 806 may generate an output y based on inputs applied from the defuzzification layer 805 . According to one embodiment, the node of the output layer 806 may be a single node, and the output y may be generated by adding the inputs applied from the defuzzification layer 805. As described above, the layer of the fuzzy inference engine Structures and aspects of fields and nodes may be designed using various methods.

Fuzzy inference is a formula used when true and false are ambiguous. For example, there is no precise answer to the feeding rate of fish. Feeding amount can vary depending on external circumstances, and it is not efficient to give a specific amount to farmed fish every time. In a word, it is to provide the right amount to the farmed fish according to the situation. Accordingly, the feed supply amount management system 200 can be optimized for result values through fuzzy reasoning. Crisp inputs a specific reference value, compares and analyzes it according to the fuzzy rule, and as a result, it is possible to find an appropriate amount. Through this, it is possible to manage appropriate feed supply for ambiguous results such as feed supply amount.

5 is a diagram showing a membership function editor for fuzzing input into membership functions according to an embodiment of the present invention.

Referring to FIG. 5, in the membership function editor 900, inputs “input 1” 910 and “input 2” 920 are graphs 940 and 950 showing the results of being graded by the membership function. . Here, the input is prepared based on the expert's knowledge data in a linguistic form. Based on this linguistic expert knowledge data, an "if-then rule" is formed and applied to the input and output of the fuzzy inference engine 210 to generate feed supply amount information.

The neuro-fuzzy reasoning system is an artificial intelligence programming method that can infer the number of cases according to various variables and calculate meticulous results by applying the neuro-networking method, which is an artificial intelligence element, to the fuzzy reasoning system, which is an existing intelligent program.

The neuro-fuzzy reasoning system can calculate optimal results according to situations based on the number of complicated and numerous cases like human neural networks.

In the case of the neuro-fuzzy reasoning system applied to the automatic feed supply management system, the optimal automatic feed supply is managed by applying four variables.

Based on the four variables, such as water temperature, dissolved oxygen content, object weight, and activity status, optimally appropriate result values can be calculated from result values for which the correct answer has not been determined. As a result, the optimal result can be output by combining the four variables. A neuro-fuzzy reasoning system can be applied using a membership function editor tool such as Matlab. When each variable range is entered, the result value according to each value is calculated according to the applied function standard. Since there is no set answer to the automatic feed supply amount, if each variable is entered in a range, the numerical value is comprehensively displayed in the form of a graph, and the optimal result value is calculated from the graph as shown in FIG. 5.

As a result, how much feed is supplied is shown by considering the number of cases to determine the optimal value according to water temperature, dissolved oxygen amount, weight of individual, and activity status, and that is calculated as the optimal value and result value shown in the graph.

6 is a diagram showing a fuzzy rule viewer showing input and output fuzzy rules according to an embodiment of the present invention.

Referring to FIG. 6 , the fuzzy rule is a collection of expert knowledge in an “if-then format” and shows the relationship between input and output and explains the operation of the overall system. This is a screen showing the result of the artificial intelligence neuro-fuzzy reasoning system that shows the optimal result when four inputs (1000) are entered. From Input 1 to Input 4 (1000), each variable is input and the optimal output value (1100) is shown according to the function determined comprehensively.

7 is a diagram showing the configuration of an intelligent feeding device according to an embodiment of the present invention.

Referring to FIG. 7, the feed supply device 300 includes a feed container 310, a first slide 320-1, a first limit switch 330-1, an instrument container 340, and a second slide 320-1. , The second limit switch 330-2, the third slide 320-3, the feed supply pipe 350, the high-pressure blower 360, and the control panel 370 (hereinafter referred to as a component device), and the manager terminal ( 380) is connected.

The manager terminal 380 is connected to the feed supply management system 200 and receives feed supply-related information from the feed supply management system 200 .

The manager terminal 380 includes a PC or mobile means, and the mobile monitors the operating state of the intelligent feed supply device 300 in the form of a responsive web app, and also feeds received from the feed supply amount management system 200 Using the supply information, the feed supply device 300 is controlled using the Mqtt protocol. That is, the component devices are controlled according to instructions such as the amount of feed supply, the timing of feed supply, and the stop of feed supply.

In one embodiment of the feed supply control method, first, according to the command of the feed supply management system 200, the third slider 320-3 slides to determine the feed direction (or the tank to be fed), Thereafter, the first slide 320-1 is opened until it comes into contact with the first limit switch 330-1, and the feed is dropped from the feed tank 310 to the measuring tube 340. When feed is accumulated in the measuring box 340 by the weight instructed by the feed supply amount management system 200, the first slider 320-1 is closed and the second slider 320-2 is opened at the same time. The feed accumulated in the second slider 320-2 is dropped, and the dropped feed is supplied to the water tank through the feed supply pipe 350 by the high-pressure blower 360. When the feed is finished, the second slider 320-2 closes and ends the feed.

1: Intelligent feeding system
100: Subsidiarity extraction unit
110: trajectory tracking unit
120: trajectory tracking model
200: feed supply management system
210: fuzzy inference engine
220: correction calculator
300: intelligent feeding device
320-1, 320-2, 320-3: 1st, 2nd, 3rd slides
330-1, 330-2: first and second limit switches
340: instrument box
360: high pressure blower
380: manager terminal
400: sensor
500: manager input device
900: membership function editor

Claims (17)

The behavior of the fish is measured by specific time - the specific time is classified by dividing the measurement start time and measurement end time of the fish trajectory into a plurality of periods -, extracting the fish trajectory image, and extracting the extracted fish trajectory image a susceptibility extracting unit that classifies using a learning classifier - the learning classifier is the horizontal distance and vertical distance that the fish has moved - and calculates the cravings of the fish in a plurality of steps;
A feed supply amount management system that calculates feed amount data to be supplied to the tank by inputting the feed desire of the fish in the plurality of steps calculated by the feed rate extractor and the tank data measured by the plurality of sensors installed in the farm; and
Including a feed supply device for receiving the calculated feed amount data, measuring the weight of feed to be supplied, and supplying it to the tank,
The prey extraction unit classifies the extracted trajectory image of the fish using the learning classifier using machine learning to calculate the need for feeding of the fish in the plurality of stages, and as the plurality of inputs of the machine learning, the specific An intelligent feeding system, characterized in that the trajectory image of the fish measured over time. delete According to claim 1,
The intelligent feeding system, characterized in that the n-th input of the plurality of inputs of the machine learning is the trajectory image of the fish measured at the n-th time among the predetermined specific times. The behavior of the fish is measured by specific time - the specific time is classified by dividing the measurement start time and measurement end time of the fish trajectory into a plurality of periods -, extracting the fish trajectory image, and extracting the extracted fish trajectory image a susceptibility extracting unit that classifies using a learning classifier - the learning classifier is the horizontal distance and vertical distance that the fish has moved - and calculates the cravings of the fish in a plurality of steps;
A feed supply amount management system that calculates feed amount data to be supplied to the tank by inputting the feed desire of the fish in the plurality of steps calculated by the feed rate extractor and the tank data measured by the plurality of sensors installed in the farm; and
Including a feed supply device for receiving the calculated feed amount data, measuring the weight of feed to be supplied, and supplying it to the tank,
The intelligent feeding system further includes a trajectory tracking unit for tracking a specific fish, assigning an object identification number to the prey extraction unit, and generating a trajectory image of the fish using a plurality of algorithms. According to claim 4,
The plurality of algorithms include at least two of YOLOv3, YOLOv2, Farneback, Pyflow, EpicFlow and FlowNet algorithms. According to claim 1 or 4,
The plurality of feeding needs are determined based on feeding degree labeling classified by an expert in advance. According to claim 6,
In the type of the feeding degree labeling, the trajectories of the fish for each of the plurality of times are classified into a plurality of types of patterns and classified into low-speed swimming, high-speed swimming, stationary state, and abnormal swimming activity state. In the case of the high-speed swimming, the need for feeding is defined as hunger, in the case of the stationary swimming, the need for feeding is defined as being very full, and in the case of the abnormal swimming, the condition of the fish is defined as disease or stress. Intelligent feeding system. According to claim 1 or 4,
The feed supply amount management system calculates feed amount data to be supplied to the tank using a fuzzy reasoning engine, and the fuzzy reasoning engine uses a neuro-fuzzy reasoning system to calculate the feed desire of the fish and the measured tank data from the plurality of sensors and an intelligent feeding system for deriving first feed amount data by inputting the expert's knowledge. According to claim 8,
The fuzzy inference engine includes a crisp input layer, an input membership function, a fuzzy rule layer, an input membership function, a defuzzification layer, and an output An intelligent feeding system comprising an output layer. According to claim 8,
The intelligent feeding system including water temperature, dissolved oxygen amount, fish weight, and fish activity status as inputs to the fuzzy reasoning engine. According to claim 8,
The feeding amount management system further comprises a correction calculator for calculating second feed amount data by calculating the first feed amount data and the feeding schedule data input by the farm manager. According to claim 11,
The first feed amount may have a positive feed weight or a negative feed weight according to the feed desire of the fish, the measured tank data from the plurality of sensors, and the expert's knowledge, and the correction calculator determines the first feed An intelligent feeding system for deriving the second feed amount by calculating the amount and the feed weight included in the feed supply schedule data. According to claim 12,
The second feed amount may be equal to, greater than, or less than the feed weight included in the feed schedule. According to claim 1 or 4,
The feeding device is connected to a manager terminal, and the manager terminal controls the feeding device by receiving the feed amount data received by the feed amount management system and using an MQTT protocol. According to claim 13,
The feed supply device is an intelligent feed supply system that accumulates the second feed amount in the feed container by using the opening and closing operation of the slide and then supplies it to the tank using a high-pressure blower while dropping it. According to claim 15,
The feed supply device includes a slide for moving the feed container in which the second amount of feed is accumulated to be coupled to a feed supply pipe connected to a water tank to be fed with the second amount of feed. According to claim 1 or 4,
The tank data includes at least one of temperature data in the tank, dissolved oxygen data, and fish weight data.

Images