System and method for managing aquafarm on basis of deep learning decision-making of ai

Abstract

The present invention relates to a system and a method for managing an aquafarm on the basis of deep learning decision-making of AI. The method for managing an aquafarm on the basis of deep learning decision-making of AI, according to the present invention, comprises steps in which: a data collection unit collects various data required for aquafarming of subjects to be aquafarmed; a data analysis unit receives the data collected by the data collection unit and selects a learning model; the data analysis unit selects parameters to be learned by the learning model after selecting the learning model; the data analysis unit selects a target to be predicted through the learning model having learned the selected parameter to be learned; the data analysis unit inputs data on reference date to be predicted for a target and derives a predicted value (result), after selecting the target to be predicted; and an aquafarm control unit controls the aquafarm on the basis of the data collected by the data collection unit and the analysis result of the data analysis unit. According to the present invention, two or more from among various growth factors, water quality environmental factors and feed supply control factors are selected and controlled through a DNN model and an LSTM model without using an image sensor, and are continuously learned (deep-learned), thereby predicting the optimal feed amount, water quality and the size of the subjects to be aquafarmed.

Description

Aquaculture farm management system and method based on AI deep learning decision-making

[0001] The present invention relates to a fish farm management system and method, and more particularly, to a fish farm management system and method based on AI deep learning decision-making that can predict optimal feed amount, water quality, and shrimp size by selecting and controlling two or more of various growth factors, water quality environmental factors, and feed supply control factors through a DNN (Deep Neural Network) model and an LSTM (Long short-term memory) model, and continuously learning (deep learning) them.

[0002] In general, aquaculture facilities can be broadly classified into land-based aquaculture facilities that cultivate various types of fish in high-density by installing aquaculture tanks indoors in a building, and cage aquaculture facilities that cultivate various types of fish in high-density by installing cage nets and mooring structures over a certain range in a water area such as a lake, river, or sea.

[0003] In land-based and cage-based aquaculture facilities such as the above, the most important aspect of aquaculture management is to provide the food necessary for fish growth, that is, feed, at appropriate times on a regular basis.

[0004] Traditionally, feed supply was done manually, with farm managers physically dispatched to the site, scooping up feed from storage bins with buckets or other implements and distributing it within the fish tanks or nets of the cages. While this isn't a significant issue for onshore farms with permanent staff, offshore cage farms, where staff are not readily available, require managers to travel by boat. This, coupled with fuel costs and the labor costs associated with hiring specialized personnel, places a significant financial burden on farm management.

[0005] Accordingly, the trend is moving towards automated feed supply devices replacing on-site managers. Representative examples of conventional automatic feed supply devices for aquaculture farms include a blower-type feed supply device that uses a rotary valve as a gate at the bottom exit of a feed storage container manufactured in the form of a hopper to discharge and block feed, and a screw-type feed supply device that directly connects a screw conveyor to the bottom exit of a feed storage container to both input and transport feed.

[0006] However, the conventional feed supply device described above cannot guarantee a uniform amount of feed per hour, and the amount of feed discharged varies depending on how the gear or screw is positioned at the feed discharge port, making it difficult to supply feed in a fixed-quantity discharge manner, which is the optimal feeding condition.

[0007] Meanwhile, Korean Patent Publication No. 10-2528286 (Patent Document 1) discloses an "AI-based automatic feeding system for a fish farm and a method thereof," wherein the AI-based automatic feeding system for a fish farm comprises: a water quality input unit (110) that includes a sensing means installed inside a fish tank of a fish farm and receives water quality-related information for the fish farm from the sensing means at preset time intervals; and an image input unit (120) that includes a photographing unit installed inside and outside the fish tank of a fish farm and receives, from each photographing unit, information on fish movement captured underwater and information on fish movement captured on the water surface as image-related information; a data input unit (100) that uses stored first and second AI models to input information by the data input unit (100) and extract information on the amount of feed supplied to the fish farm; And it is characterized by comprising a supply control unit (300) that generates a control signal for controlling a feed supply means in a linked aquaculture farm based on the feed supply amount information by the AI ​​analysis unit (200).

[0008] In the case of Patent Document 1 as described above, although it has the advantage of being able to determine the optimal feed supply amount by using information obtained while operating a fish farm (water quality information of the fish farm and movement status information of fish within the fish farm), it is ultimately focused on determining the feed supply amount, and thus has the problem of being relatively vulnerable to predicting water quality or the size of the fish farm object.

[0009] -Patent Document 1-

[0010] Korean Patent Publication No. 10-2528286 (Published on May 3, 2023)

[0011] The present invention was created by comprehensively considering the above matters, and the purpose of the present invention is to provide an aquaculture management system and method based on AI deep learning decision-making that can predict the optimal feed amount, water quality, and size of aquaculture target by selecting and controlling two or more of various growth factors, water quality environmental factors, and feed supply control factors through a DNN (Deep Neural Network) model and an LSTM (Long short-term memory) model without using an image sensor, and continuously learning (deep learning) the same.

[0012] In order to achieve the above purpose, the aquaculture farm management system based on AI deep learning decision-making according to the present invention is:

[0013] A data collection unit that collects various data required for the form of the form target and sends control commands for controlling the form farm;

[0014] A data analysis unit that analyzes the data collected by the data collection unit; and

[0015] Includes a farm control unit that controls the farm based on data collected by the data collection unit and data analyzed by the data analysis unit,

[0016] The above data analysis unit selects a learning model based on the data collected by the data collection unit, then selects parameters to be learned by the learning model, selects a target to be predicted through the learning model that has learned the selected parameters to be learned, and after selecting the target to be predicted, inputs data on a reference date to be predicted for the target to derive a predicted value (result), and the aquaculture control unit derives optimal results for various targets including feed amount, solution dosage, and stocking density based on the data collected by the data collection unit and the predicted value (result) derived by the data analysis unit, thereby automatically controlling to accelerate the growth of aquaculture targets, minimize mortality, and maintain optimal water quality.

[0017] Here, the data collected by the data collection unit may include water quality information of the water in the tank where the aquaculture target is cultivated, such as water temperature, DO (dissolved oxygen), pH, salinity concentration, ammonia, nitrite, nitrate, and CO2.

[0018] In addition, the data collection unit may be equipped with an application for data collection and aquaculture control that collects various data required for the form of the aquaculture target, provides convenient functions, and sends control commands for aquaculture control.

[0019] At this time, the data collection and form control app (application) may be equipped with an operation guide function that summarizes and guides which stage of the current form stage it belongs to and how to manage the schedule, and a quest-type task management function that presents tasks to be done in the form of quests and tracks and manages the tasks to be done.

[0020] At this time, a search function can also be added to allow searching for detailed formatting methods by linking to the manual when necessary.

[0021] In addition, the data analysis unit may be equipped with an AI program that receives data collected by the data collection unit, selects a learning model, selects parameters to be learned, selects a target to be predicted, inputs data on a reference date to be predicted, and derives a predicted value.

[0022] In addition, the above-mentioned form control unit,

[0023] Main controller board that comprehensively controls various devices and controllers installed in the farm;

[0024] A tank controller electrically connected to the above main controller board, installed in a tank where aquaculture objects are cultured, and controlling various aquaculture-related devices installed directly in the tank or installed around the tank;

[0025] A plurality of input / output boards electrically connected to the above tank controller, receiving control signals from the above tank controller and outputting them to various aquaculture-related devices installed directly in the tank or installed around the tank;

[0026] A sensor connected to the input / output board and receiving a control signal from the tank controller through the input / output board to measure DO, water temperature, pH, salinity concentration, etc. of water in the tank;

[0027] A feeder connected to the input / output board and automatically supplies feed (food) into the tank by receiving a control signal from the tank controller through the input / output board;

[0028] A nutrient solution device connected to the input / output board and receiving a control signal from the tank controller through the input / output board to perform water quality management functions such as pH control of water in the tank, removal of nitrogen pollutants, mineral control, and administration of useful microorganisms; and

[0029] It may be configured to include a circulation filtration system that is connected to the input / output board and receives a control signal from the tank controller through the input / output board to periodically filter impurities present in the water in the tank.

[0030] Here, the sensor may include a DO sensor, a water temperature sensor, a pH sensor, and a salt concentration sensor.

[0031] In addition, the nutrient solution device includes a first nutrient solution device that injects a solution for adjusting the pH of the water in the tank, a second nutrient solution device that injects a solution for removing nitrogen pollutants, a third nutrient solution device that injects a solution for adjusting minerals, a fourth nutrient solution device that injects useful microorganisms, and a fourth nutrient solution device that automatically injects seawater according to a decrease in salinity concentration.

[0032] It may include a fifth nutrient solution.

[0033] In addition, in order to achieve the above purpose, the method for managing aquaculture farms based on AI deep learning decision-making according to the present invention,

[0034] An AI deep learning decision-making-based aquaculture farm management method based on an AI deep learning decision-making-based aquaculture farm management system including a data collection unit, a data analysis unit, and a farm control unit,

[0035] a) A step in which the data collection unit collects various data required for the form of the form object;

[0036] b) A step in which the data analysis unit receives data collected by the data collection unit and selects a learning model;

[0037] c) A step in which the data analysis unit selects parameters to be learned by the learning model after selecting the learning model;

[0038] d) A step in which the data analysis unit selects a target to be predicted through the learning model that has learned the selected learning parameters;

[0039] e) A step in which the data analysis unit, after selecting the target to be predicted, inputs data on the reference date to be predicted for the target and derives a predicted value (result); and

[0040] f) It is characterized in that it includes a step in which the aquaculture control unit controls the aquaculture unit based on data collected by the data collection unit and analysis results by the data analysis unit.

[0041] Here, the data collected by the data collection unit in the step a) above may include water quality information of the water in the tank in which the aquaculture target is cultivated, such as water temperature, DO (dissolved oxygen), pH, salinity concentration, ammonia, nitrite, nitrate, and CO2.

[0042] In addition, in step b), when the data analysis unit receives data collected by the data collection unit, the data collection unit can input the collected data into the data analysis unit by uploading a CSV (Comma Separated Value) file.

[0043] Additionally, in the above step b), the learning model may include a DNN (Deep Neural Network) and an LSTM (Long short-term memory).

[0044] In addition, the parameters to be learned by the learning model in the step c) above may include 'age (aquaculture period)', 'size (size)', 'feed (feed amount)', 'DO (dissolved oxygen)', 'pH', 'TEMP (water temperature)', 'SALT (salt concentration)', 'ammonia (ammonia)', 'no2 (nitrite)', 'no3 (nitrate)', 'feed Num (number of times feed is supplied)', etc.

[0045] In addition, the target to be predicted in the above step d) may include 'age (culture period)', 'size', 'feed (feed amount)', 'DO (dissolved oxygen)', 'pH', 'TEMP (water temperature)', 'SALT (salt concentration)', 'ammonia (ammonia)', 'no2 (nitrite)', 'no3 (nitrate)', 'feed Num (number of times feed is supplied)', etc.

[0046] According to the present invention, without using an image sensor, two or more of various growth factors, water quality environmental factors, and feed supply control factors are selected and controlled through a DNN (Deep Neural Network) model and an LSTM (Long short-term memory) model, and by continuously learning (deep learning) them, there is an advantage in that the optimal feed amount, water quality, and size of the water to be aquacultured can be predicted.

[0047] Figure 1 is a schematic diagram showing the configuration of a fish farm management system based on AI deep learning decision-making according to the present invention.

[0048] Figure 2 is a flowchart showing the execution process of a method for managing aquaculture farms based on AI deep learning decision-making according to the present invention.

[0049] FIG. 3 is a drawing showing an operation guide and a screen for quest-type task management in an app for data collection and form control employed in the present invention.

[0050] FIG. 4 is a drawing showing a screen for checking the status of a tank and a screen for checking detailed status of a tank in an app for data collection and aquaculture control employed in the present invention.

[0051] FIG. 5 is a drawing showing a water quality information screen and a sensor data confirmation screen in an app for data collection and aquaculture control employed in the present invention.

[0052] Figure 6 is a drawing showing the screens for real-time equipment control, feeder control, and nutrient solution control in the app for data collection and aquaculture farm control employed in the present invention.

[0053] Figure 7 is a drawing showing the water quality information recording and feeding information recording screens in the data collection and aquaculture farm control app employed in the present invention, respectively.

[0054] Figure 8 is a drawing showing an additive dosing record screen in an app for data collection and form control employed in the present invention.

[0055] Figure 9 is a diagram showing a screen for uploading a learning data file, selecting a learning model, selecting parameters to be learned, selecting a prediction target, and setting a prediction date in an AI program employed in the present invention.

[0056] Figure 10 is a diagram showing the input data of the reference date, the predicted result value, and the learned learning data on the screen after learning in the AI ​​program employed in the present invention.

[0057] FIG. 11 is a diagram showing a data file to be learned provided from a data collection unit to a data analysis unit of a fish farm management system based on AI deep learning decision-making according to the present invention.

[0058] According to one embodiment of the present invention, a fish farm management system based on AI deep learning decision-making includes a data collection unit that collects various data necessary for the cultivation of aquaculture objects and transmits a control command for controlling a fish farm, a data analysis unit that analyzes the data collected by the data collection unit, and a fish farm control unit that controls a fish farm based on the data collected by the data collection unit and the data analyzed by the data analysis unit, wherein the data analysis unit selects a learning model based on the data collected by the data collection unit, selects a parameter to be learned by the learning model, selects a target to be predicted through the learning model that has learned the selected parameter to be learned, and after the target to be predicted is selected, data on a reference date for prediction of the target is input to derive a predicted value (result), and the fish farm control unit derives an optimal result for various targets including feed amount, solution dosage, and stocking density based on the data collected by the data collection unit and the predicted value (result) derived by the data analysis unit, thereby automatically controlling the growth of aquaculture objects to be faster, minimize mortality, and maintain optimal water quality.

[0059] Terms or words used in this specification and claims should not be construed as limited to their usual or dictionary meanings, but should be construed as meanings and concepts that conform to the technical spirit of the present invention, based on the principle that the inventor can appropriately define the concept of the term to best explain his or her own invention.

[0060] Throughout the specification, when a part is said to "include" a component, this does not exclude other components, but rather implies that other components may be included, unless otherwise specifically stated. Furthermore, terms such as "part," "unit," "module," and "device" used in the specification mean a unit that processes at least one function or operation, which may be implemented using hardware, software, or a combination of hardware and software.

[0061] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

[0062] FIG. 1 is a diagram schematically illustrating the configuration of a fish farm management system based on AI deep learning decision-making according to an embodiment of the present invention.

[0063] Referring to FIG. 1, the aquaculture management system (100) based on AI deep learning decision-making according to the present invention is configured to include a data collection unit (110), a data analysis unit (120), and aquaculture control unit (130).

[0064] The data collection unit (110) collects various data required for the cultivation of aquaculture objects (e.g., whiteleg shrimp) and transmits control commands for aquaculture farm control. Here, the data collected by the data collection unit (110) may include water quality information of the water (e.g., seawater + freshwater) in the tank where the aquaculture objects are cultivated, such as water temperature, DO (dissolved oxygen), pH, salinity, ammonia, nitrite, nitrate, and CO2.

[0065] In addition, the data collection unit (110) as described above may be equipped with a data collection and aquaculture control app (110a) that collects various data required for the form of the aquaculture target, provides convenient functions, and transmits control commands for aquaculture control.

[0066] At this time, the data collection and aquaculture control app (application) may be equipped with an operation guide function that summarizes and guides which stage of the current aquaculture stage it belongs to and how to manage the schedule, as shown in (a) of Fig. 3, and a task management function in the form of a quest that presents tasks to be done in the aquaculture field in the form of a quest to track and manage the tasks, as shown in (b) of Fig. 3. At this time, a search function that allows for searching for detailed aquaculture methods by linking to a manual when necessary may be additionally equipped.

[0067] The data analysis unit (120) analyzes the data collected by the data collection unit (110). The data analysis unit (120) may be equipped with an AI program (120a) that receives the data collected by the data collection unit (110), selects a learning model, selects parameters to be learned, selects a target to be predicted, inputs data for a reference date to be predicted, and derives a predicted value.

[0068] The aquaculture farm control unit (130) controls the aquaculture farm based on data collected by the data collection unit (110) and data analyzed by the data analysis unit (120).

[0069] In the aquaculture management system (100) based on AI deep learning decision-making having the above configuration, the data analysis unit (120) selects a learning model based on data collected by the data collection unit (110), then selects parameters to be learned by the learning model, selects a target to be predicted through the learning model that has learned the selected parameters to be learned, and after selecting the target to be predicted, inputs data on a reference date for prediction for the target to derive a predicted value (result), and the aquaculture control unit (130) derives optimal results for various targets including feed amount, solution dosage, and stocking density based on the data collected by the data collection unit (110) and the predicted value (result) derived by the data analysis unit (120), thereby automatically controlling the aquaculture target to grow faster, minimize mortality, and maintain optimal water quality.

[0070] Meanwhile, the above-described aquaculture control unit (130) comprises a main controller board (131) that comprehensively controls various devices and controllers installed in the aquaculture farm; a tank controller (132) that is electrically connected to the main controller board (131) and installed in a tank in which aquaculture objects are cultivated, and controls various aquaculture-related devices installed directly in the tank or installed around the tank; a plurality of input / output boards (133) that are electrically connected to the tank controller (132) and receive control signals from the tank controller (132) and output them to various aquaculture-related devices installed directly in the tank or installed around the tank; a sensor (134) that is connected to the input / output board (133) and receives control signals from the tank controller (132) through the input / output board (133) and measures DO, water temperature, pH, salinity concentration, etc. of water in the tank; It may be configured to include a feeder (135) that is connected to the input / output board (133) and automatically supplies feed (food) into the tank by receiving a control signal from the tank controller (132) through the input / output board (133); a nutrient solution device (136) that is connected to the input / output board (133) and receives a control signal from the tank controller (132) through the input / output board (133) and performs water quality management functions such as pH control of water in the tank, removal of nitrogen pollutants, mineral control, and administration of useful microorganisms; and a circulation filtration system (137) that is connected to the input / output board (133) and receives a control signal from the tank controller (132) through the input / output board (133) and periodically filters impurities present in the water in the tank.

[0071] Here, a plurality of (e.g., 15) tank controllers (132) can be connected to the main controller board (131). In addition, a plurality of (e.g., 15) input / output boards (133) can be connected to the tank controllers (132). In addition, each input / output board (133) can be connected to one device or sensor.

[0072] The above sensor (134) may include a DO sensor, a water temperature sensor, a pH sensor, and a salt concentration sensor. In addition, the nutrient solution device (136) may include a first nutrient solution device that injects a solution (additive) for adjusting the pH of the water in the tank, a second nutrient solution device that injects a solution (additive) for removing a nitrogen pollutant, a third nutrient solution device that injects a solution (additive) for adjusting minerals, a fourth nutrient solution device that injects useful microorganisms, and a fifth nutrient solution device that automatically injects seawater as the salt concentration decreases.

[0073] Then, below, we will explain an AI deep learning decision-making-based aquaculture farm management method based on an AI deep learning decision-making-based aquaculture farm management system according to the present invention having the above-described configuration.

[0074] Figure 2 is a flowchart showing the execution process of a method for managing aquaculture farms based on AI deep learning decision-making according to an embodiment of the present invention.

[0075] Referring to FIGS. 1 and 2, the aquaculture management method based on AI deep learning decision-making according to the present invention is an AI deep learning decision-making-based aquaculture management method based on AI deep learning decision-making-based aquaculture management system (100) including the data collection unit (110), data analysis unit (120), and aquaculture control unit (130) as described above. First, the data collection unit (110) collects various data necessary for the aquaculture of aquaculture objects (e.g., whiteleg shrimp) (step S201). Here, the data collected by the data collection unit (110) may include water quality information of the water in the tank where the aquaculture objects are aquacultured, such as water temperature, DO (dissolved oxygen), pH, salinity concentration, ammonia, nitrite, nitrate, and CO2.

[0076] When data collection is completed by the data collection unit (110), the data analysis unit (120) receives the data collected by the data collection unit (110) and selects a learning model (step S202). Here, when the data analysis unit (120) receives the data collected by the data collection unit (110), the data collection unit (110) can input the collected data into the data analysis unit (120) by uploading a CSV (Comma Separated Value) file. In addition, the learning model may include a DNN (Deep Neural Network) and an LSTM (Long short-term memory). In this embodiment, it is assumed that LSTM is selected as the learning model.

[0077] As described above, after the data analysis unit (120) selects a learning model, it selects parameters to be learned by the learning model (step S203). Here, the parameters to be learned by the learning model may include 'age (culture period)', 'size (size)', 'feed (feed amount)', 'DO (dissolved oxygen)', 'pH', 'TEMP (water temperature)', 'SALT (salinity)', 'ammonia (ammonia)', 'no2 (nitrite)', 'no3 (nitrate)', 'feed Num (number of feed supplies)', etc. This will be explained again later.

[0078] When the selection of the parameters to be learned is completed as described above, the data analysis unit (120) selects a target to be predicted through the learning model that learned the selected parameters to be learned (step S204). Here, the target to be predicted may include 'age (culture period)', 'size (size)', 'feed (feed amount)', 'DO (dissolved oxygen)', 'pH', 'TEMP (water temperature)', 'SALT (salinity)', 'ammonia (ammonia)', 'no2 (nitrite)', 'no3 (nitrate)', 'feed Num (number of times feed is supplied)', etc.

[0079] Afterwards, the data analysis unit (120) selects the target to be predicted, inputs data on the reference date for the target to be predicted, and derives a predicted value (result) (step S205).

[0080] Accordingly, the aquaculture control unit (130) controls the aquaculture farm based on the data collected by the data collection unit (110) and the analysis results by the data analysis unit (120) (step S206). That is, as described above, the aquaculture control unit (130) automatically controls the aquaculture farm to derive optimal results for various targets including feed amount, solution dosage, and stocking density based on the data collected by the data collection unit (110) and the predicted values ​​(results) derived by the data analysis unit (120) so as to accelerate the growth of the aquaculture target, minimize the amount of mortality, and maintain optimal water quality.

[0081] Below, a further explanation will be provided regarding the aquaculture management system and method based on AI deep learning decision-making according to the present invention.

[0082] FIG. 4 is a drawing showing a screen for checking the tank status and a screen for checking the detailed status of a tank in an app for data collection and aquaculture control employed in the present invention.

[0083] Referring to FIG. 4, (a) is a screen for checking the tank status, and when the sensor data exceeds the set value or there is no signal from the board, an abnormality can be determined through a color change. (b) is a screen for checking the detailed status of the tank, and this is a screen for checking the status of the equipment and sensors connected to the tank controller (132) (see FIG. 1). In FIG. 4 (b), the status of the pH sensor, NO3 sensor, NH3 sensor, and Salinity sensor connected to the tank controller (132) No. 01 are respectively shown.

[0084] Figure 5 is a drawing showing a water quality information screen and a sensor data confirmation screen in an app for data collection and aquaculture control employed in the present invention.

[0085] Referring to Figure 5, (a) is a water quality information screen, and as shown, eight types of water quality information can be collected: water temperature, DO, pH, salinity, ammonia, nitrite, nitrate, and CO2. Of course, the water quality information is not limited to the above eight types, and the water quality information that can be collected can also increase as the types of sensors are increased. (b) is a sensor data confirmation screen, and as shown, data accumulated through the sensors can be confirmed in a graph.

[0086] Figure 6 is a drawing showing the screens for real-time equipment control, feeder control, and nutrient solution control in the app for data collection and aquaculture farm control employed in the present invention.

[0087] Referring to FIG. 6, (a) is a screen for real-time equipment control, and equipment (e.g., a feeder) can be controlled in real time through the data collection and aquaculture control app (110a) employed in the present invention (see FIG. 1). (b) is a screen for controlling a feeder, and scheduled feeding and immediate feeding of the feeder are possible through the data collection and aquaculture control app (110a). (c) is a screen for controlling a nutrient solution device, and scheduled and immediate administration of solution administration is possible through the data collection and aquaculture control app (110a), thereby enabling water quality management.

[0088] Figure 7 is a drawing showing the water quality information recording and feeding information recording screens in the data collection and aquaculture farm control app employed in the present invention, respectively.

[0089] Referring to Figure 7, (a) is the water quality information recording screen, where sensor values ​​can be retrieved and water quality information recorded. At this time, values ​​directly measured through the kit, excluding the sensor, can also be recorded. (b) is the feeding information recording screen, where feed feeding time, feed amount, feed type, etc. can be recorded.

[0090] Figure 8 is a drawing showing an additive dosing record screen in an app for data collection and form control employed in the present invention.

[0091] Referring to Figure 8, this is an additive administration record screen where the additive administration time, additive administration amount, and additive type can be recorded.

[0092] Figure 9 is a diagram showing a screen for uploading a learning data file, selecting a learning model, selecting parameters to be learned, selecting a prediction target, and setting a prediction date in an AI program employed in the present invention.

[0093] Referring to Fig. 9, ① represents the upload of the learning data file, ② represents the selection of LSTM as the learning model selection (DNN, LSTM), and ③ represents the selection of parameters to be learned. At this time, the parameters are automatically loaded according to the rows of the learning file (see ① in Fig. 11). This example screen shows that 'age (aquaculture period)', 'size (size)', 'feed (feed amount)', 'DO (dissolved oxygen)', 'pH', 'TEMP (water temperature)', 'SALT (salinity)', 'ammonia (ammonia)', 'no2 (nitrite)', 'no3 (nitrate)', and 'feed Num (number of feed supplies)' have been selected as parameters. In addition, ④ is a prediction target selection table. For example, if you want to predict shrimp size, you can use 'Size', if you want to predict feed amount, you can use 'feed', and so on. You can make predictions by target as desired. ⑤ is the predicted date. For example, if you set the number '5', the predicted result for 5 days later will be displayed.

[0094] Figure 10 is a diagram showing the input data of the reference date, the predicted result value, and the learned learning data on the screen after learning in the AI ​​program employed in the present invention.

[0095] Referring to Figure 11, ① shows that data for the reference date to be predicted is entered in the prediction tab after learning, ② shows a status window for checking the predicted result value, and ③ shows a window for checking the learned learning data.

[0096] Figure 11 is a diagram showing a data file to be learned provided from a data collection unit to a data analysis unit of a fish farm management system based on AI deep learning decision-making according to the present invention.

[0097] Referring to Fig. 11, when the learning file is submitted to the AI ​​program (120a) (see Fig. 1), the first column (①) indicated by the red box is recognized as a parameter, and the parameter to be predicted among the parameters is selected to perform subsequent operations. At this time, the prediction model can be analyzed by selecting two models, DNN and LSTM.

[0098]

[0099] As described above, the aquaculture management system and method based on AI deep learning decision-making according to the present invention has the advantage of being able to predict the optimal feed amount, water quality, and size of aquaculture target by selecting and controlling two or more of various growth factors, water quality environmental factors, and feed supply control factors through a DNN (Deep Neural Network) model and an LSTM (Long short-term memory) model without using an image sensor and continuously learning (deep learning) them.

[0100] In addition, by introducing AI programs, if sufficient data is secured, anything desired, such as water quality, feed amount, and growth prediction, can be predicted, and the parameters to be analyzed can be set as desired, so there is an advantage in being able to derive the optimal conclusion for aquaculture through various analyses.

[0101] In addition, it has the advantage of increasing convenience by naturally inducing data entry through the form of an operation log and quests, including functions that can help with the operation of the farm (e.g., automatic feeder, nutrient solution machine, etc.) rather than simply receiving data.

[0102] While the present invention has been described in detail through preferred embodiments, it is not limited thereto, and it will be apparent to those skilled in the art that various modifications and applications can be made without departing from the technical spirit of the present invention. Therefore, the true scope of protection of the present invention should be interpreted by the following claims, and all technical ideas within the scope equivalent thereto should be construed as being included within the scope of the rights of the present invention.

[0103] -Explanation of the symbol-

[0104] 100: (The present invention) Fish farm management system based on AI deep learning decision-making

[0105] 110: Data Collection Department

[0106] 120: Data Analysis Department

[0107] 120a: AI program 110a: App for data collection and farm control

[0108] 130: Formwork Control Unit

[0109] 131: Main controller board

[0110] 132: Tank Controller

[0111] 133: Input / Output Board

[0112] 134: Sensor

[0113] 135: Grade

[0114] 136: Nutrient solution machine

[0115] 137: Circulating filtration system

[0116] The present invention can be used in a fish farm management system and a fish farm management method.

Claims

1. A data collection unit that collects various data required for the form of the form target and sends control commands for the form control; A data analysis unit that analyzes data collected by the data collection unit; and a farm control unit that controls the farm based on the data collected by the data collection unit and the data analyzed by the data analysis unit. The data analysis unit selects a learning model based on data collected by the data collection unit, and then selects parameters to be learned by the learning model, selects a target to be predicted through the learning model that has learned the selected parameters to be learned, and after selecting the target to be predicted, inputs data on a reference date for prediction for the target to derive a predicted value (result), and the aquaculture control unit derives optimal results for various targets including feed amount, solution dosage, and stocking density based on the data collected by the data collection unit and the predicted value (result) derived by the data analysis unit, thereby automatically controlling the aquaculture management system to accelerate the growth of aquaculture targets, minimize mortality, and maintain optimal water quality.

2. In paragraph 1, The data collected by the above data collection unit is water quality information of the water in the tank where the aquaculture target is cultivated, including water temperature, DO (dissolved oxygen), pH, salinity concentration, ammonia, nitrite, nitrate, and CO2. An aquaculture farm management system based on AI deep learning decision-making.

3. In paragraph 1, An AI deep learning decision-making-based aquaculture management system characterized in that the above data collection unit is equipped with a data collection and aquaculture control app that collects various data necessary for the aquaculture of aquaculture objects, provides convenient functions, and sends control commands for aquaculture control.

4. In paragraph 3, The above data collection and form control app has an operation guide function that summarizes which stage of the current form stage it is in and how to manage the schedule, and presents the tasks to be done in the form of a quest to track the tasks to be done. An AI deep learning decision-making-based farm management system characterized by having a task management function in the form of a quest and management.

5. In paragraph 4, An AI deep learning decision-making-based aquaculture management system characterized in that the above data collection and aquaculture control app further has a search function that allows searching for detailed aquaculture methods by linking to a manual when necessary.

6. In paragraph 1, An AI deep learning decision-making-based aquaculture management system characterized in that the data analysis unit is equipped with an AI program that receives data collected by the data collection unit, selects a learning model, selects parameters to be learned, selects a target to be predicted, inputs data on a reference date to be predicted, and derives a predicted value.

7. In paragraph 1, The above form control unit is, Main controller board that comprehensively controls various devices and controllers installed in the farm; A tank controller electrically connected to the above main controller board, installed in a tank where aquaculture objects are cultured, and controlling various aquaculture-related devices installed directly in the tank or installed around the tank; A plurality of input / output boards electrically connected to the above tank controller, receiving control signals from the above tank controller and outputting them to various aquaculture-related devices installed directly in the tank or installed around the tank; A sensor connected to the input / output board and receiving a control signal from the tank controller through the input / output board to measure DO, water temperature, pH, salinity concentration, etc. of water in the tank; A feeder connected to the input / output board and automatically supplies feed (food) into the tank by receiving a control signal from the tank controller through the input / output board; A nutrient solution device connected to the input / output board and receiving a control signal from the tank controller through the input / output board to perform water quality management functions such as pH control of water in the tank, removal of nitrogen pollutants, mineral control, and administration of useful microorganisms; and An AI deep learning decision-making-based aquaculture management system characterized in that it comprises a circulation filtration system that is connected to the input / output board and receives a control signal from the tank controller through the input / output board to periodically filter impurities present in the water in the tank.

8. In paragraph 7, The above sensor is an AI deep learning decision-making-based aquaculture farm management system including a DO sensor, a water temperature sensor, a pH sensor, and a salinity sensor.

9. In paragraph 7, The above nutrient solution device is an AI deep learning decision-making-based aquaculture farm management system including a first nutrient solution device that injects a solution for controlling the pH of water in the tank, a second nutrient solution device that injects a solution for removing nitrogen pollutants, a third nutrient solution device that injects a solution for controlling minerals, a fourth nutrient solution device that injects useful microorganisms, and a fifth nutrient solution device that automatically injects seawater according to a decrease in salinity concentration.

10. An AI deep learning decision-making-based aquaculture farm management system that includes a data collection unit, a data analysis unit, and a farm control unit. As a management method, a) A step in which the data collection unit collects various data required for the form of the form object; b) A step in which the data analysis unit receives data collected by the data collection unit and selects a learning model; c) A step in which the data analysis unit selects parameters to be learned by the learning model after selecting the learning model; d) A step in which the data analysis unit selects a target to be predicted through the learning model that has learned the selected learning parameters; e) A step in which the data analysis unit, after selecting the target to be predicted, inputs data on the reference date to be predicted for the target and derives a predicted value (result); and f) An AI deep learning decision-making-based fish farm management method, comprising a step of controlling the fish farm based on data collected by the data collection unit and analysis results by the data analysis unit.

11. In paragraph 10, The data collected by the data collection unit in the above step a) is water quality information of the water in the tank where the aquaculture target is cultivated, including water temperature, DO (dissolved oxygen), pH, salinity concentration, ammonia, nitrite, nitrate, and CO2. An aquaculture farm management method based on AI deep learning decision-making.

12. In paragraph 10, A method for managing aquaculture farm based on AI deep learning decision-making, characterized in that in step b), the data analysis unit inputs the data collected by the data collection unit, and the data collection unit inputs the collected data into the data analysis unit by uploading a CSV (Comma Separated Value) file.

13. In paragraph 10, In the above step b), the learning model is an AI deep learning decision-making-based aquaculture farm management method including a DNN (Deep Neural Network) and an LSTM (Long short-term memory).

14. In paragraph 10, In the above step c), the parameters to be learned by the learning model include 'age (aquaculture period)', 'size', 'feed (feed amount)', 'DO (dissolved oxygen)', 'pH', 'TEMP (water temperature)', 'SALT (salinity)', 'ammonia (ammonia)', 'no2 (nitrite)', 'no3 (nitrate)', and 'feed Num (number of times feed is supplied)'.

15. In paragraph 10, In the above step d), the target to be predicted includes 'age (aquaculture period)', 'size (size)', 'feed (feed amount)', 'DO (dissolved oxygen)', 'pH', 'TEMP (water temperature)', 'SALT (salinity)', 'ammonia (ammonia)', 'no2 (nitrite)', 'no3 (nitrate)', and 'feed Num (number of times feed is supplied)'.

Images