Water quality prediction and intelligent regulation and control system based on machine learning

Abstract

The invention provides a water quality prediction and intelligent regulation and control system based on machine learning, and the system comprises a real-time collection module which is used for collecting the current water quality data of a culture pond in real time; the water quality prediction module is used for determining a water quality prediction result according to the current water quality data based on a pre-trained water quality prediction model; and the intelligent regulation and control module is used for carrying out adaptive regulation and control on a future water quality regulation plan and a bait feeding plan of the culture pond based on the water quality prediction result. Based on a machine learning algorithm, a water quality prediction model is trained, based on the water quality prediction model, a water quality prediction result is determined according to current water quality data, and based on the water quality prediction result, a future water quality regulation plan and a bait feeding plan of a culture pond are adaptively regulated and controlled. According to the invention, accurate water quality management and bait feeding management are automatically carried out on the culture pond, the labor cost is reduced, real-time accurate regulation and control are realized, and the aquaculture efficiency and the aquaculture quality are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and in particular to a water quality prediction and intelligent control system based on machine learning. Background Art

[0002] At present, in the aquaculture industry, the water quality of breeding ponds not only directly affects the growth, reproduction and health of aquatic animals, but is also closely related to the ecological balance of the breeding environment. The feed for aquatic animals is also determined by the water quality.

[0003] Traditional water quality management and bait delivery management mainly rely on manual monitoring and manual experience judgment. The operation is cumbersome, the labor cost is high, and it is difficult to achieve real-time and precise regulation, which has a great impact on breeding efficiency and aquatic product quality.

[0004] Therefore, a solution is urgently needed. Summary of the invention

[0005] One of the purposes of the present invention is to provide a water quality prediction and intelligent control system based on machine learning. Based on the machine learning algorithm, a water quality prediction model is trained, and the current water quality data of the breeding pond is collected in real time. Based on the water quality prediction model, the water quality prediction result is determined according to the current water quality data. Based on the water quality prediction result, the future water quality adjustment plan and feed feeding plan of the breeding pond are adaptively controlled to realize automated and precise water quality management and feed delivery management of the breeding pond, thereby reducing labor costs, realizing real-time and precise control, and improving aquaculture efficiency and aquaculture quality.

[0006] The embodiment of the present invention provides a water quality prediction and intelligent control system based on machine learning, comprising:

[0007] Real-time data collection module, used to collect the current water quality data of the breeding pond in real time;

[0008] A water quality prediction module is used to determine the water quality prediction result based on the pre-trained water quality prediction model and the current water quality data;

[0009] Intelligent control module, used to adaptively control the future water quality adjustment plan and bait feeding plan of the aquaculture pond based on the water quality prediction results;

[0010] Among them, the pre-training steps of the water quality prediction model are as follows:

[0011] Obtain a large amount of historical water quality data from different aquaculture ponds;

[0012] Preprocess historical water quality data;

[0013] Based on the machine learning algorithm, the water quality prediction model is trained according to the preprocessed historical water quality data.

[0014] Optionally, the real-time acquisition module collects the current water quality data of the breeding pond in real time, including:

[0015] The current water quality data of the breeding pond is collected in real time through the sensor network arranged in the breeding pond.

[0016] Optionally, the current water quality data includes at least: salinity, temperature, pH value, and dissolved oxygen.

[0017] Optionally, the preprocessing of historical water quality data includes:

[0018] The historical water quality data are processed in turn by removing missing values, removing outliers and standardizing.

[0019] Optional, machine learning-based water quality prediction and intelligent control system, also includes:

[0020] Optimization modules for:

[0021] Obtain the error conditions determined by the water quality prediction results;

[0022] Based on the error situation, the water quality prediction model is optimized.

[0023] Optionally, the intelligent control module adaptively controls the future water quality adjustment plan and bait feeding plan of the aquaculture pond based on the water quality prediction result, including:

[0024] Performing feature description on the water quality prediction result to obtain a first feature description vector;

[0025] Determine a plan control scheme corresponding to the first feature description vector from a plan control scheme library;

[0026] Based on the planned control plan, the future water quality adjustment plan and feed feeding plan of the breeding pond are regulated.

[0027] Optional, machine learning-based water quality prediction and intelligent control system, also includes:

[0028] Auxiliary modules for:

[0029] Generate a digital model based on current water quality data, water quality forecast results, and process information for adaptive regulation of future water quality regulation plans and feed feeding plans for aquaculture ponds;

[0030] When a user brings a smart terminal close to a breeding pond, the user's movement trajectory is continuously obtained through the smart terminal;

[0031] Determine whether the movement route indicates that a shooting prompt has been entered;

[0032] When the answer is yes, the user is prompted to take photos of the intended area of ​​the breeding pond through the smart terminal;

[0033] When the user starts to shoot the aquaculture pond, the user's shooting trajectory is continuously obtained through the smart terminal;

[0034] When the user is shooting the aquaculture pond, the smart terminal determines whether it is in the camera flipping time based on the movement trajectory and shooting trajectory;

[0035] When the answer is yes, the smart terminal is controlled to flip the camera;

[0036] When the user finishes photographing the breeding pond, the intended area of ​​the user's photography is obtained;

[0037] Determine the slice model of the intended area from the digitized model;

[0038] Determine regulatory recommendations based on the slice model;

[0039] Display slicing models and tuning suggestions to users;

[0040] When the user inputs a control plan, the breeding pond is relay-controlled based on the control plan.

[0041] Optionally, the auxiliary module determines whether the moving route indicates entering a shooting prompt time, including:

[0042] Determine a target sub-pond from the breeding pond; wherein the shortest distance between the center position of the target sub-pond and the moving route does not exceed a distance threshold;

[0043] When the ratio of the total number of target sub-pools to the total number of sub-pools in the breeding pond exceeds a ratio threshold or at least two target sub-pools have at least one standard association relationship, determining that the moving route indicates entering a shooting prompt time;

[0044] Otherwise, it is determined that the moving route does not indicate a shooting prompt timing.

[0045] Optionally, the auxiliary module determines whether the smart terminal enters a camera flipping opportunity based on the moving trajectory and the shooting trajectory, including:

[0046] Determine whether the moving trajectory and the shooting trajectory meet the timing entry constraint, and if so, determine the smart terminal enters the camera flip timing;

[0047] The timing entry constraints include:

[0048] Constraint 1: the average moving speed of the second target segment within the first target segment on the shooting trajectory exceeds the first speed threshold; the first target segment is the partial trajectory generated within the first time on the shooting trajectory; the second target segment is the partial trajectory of the previous target ratio within the first target segment;

[0049] as well as,

[0050] Constraint 2: the average moving speed of the third target segment within the first target segment on the shooting trajectory does not exceed the second speed threshold; the third target segment is the partial trajectory within the first target segment excluding the second target segment; the second speed threshold is less than the first speed threshold;

[0051] as well as,

[0052] Constraint 3: the maximum matching degree between the fourth target segment on the moving trajectory and each standard trajectory in the standard trajectory library exceeds the matching degree threshold; the fourth target segment is a part of the moving trajectory generated within the second time; the second time is less than the first time.

[0053] Optionally, the auxiliary module determines the control suggestion based on the slice model, including:

[0054] Performing feature description on the slice model to obtain a second feature description vector;

[0055] A control suggestion corresponding to the second feature description vector is determined from the control suggestion library.

[0056] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0057] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0059] Figure 1 Schematic diagram of a water quality prediction and intelligent control system based on machine learning in an embodiment of the present invention;

[0060] Figure 2 This is another schematic diagram of a water quality prediction and intelligent control system based on machine learning in an embodiment of the present invention;

[0061] Figure 3 This is another schematic diagram of a water quality prediction and intelligent control system based on machine learning in an embodiment of the present invention;

[0062] Figure 4 The figure is a flow chart of specific operation steps performed by the auxiliary module in an embodiment of the present invention. DETAILED DESCRIPTION

[0063] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0064] The embodiment of the present invention provides a water quality prediction and intelligent control system based on machine learning, such as Figure 1 As shown, including:

[0065] Real-time collection module 1, used for collecting current water quality data of the breeding pond in real time;

[0066] The water quality prediction module 2 is used to determine the water quality prediction result based on the pre-trained water quality prediction model and the current water quality data;

[0067] Intelligent control module 3 is used to adaptively control the future water quality adjustment plan and bait feeding plan of the aquaculture pond based on the water quality prediction results;

[0068] Among them, the pre-training steps of the water quality prediction model are as follows:

[0069] Obtain a large amount of historical water quality data from different aquaculture ponds;

[0070] Preprocess historical water quality data;

[0071] Based on machine learning algorithms, water quality prediction models are trained according to pre-processed historical water quality data;

[0072] The real-time acquisition module collects the current water quality data of the breeding pond in real time, including:

[0073] The current water quality data of the aquaculture pond is collected in real time through the sensor network arranged in the aquaculture pond;

[0074] Wherein, the current water quality data at least includes: salinity, temperature, pH value, dissolved oxygen;

[0075] Wherein, the preprocessing of historical water quality data includes:

[0076] The historical water quality data are processed in turn by removing missing values, removing outliers and standardizing.

[0077] In the above technical solution, the real-time acquisition module monitors and collects the current water quality data of the breeding pond in real time through the sensor network deployed in the breeding pond. The sensor network includes at least: salinity sensor, temperature sensor, pH sensor, dissolved oxygen sensor, etc.; when collecting a large amount of historical water quality data of different breeding ponds, historical data of various water quality parameters (such as salinity, temperature, pH value, dissolved oxygen, etc.) are collected from different breeding ponds; then data preprocessing is performed, missing values ​​are processed by interpolation, filling or discarding, outliers in the data (such as sudden sensor errors or extreme water quality fluctuations) are identified and eliminated, and the data are standardized (such as normalization or z-score standardization) to ensure that the dimensions of each eigenvalue are consistent; then based on the machine The machine learning algorithm is used to train the water quality prediction model based on the pre-processed historical water quality data. The trained water quality prediction model can determine the water quality prediction results based on the current water quality data. The water quality prediction results include the water quality change trend data of the breeding pond in the future (such as 24 hours, 48 ​​hours, etc.); finally, based on the water quality prediction results, the future water quality adjustment plan and feed feeding plan of the breeding pond can be adaptively regulated. For example, if the dissolved oxygen level is predicted to decrease, the intelligent control module will start the oxygenation device in advance or adjust the water fluidity to ensure that the aquatic animals in the breeding pond obtain sufficient oxygen. For example, when the system predicts that the dissolved oxygen is low and the ammonia nitrogen concentration is high, the system can appropriately reduce the amount of feed to avoid further deterioration of water quality due to excessive feed.

[0078] This application is based on a machine learning algorithm to train a water quality prediction model, collect current water quality data of the breeding pond in real time, and determine the water quality prediction result based on the water quality prediction model and the current water quality data. Based on the water quality prediction result, the future water quality adjustment plan and feed feeding plan of the breeding pond are adaptively adjusted to achieve automated and precise water quality management and feed delivery management of the breeding pond, thereby reducing labor costs, achieving real-time and precise regulation, and improving aquaculture efficiency and quality.

[0079] In one embodiment, Figure 2 As shown, the water quality prediction and intelligent control system based on machine learning also includes:

[0080] Optimization module 4, for:

[0081] Obtain the error conditions determined by the water quality prediction results;

[0082] Based on the error situation, the water quality prediction model is optimized.

[0083] Compare the water quality prediction results with the water quality data actually collected later to determine the error situation; based on the error situation, optimize the water quality prediction model, such as adaptively adjusting the hyperparameters of the machine learning model according to the error situation.

[0084] In one embodiment, the intelligent control module adaptively controls the future water quality adjustment plan and bait feeding plan of the aquaculture pond based on the water quality prediction result, including:

[0085] Performing feature description on the water quality prediction result to obtain a first feature description vector;

[0086] Determine a plan control scheme corresponding to the first feature description vector from a plan control scheme library;

[0087] Based on the planned control plan, the future water quality adjustment plan and feed feeding plan of the breeding pond are regulated.

[0088] When performing feature description, key features in the water quality prediction results are extracted, such as: dissolved oxygen change trend, temperature fluctuation amplitude, pH stability, etc.; the extracted key features are converted into numerical vectors to obtain the first feature description vector; the plan control scheme library has pre-set plan control schemes corresponding to different first feature description vectors, and the plan control scheme indicates how to control the future water quality adjustment plan and feed feeding plan of the aquaculture pond.

[0089] In one embodiment, Figure 3 As shown, the water quality prediction and intelligent control system based on machine learning also includes:

[0090] Auxiliary module 5, such as Figure 4 As shown, it is used for:

[0091] S1. Generate a digital model based on current water quality data, water quality prediction results, and process information for adaptive regulation of future water quality regulation plans and feed feeding plans for aquaculture ponds;

[0092] In S1, the process information for adaptively regulating the future water quality regulation plan and bait feeding plan of the aquaculture pond includes at least: the adjustment results of the water quality regulation plan and the bait feeding plan, the comparison information before and after the adjustment, etc.; the digital model is a digital three-dimensional model that visualizes the current water quality data, water quality prediction results and process information;

[0093] S2. When the user brings the smart terminal close to the breeding pond, the user's movement trajectory is continuously obtained through the smart terminal;

[0094] In S2, the smart terminal may be a smart phone or the like; the smart terminal has a built-in positioning device and a millimeter wave radar, the positioning device collects the real-time position of the smart terminal, and the millimeter wave radar collects the real-time relative position relationship between the user and the smart terminal. The real-time position of the user can be determined based on the real-time position of the smart terminal and the real-time relative position relationship between the user and the smart terminal, and the real-time recording of the real-time position of the user forms the user's movement trajectory;

[0095] S3, determining whether the moving route indicates that a shooting prompt timing has been entered;

[0096] In S3, the shooting prompting opportunity refers to the opportunity to prompt the user to shoot the intended area of ​​the breeding pond, and the intended area may be the breeding pond area where the user needs to re-determine the control plan, the breeding pond area where the user needs to obtain the control suggestions of the system, etc.;

[0097] S4. When the answer is yes, the user is prompted to take photos of the intended area of ​​the breeding pond through the smart terminal;

[0098] In S4, when the answer is yes, a corresponding prompt is given to the user, and the prompt may be given by pushing relevant prompt information to the smart terminal;

[0099] S5. When the user starts to shoot the aquaculture pond, the user's shooting trajectory is continuously obtained through the smart terminal;

[0100] In S5, when the user receives the prompt, he / she will start to shoot the breeding pond. When he / she starts shooting, the shooting trajectory of the user is continuously obtained through the smart terminal. The smart terminal is equipped with a front camera and a rear camera. As the user adjusts the shooting area of ​​the front camera or the rear camera, the center position of the shooting area will change continuously. The trajectory formed by the continuous change of the center position of the shooting area is the shooting trajectory.

[0101] S6. When the user is photographing the breeding pond, determine whether the smart terminal has entered a camera flipping time based on the movement trajectory and the shooting trajectory;

[0102] In S6, the camera flip timing refers to the timing of controlling the smart terminal to flip the camera;

[0103] S7, when the answer is yes, control the smart terminal to flip the camera;

[0104] In S6, when the camera flip timing comes, the smart terminal is controlled to flip the camera;

[0105] S8. After the user finishes photographing the breeding pond, the intended area of ​​the user's photographing is obtained;

[0106] In S8, when obtaining the intended area photographed by the user, the entire photographed area photographed by the smart terminal from the time when the user starts photographing the breeding pond to the time when the photographing ends is taken as the intended area;

[0107] S9, determining a slice model of the intended area from the digitized model;

[0108] In S9, the intended area is a geographical location area in the breeding pond, and the digital model will display information related to the geographical location area, and the part of the model displaying the related information is used as a slice model;

[0109] S10. Determine the regulation suggestions based on the slice model;

[0110] In S10, the user may want to obtain regulation suggestions related to the slice model, therefore, the disadvantage of regulation suggestions based on the slice model;

[0111] S11, displaying the slicing model and regulation suggestions to the user;

[0112] In S11, the slice model and the control suggestions are displayed to the user. The user can view the slice model to learn the situation of the intended area and make a decision on a new control plan based on the control suggestions.

[0113] S12. When the user inputs a control plan, the breeding pond is controlled in a relay manner based on the control plan.

[0114] In S12, when the user decides on a new control plan, relay control is performed on the aquaculture pond based on the control plan; relay control means that based on the water quality prediction results, the future water quality control plan and the feed feeding plan of the aquaculture pond are adaptively controlled, and the water quality control plan and the feed feeding plan are continued to be controlled.

[0115] Generally, when the system performs intelligent regulation, the user (the manager of the breeding pond) may want to conduct appropriate manual intervention and regulation of the breeding pond. He will approach the breeding pond to observe the working conditions of relevant sensors and the actual water quality, etc. At this time, the system can make regulation suggestions, but the premise of the suggestion is to know which breeding pond areas the user wants to make regulation decisions on, that is, how to determine the intention.

[0116] Usually, when a user wants to select an intended area, he needs to select a frame on the map of the breeding pond, which is cumbersome. The embodiment of the present invention attempts to assist the user in photographing the intended area, but the following two problems may occur:

[0117] 1. There are a large number of sub-pools in the breeding pond, which are arranged in several rows and columns. When shooting the intended area, the user may need to shoot at least two sub-pools in relative directions (such as front and back). However, if the user wants to shoot at least two sub-pools, it is impossible to shoot continuously. In order to avoid irrelevant areas in the shooting area, it is necessary to manually pause, which is cumbersome;

[0118] 2. The space available for users to walk between different pools is small. If the user manually pauses the camera and then controls the camera to face the next pool to be photographed, it is cumbersome to move.

[0119] The embodiments of the present invention can solve the above two problems. When the user takes pictures of the breeding pond, it is determined whether the smart terminal has entered the camera flipping time based on the movement trajectory and the shooting trajectory. When the camera flipping time is entered, the smart terminal is controlled to flip the camera. The user does not need to manually pause shooting to continuously shoot at least two sub-pool areas that the user wants to be the intended area. The user does not need to flip the body to move, which suggests convenience and is suitable for smaller moving spaces between different sub-pools.

[0120] Secondly, the embodiment of the present invention determines whether the moving route indicates entering a shooting prompt time. If so, the user is prompted to shoot the intended area of ​​the breeding pond through the smart terminal, accurately prompting the user to shoot the intended area, which is humane.

[0121] In one embodiment, the auxiliary module determines whether the moving route indicates entering a shooting prompt opportunity, including:

[0122] S31, determining a target sub-pond from the breeding pond; wherein the shortest distance between the center position of the target sub-pond and the moving route does not exceed a distance threshold;

[0123] In S31, the distance threshold may be 50 meters; if the shortest distance between the center position of the target sub-pool and the moving route does not exceed the distance threshold, it means that the user has viewed the target sub-pool during the moving process;

[0124] S32, when the ratio of the total number of target sub-pools to the total number of sub-pools in the breeding pond exceeds a ratio threshold or at least two target sub-pools have at least one standard association relationship, determining that the moving route indicates entering a shooting prompt timing;

[0125] In S31, the ratio threshold may be, for example, 3 / 10, and the standard association relationship may be that the aquatic animals cultured in the target sub-pools are the same, etc. When the ratio of the total number of target sub-pools to the total number of sub-pools in the breeding pond exceeds the ratio threshold or at least two target sub-pools have at least one standard association relationship, it means that the user has viewed more sub-pools of the breeding pond or the viewed sub-pools of the breeding pond indicate that the user needs to make a control decision, then determining the moving route indicates entering the shooting prompt time;

[0126] S33: Otherwise, determine that the moving route does not indicate entering a shooting prompting opportunity.

[0127] The embodiment of the present invention accurately determines whether the moving route indicates entering a shooting prompting opportunity, thereby improving the suitability of the shooting prompting and greatly improving the applicability of the system.

[0128] In one embodiment, the auxiliary module determines whether the smart terminal enters the camera flipping time based on the movement trajectory and the shooting trajectory, including:

[0129] S61, determining whether the moving trajectory and the shooting trajectory meet the timing entry constraint, and if so, determining the smart terminal to enter the camera flip timing;

[0130] The timing entry constraints include:

[0131] Constraint 1: the average moving speed of the second target segment within the first target segment on the shooting trajectory exceeds the first speed threshold; the first target segment is the partial trajectory generated within the first time on the shooting trajectory; the second target segment is the partial trajectory of the previous target ratio within the first target segment;

[0132] as well as,

[0133] Constraint 2: the average moving speed of the third target segment within the first target segment on the shooting trajectory does not exceed the second speed threshold; the third target segment is the partial trajectory within the first target segment excluding the second target segment; the second speed threshold is less than the first speed threshold;

[0134] as well as,

[0135] Constraint 3: the maximum matching degree between the fourth target segment on the moving trajectory and each standard trajectory in the standard trajectory library exceeds the matching degree threshold; the fourth target segment is a part of the moving trajectory generated within the second time; the second time is less than the first time.

[0136] In constraint 1 and constraint 2, the first speed threshold is a speed representing a faster user movement; the average moving speed of the second target segment can be obtained by dividing the total travel length of the second target segment by the total time length of the second target segment; the most recent first time can be, for example, the most recent 100 seconds; the front target ratio can be, for example, the first 4 / 5; the average moving speed of the third target segment can be obtained by dividing the total travel length of the third target segment by the total time length of the third target segment; the second speed threshold is a speed representing a slower user movement. When the user wants to shoot at least two sub-pool areas in relative directions, he will first shoot one of the sub-pool areas with the rear camera of the smart terminal. When shooting, the shooting speed is fast at the beginning. If the camera of the smart terminal is to be flipped, the shooting speed drops sharply and waits for the system to automatically flip, which satisfies constraint 1 and constraint 2.

[0137] In constraint three, the second most recent time may be, for example, the most recent 20 seconds; the standard trajectory in the standard trajectory library is a moving trajectory generated when the user wants to flip the camera of the smart terminal, such as: the user starts shooting with the rear camera of the smart terminal, and when he wants to flip the camera, the moving trajectory is generated from the original position of the front camera facing the smart terminal to the position of the front camera facing the side of the smart terminal (getting out of the way so as not to block the shooting area of ​​the front camera); the matching degree threshold may be 85%; when the maximum matching degree between the fourth target segment and each standard trajectory in the standard trajectory library exceeds the matching degree threshold, it indicates that the user wants to flip the camera of the smart terminal.

[0138] By setting constraint one, constraint two, and constraint three, it is possible to accurately determine whether the smart terminal has entered the camera flipping timing, which greatly improves the intelligence level of the system.

[0139] In one embodiment, the auxiliary module determines the regulation suggestion based on the slice model, including:

[0140] Performing feature description on the slice model to obtain a second feature description vector;

[0141] A control suggestion corresponding to the second feature description vector is determined from the control suggestion library.

[0142] When performing feature description, the system will extract features related to the status of the aquaculture pond in the slice model, such as: the average value of dissolved oxygen, the range of change, the fluctuation of pH value, the change trend of water temperature, etc.; then the feature is represented in the form of a vector to obtain the second feature description vector; there are control suggestions corresponding to different second feature description vectors in the control suggestion library, so the library is checked to determine the control suggestion.

[0143] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A water quality prediction and intelligent control system based on machine learning, characterized in that: include: Real-time data collection module, used to collect the current water quality data of the breeding pond in real time; A water quality prediction module is used to determine the water quality prediction result based on the pre-trained water quality prediction model and the current water quality data; Intelligent control module, used to adaptively control the future water quality adjustment plan and bait feeding plan of the aquaculture pond based on the water quality prediction results; Among them, the pre-training steps of the water quality prediction model are as follows: Obtain a large amount of historical water quality data from different aquaculture ponds; Preprocess historical water quality data; Based on the machine learning algorithm, the water quality prediction model is trained according to the preprocessed historical water quality data.

2. The water quality prediction and intelligent control system based on machine learning according to claim 1, characterized in that: The real-time acquisition module collects the current water quality data of the breeding pond in real time, including: The current water quality data of the breeding pond is collected in real time through the sensor network arranged in the breeding pond.

3. The water quality prediction and intelligent control system based on machine learning according to claim 1, characterized in that: The current water quality data at least includes: salinity, temperature, pH value, and dissolved oxygen.

4. The water quality prediction and intelligent control system based on machine learning according to claim 1, characterized in that: The preprocessing of the historical water quality data includes: The historical water quality data are processed in turn by removing missing values, removing outliers and standardizing.

5. The water quality prediction and intelligent control system based on machine learning according to claim 1, characterized in that: Also includes: Optimization modules for: Obtain the error conditions determined by the water quality prediction results; Based on the error situation, the water quality prediction model is optimized.

6. The water quality prediction and intelligent control system based on machine learning according to claim 1, characterized in that: The intelligent control module adaptively controls the future water quality adjustment plan and feed feeding plan of the aquaculture pond based on the water quality prediction results, including: Performing feature description on the water quality prediction result to obtain a first feature description vector; Determine a plan control scheme corresponding to the first feature description vector from a plan control scheme library; Based on the planned control plan, the future water quality adjustment plan and feed feeding plan of the breeding pond are regulated.

7. The water quality prediction and intelligent control system based on machine learning according to claim 1, characterized in that: Also includes: Auxiliary modules for: Generate a digital model based on current water quality data, water quality forecast results, and process information for adaptive regulation of future water quality regulation plans and feed feeding plans for aquaculture ponds; When a user brings a smart terminal close to a breeding pond, the user's movement trajectory is continuously obtained through the smart terminal; Determine whether the movement route indicates that a shooting prompt has been entered; When the answer is yes, the user is prompted to take photos of the intended area of ​​the breeding pond through the smart terminal; When the user starts to shoot the aquaculture pond, the user's shooting trajectory is continuously obtained through the smart terminal; When the user is shooting the aquaculture pond, the smart terminal determines whether it is in the camera flipping time based on the movement trajectory and shooting trajectory; When the answer is yes, the smart terminal is controlled to perform camera flipping; When the user finishes photographing the breeding pond, the intended area of ​​the user's photography is obtained; Determine the slice model of the intended area from the digitized model; Determine regulatory recommendations based on the slice model; Display slicing models and tuning suggestions to users; When the user inputs a control plan, the breeding pond is controlled in a relay manner based on the control plan.

8. The water quality prediction and intelligent control system based on machine learning according to claim 7, characterized in that: The auxiliary module determines whether the moving route indicates entering a shooting prompt time, including: Determine a target sub-pond from the breeding pond; wherein the shortest distance between the center position of the target sub-pond and the moving route does not exceed a distance threshold; When the ratio of the total number of target sub-pools to the total number of sub-pools in the breeding pond exceeds a ratio threshold or at least two target sub-pools have at least one standard association relationship, determining that the moving route indicates entering a shooting prompt time; Otherwise, it is determined that the moving route does not indicate a shooting prompt timing.

9. The water quality prediction and intelligent control system based on machine learning according to claim 7, characterized in that: The auxiliary module determines whether the smart terminal enters the camera flipping time based on the moving trajectory and the shooting trajectory, including: Determine whether the moving trajectory and the shooting trajectory meet the timing entry constraint, and if so, determine the smart terminal enters the camera flip timing; The timing entry constraints include: Constraint 1: the average moving speed of the second target segment within the first target segment on the shooting trajectory exceeds the first speed threshold; the first target segment is the partial trajectory generated within the first time on the shooting trajectory; the second target segment is the partial trajectory of the previous target ratio within the first target segment; as well as, Constraint 2: the average moving speed of the third target segment within the first target segment on the shooting trajectory does not exceed the second speed threshold; the third target segment is the partial trajectory within the first target segment excluding the second target segment; the second speed threshold is less than the first speed threshold; as well as, Constraint 3: the maximum matching degree between the fourth target segment on the moving trajectory and each standard trajectory in the standard trajectory library exceeds the matching degree threshold; the fourth target segment is a part of the moving trajectory generated within the second time; the second time is less than the first time.

10. The water quality prediction and intelligent control system based on machine learning according to claim 7, characterized in that: The auxiliary module determines the control suggestions based on the slice model, including: Performing feature description on the slice model to obtain a second feature description vector; A control suggestion corresponding to the second feature description vector is determined from the control suggestion library.

Images