Fish disease comprehensive regulation method, system, device and medium fusing artificial intelligence and mechanism model

Abstract

The present application aims at the problems of lack of ecological process mechanism explanation, low credibility of prediction results and difficulty in guiding precise intervention in the prior art, and discloses a fish disease comprehensive regulation method, system, equipment and medium fusing artificial intelligence and mechanism model, the method comprising: collecting fish pond data and preprocessing; establishing a mechanism model PCLake of the fish pond ecosystem, calibrating and verifying the mechanism model PCLake by using the preprocessed fish pond data, simulating the ecological process by using the calibrated mechanism model PCLake, analyzing the simulation results, and obtaining the ecological safety boundary of each index; searching for the optimal solution within the ecological safety boundary range of each index given by the mechanism model PCLake by using a deep reinforcement learning algorithm, and giving a final regulation measure scheme; receiving the instructions output by the regulation measure scheme for spatio-temporal decomposition, and starting the related equipment through the Internet of Things after the decomposition. The present application can more accurately regulate fish diseases and effectively reduce drug abuse.

Description

Technical Field

[0001] This invention relates to aquaculture technology, specifically to a method, system, equipment, and medium for the comprehensive regulation of fish diseases that integrates artificial intelligence and mechanistic models. Background Technology

[0002] The mainstream technologies in the field of fish disease control can be summarized in the following aspects: First, threshold alarm systems based on human experience and regular water quality testing, which trigger audible and visual alarms when parameters exceed preset safety thresholds by deploying sensors for dissolved oxygen, ammonia nitrogen, etc.; Second, traditional control theories, such as PID controllers, are used to independently regulate single devices (such as aerators and feeders) in a closed loop to maintain a certain water quality parameter; Third, statistical prediction models based on historical data, such as time series analysis or regression models, are used to make short-term predictions of water quality trends or the probability of disease occurrence, but the coordination between various links is weak and it is heavily dependent on stable networks and human intervention.

[0003] The PCLake model is the earliest composite dynamic model developed for temperate shallow lakes. It mainly studies the structure, function and spatiotemporal evolution of shallow lake ecosystems. By simulating the impact and feedback mechanisms of physical, chemical and biological processes in the water body and surface sediments on the shallow lake ecosystem, it predicts the dynamic changes of the system.

[0004] Existing purely data-driven AI methods are like "black boxes," lacking explanations of ecological process mechanisms, resulting in low reliability of predictions and difficulty in guiding precise interventions. Summary of the Invention

[0005] This invention addresses the shortcomings of existing technologies by providing a comprehensive regulation method, system, device, and medium for fish diseases that integrates artificial intelligence and mechanistic models. The coupling of artificial intelligence and mechanistic models enables more accurate regulation of fish diseases and effectively reduces drug abuse.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A comprehensive method for the control of fish diseases that integrates artificial intelligence and mechanistic models includes the following steps: Collect and preprocess fishpond data; A mechanistic model PCLake for fishpond ecosystems was established. The mechanistic model PCLake was calibrated and validated using preprocessed fishpond data. Ecological process simulations were performed using the calibrated mechanistic model PCLake. The simulation results were analyzed to obtain the ecological safety boundaries of each indicator. By using deep reinforcement learning algorithms, the optimal solution is searched within the ecological security boundary of each indicator given by the mechanistic model PCLake, and the final control measures are proposed. The system receives instructions from the control measures plan, performs spatiotemporal decomposition, and then uses the decomposed instructions to activate relevant devices via the Internet of Things.

[0008] To optimize the above technical solution, the specific measures also include: Furthermore, the mechanistic model PCLake includes an abiotic module, a phytoplankton module, an aquatic plant module, a zooplankton module, a fish module, a benthic animal module, and an auxiliary module. The fish module includes a dynamic feeding submodule and a fish disease submodule.

[0009] Furthermore, the specific process of simulating ecological processes using the PCLake model with a predetermined utilization rate, analyzing the simulation results, and obtaining the ecological security boundaries of each indicator is as follows: Different gradients were set for different indicators within ±50% of the standard value to simulate the process of fishpond water transitioning from clear water to turbid water. The simulation results were analyzed to find the critical point at which the fishpond ecosystem undergoes abrupt change or irreversible degradation. This critical point constitutes the ecological safety range of the parameter. Similarly, the accumulation of ammonia nitrogen and nitrite under different water temperatures and feeding amounts was simulated to find the concentration thresholds for ammonia nitrogen poisoning and nitrite poisoning in fish, thus obtaining a set of quantitative ecological safety boundaries.

[0010] Furthermore, the dynamic feeding submodule specifically comprises: The dynamic feeding submodule takes the feeding amount as input and outputs the rate of change of organic matter concentration, nitrogen concentration, and phosphorus concentration in the water body. It calculates these rates based on the feeding amount, specifically as follows: The amount of feed in the water can be calculated based on the amount of feed given, expressed by the following formula:

[0011] In the formula, Let t represent the amount of bait in the water at time t, and n be the total number of times bait is thrown. It is the first i Feeding amount per feeding It is the first i The timing of the second feeding. It is a Dirac function; The formula for calculating the rate of change of organic matter concentration in water is derived based on the principle of decomposition and release of organic matter in feed:

[0012] In the formula, O ( t This indicates the concentration of organic matter in the water. α This indicates the rate of release of organic matter from the bait. This represents the amount of food in the water at time t. kdecay Indicates the decomposition rate; The organic matter concentration in water is obtained based on the formula for calculating the rate of change of organic matter concentration in water. O ( t The rates of change in nitrogen and phosphorus concentrations are obtained based on the organic matter concentration in the water body, using the following formulas:

[0013]

[0014] In the formula, Indicates the nitrogen concentration in the water. f N This indicates the mass fraction of nitrogen (N) in the bait. η N This indicates the mineralization efficiency of nitrogen (N) during the decomposition of organic matter. Indicates the phosphorus concentration in the water. f P This indicates the mass fraction of phosphorus (P) in the feed. η P This indicates the mineralization efficiency of phosphorus (P) during the decomposition of organic matter.

[0015] Furthermore, the fish disease submodule is used to dynamically calculate the fish disease risk index based on the concentration of water quality parameters, and to determine the level of fish disease based on the fish disease risk index, specifically as follows: The fish disease submodule calls upon the rate of change of water quality parameters from the abiotic module, phytoplankton module, aquatic plant module, zooplankton module, benthic animal module, and dynamic feeding submodule. Based on the rate of change of water quality parameters output by each module, it updates the concentration of each water quality parameter. Based on the concentration of water quality parameters, it calculates the fish disease risk index, expressed by the formula:

[0016] In the formula, D ( i () is a fish disease risk index, which represents the probability of fish contracting different levels of diseases. It is the first i Concentration of various water quality parameters C i,threshold It is the first i Safety thresholds for various water quality parameters w i These are weighting coefficients; The disease level corresponding to the highest fish disease risk index is determined as the current fish disease level.

[0017] Furthermore, the deep reinforcement learning algorithm employs a proximal policy optimization algorithm, which uses water quality parameters, environmental parameters, meteorological parameters, and fish disease levels to form the state space of the fishpond ecosystem. The combination of control measures serves as the action space of the agent, including oxygenation intensity, water exchange volume, feed amount, and medication dosage. A reward function is used to evaluate the reward value of each action; the reward function expression is as follows:

[0018] In the formula, It's a reward value. α 1 represents the weight of disease control; This indicates the difference in disease severity between affected and healthy fish. α 2 is the economic cost weight; Cost This indicates the expenses incurred in response to this disease. α 3 is the ecological stability weight. This indicates the ecological stability of a fishpond when diseases occur; The specific steps of using deep reinforcement learning algorithms to search for the optimal solution within the ecological security boundary of each indicator given by the PCLake mechanism model are as follows: Based on the current state of the fishpond ecosystem, a candidate regulatory action is generated. The mechanistic model PCLake is then activated to simulate the ecological process. The simulated future water quality change curve is monitored to check whether any indicators exceed the ecological safety boundary. If no boundary is exceeded and the simulation results tend to improve, the candidate regulatory action is considered safe and feasible, and its reward value is added to the experience pool. If a boundary is exceeded, the candidate regulatory action is marked as a dangerous action, the reward value is negative, and the probability of selecting such a regulatory action is reduced in the strategy update.

[0019] Furthermore, the instructions output by the received control measures scheme are spatiotemporally decomposed, and the decomposed instructions are used to activate relevant devices via the Internet of Things, specifically as follows: The instructions are decomposed into three dimensions: time axis, spatial domain, and measure set. The time axis is divided into three time periods: 0-2h, 2-6h, and 6-24h. The spatial domain is divided into surface, middle, and bottom layers. The surface layer specifically includes surface aerators to promote gas exchange and inject new external water. The middle layer specifically includes diffusers for targeted drug administration to inhibit pathogens. The bottom layer specifically includes submersible aerators to promote gas exchange and extract aquaculture water containing organic matter. The measure set includes aeration rate, water exchange rate, and drug administration. The decomposed instructions are then used to activate relevant equipment via the Internet of Things (IoT).

[0020] This invention also proposes a comprehensive fish disease control system integrating artificial intelligence and mechanistic models, comprising: The intelligent data module is used to collect and preprocess data from the fishpond. The mechanism model module is used to establish the mechanism model PCLake of the fishpond ecosystem. The mechanism model PCLake is calibrated and validated using preprocessed fishpond data. The calibrated mechanism model PCLake is used to simulate ecological processes, analyze the simulation results, and obtain the ecological safety boundaries of each indicator. The artificial intelligence module is used to search for the optimal solution within the ecological security boundary of each indicator given by the mechanistic model PCLake using deep reinforcement learning algorithms, and to provide the final control measures. The equipment control module is used to receive the instructions output by the control measures plan, perform spatiotemporal decomposition, and activate the relevant equipment through the Internet of Things.

[0021] The present invention also proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the integrated regulation method for fish diseases that integrates artificial intelligence and mechanistic models as described above.

[0022] The present invention also proposes a computer-readable storage medium storing a computer program that enables a computer to execute the integrated fish disease control method that integrates artificial intelligence and mechanistic models as described above.

[0023] The beneficial effects of this invention are: This invention effectively solves the problems of decision lag, model "black box," system fragmentation, and insufficient long-term extrapolation capabilities in existing technologies by organically coupling an ecological mechanism model with process interpretation capabilities with a data-driven, real-time optimized deep reinforcement learning model. It transforms traditional passive emergency response and single-point control into a proactive, predictive, and systematic comprehensive prevention and control model within the scope of ecological security. This model can adjust water quality, more accurately regulate fish diseases, and effectively reduce drug overuse. Attached Figure Description

[0024] Figure 1 This is a block diagram of the integrated fish disease control system that combines artificial intelligence and mechanistic models, as proposed in this invention.

[0025] Figure 2 This is a flowchart illustrating the interaction between the dynamic feeding submodule and the fish disease submodule.

[0026] Figure 3 A flowchart for obtaining ecological security boundaries.

[0027] Figure 4 This is a flowchart of the PCLake-PPO model operation. Detailed Implementation

[0028] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0029] Example 1 This invention proposes a comprehensive regulation method for fish diseases that integrates artificial intelligence and mechanistic models, comprising the following steps: Data collection and preprocessing of fishponds were performed. This data included meteorological, water quality, topographic, hydrological, and feeding data. Specific data included topography, water depth, inflow and outflow rates, exogenous nutrient loads (particulate organic phosphorus, dissolved organic phosphorus, phosphate, particulate organic nitrogen, dissolved organic nitrogen, nitrate, and ammonia nitrogen), wind speed, wind direction, air pressure, air temperature, dew point temperature, cloud cover, feeding frequency, and feed amount. Preprocessing included data summarization, outlier removal, and sorting. The collected meteorological, water quality, and feeding data were used to establish the PCLake mechanistic model in the mechanistic model module, completing the calibration and validation of the PCLake model. The collected fish disease, aeration, water exchange, and medication data were used for training, validation, and testing of the PPO algorithm in the artificial intelligence model module.

[0030] A mechanistic model of a fishpond ecosystem, PCLake, was established. The original PCLake model includes abiotic modules, phytoplankton modules, aquatic plant modules, zooplankton modules, fish modules, benthic animal modules, and auxiliary modules. To address fish disease issues, PCLake was localized by adding dynamic feeding submodules and fish disease submodules to the fish module.

[0031] The dynamic feeding submodule quantifies the driving effect of feed input on the nutrient cycle of water bodies during aquaculture by dynamically inputting feed amount, feed composition and residual feed decomposition process, and quantifies the impact of feeding behavior on the input of nutrients (nitrogen, phosphorus, etc.) and organic matter in water bodies.

[0032] The dynamic feeding submodule takes the feeding amount as input and outputs the rate of change of organic matter concentration, nitrogen concentration, and phosphorus concentration in the water body. It calculates these rates based on the feeding amount, specifically as follows: The amount of feed in the water can be calculated based on the amount of feed given, expressed by the following formula:

[0033] In the formula, Let t represent the amount of bait in the water at time t, and n be the total number of times bait is thrown. It is the firsti Feeding amount per feeding It is the first i The timing of the second feeding. It is a Dirac function; The formula for calculating the rate of change of organic matter concentration in water is derived based on the principle of decomposition and release of organic matter in feed:

[0034] In the formula, O ( t This indicates the concentration of organic matter in the water. α This indicates the release rate of organic matter from the bait, determined experimentally, with a default value of 0.3. This represents the amount of food in the water at time t. k decay Indicates the decomposition rate; The organic matter concentration in water is obtained based on the formula for calculating the rate of change of organic matter concentration in water. O ( t The rates of change in nitrogen and phosphorus concentrations are obtained based on the organic matter concentration in the water body, using the following formulas:

[0035]

[0036] In the formula, Indicates the nitrogen concentration in the water. f N This indicates the mass fraction of nitrogen (N) in the bait. η N This indicates the mineralization efficiency of nitrogen (N) during the decomposition of organic matter. Indicates the phosphorus concentration in the water. f P This indicates the mass fraction of phosphorus (P) in the feed. η P This indicates the mineralization efficiency of phosphorus (P) during the decomposition of organic matter.

[0037] The fish disease submodule is used to dynamically calculate the fish disease risk index based on the concentration of water quality parameters, and to determine the level of fish diseases based on the fish disease risk index. Specifically: The fish disease submodule calls upon the rate of change of water quality parameters from the abiotic module, phytoplankton module, aquatic plant module, zooplankton module, benthic animal module, and dynamic feeding submodule. Based on the rate of change of water quality parameters output by each module, it updates the concentration of each water quality parameter. Based on the concentration of water quality parameters, it calculates the fish disease risk index, expressed by the formula:

[0038] In the formula, D ( i() is a fish disease risk index, which represents the probability of fish contracting different levels of diseases. It is the first i Concentration of various water quality parameters C i,threshold It is the first i Safety thresholds for various water quality parameters w i These are weighting coefficients; The disease level corresponding to the highest fish disease risk index is determined as the current fish disease level.

[0039] The interaction process between the dynamic feeding submodule and the fish disease submodule is as follows: Figure 2 .

[0040] The mechanistic model PCLake was calibrated and validated using preprocessed fishpond data. Ecological process simulations were then performed using the calibrated PCLake model, and the simulation results were analyzed to obtain the ecological safety boundaries for each indicator. These indicators included water temperature, dissolved oxygen (DO), pH, ammonia nitrogen, nitrate nitrogen, total nitrogen, nitrite, orthophosphate, total phosphorus, and chlorophyll a concentration. The process for obtaining the ecological safety boundaries is as follows: Figure 3 The specific process is as follows: Different gradients were set for different indicators within ±50% of the standard value to simulate the process of fishpond water transitioning from clear water (dominated by aquatic plants) to turbid water (algal bloom). The simulation results were analyzed to find the critical point at which the fishpond ecosystem undergoes abrupt change or irreversible degradation (for example, it was found that when the total phosphorus concentration in the water exceeds 0.03 mg / L, algae will proliferate rapidly, causing drastic diurnal fluctuations in DO, which can easily lead to hypoxia). This critical point constitutes the ecological safety range of the parameter. Similarly, the accumulation of ammonia nitrogen and nitrite under different water temperatures and feeding amounts was simulated to find the concentration thresholds for ammonia nitrogen poisoning and nitrite poisoning in fish, thus obtaining a set of quantitative ecological safety boundaries.

[0041] This study utilizes deep reinforcement learning algorithms to search for optimal solutions within the ecological safety boundaries of each indicator given by the PCLake mechanistic model, ultimately proposing a final control measure scheme. The deep reinforcement learning algorithm employs a proximal policy optimization (PPO) algorithm, supporting offline policy learning, reusing historical monitoring data, and leveraging the high-cost simulation data generated by PCLake to reduce trial-and-error costs in real-world environments. Water quality parameters, environmental parameters, meteorological parameters, and fish disease levels collectively constitute the state space of the fishpond ecosystem. The combination of control measures serves as the action space for the agent, including oxygenation intensity, water exchange volume, feed quantity, and medication dosage. A reward function is used to evaluate the reward value of each action; the reward function expression is as follows:

[0042] In the formula, It's a reward value. α 1 represents the weight of disease control; This indicates the difference in disease severity between affected and healthy fish. α 2 is the economic cost weight; Cost This indicates the expenses incurred in response to this disease. α 3 is the ecological stability weight. This indicates the ecological stability of a fishpond when diseases occur; The specific steps of using deep reinforcement learning algorithms to search for the optimal solution within the ecological security boundary of each indicator given by the PCLake mechanism model are as follows: Based on the current state of the fishpond ecosystem, a candidate regulatory action is generated (e.g., increasing aerator power by 50%, reducing feed by 30%, or changing water by 10%). The PCLake mechanistic model is then activated to simulate the ecological process, monitoring the simulated future water quality change curve. It is checked whether any indicators exceed the ecological safety boundary. If no boundary is exceeded and the simulation results show improvement, the candidate regulatory action is considered safe and feasible, and its reward value is added to the experience pool. If a boundary is exceeded, the candidate regulatory action is marked as a dangerous action, its reward value is negative, and the probability of selecting such a regulatory action is reduced in the strategy update. The PCLake-PPO model operation flow is as follows: Figure 4 As shown.

[0043] The system receives and decomposes the instructions from the control measures plan into temporal and spatial parameters. These decomposed instructions are then used to activate relevant equipment via the Internet of Things (IoT), thereby achieving the goal of adjusting water quality and controlling fish diseases. Specifically: The instructions are decomposed into three dimensions: time axis, spatial domain, and measure set. The time axis is divided into three time periods: 0-2h, 2-6h, and 6-24h. The spatial domain is divided into surface, middle, and bottom layers. The surface layer specifically includes surface aerators to promote gas exchange and inject new external water. The middle layer specifically includes diffusers for targeted drug administration to inhibit pathogens. The bottom layer specifically includes submersible aerators to promote gas exchange and extract aquaculture water containing organic matter. The measure set includes aeration rate, water exchange rate, and drug administration. The decomposed instructions are then used to activate relevant equipment via the Internet of Things (IoT).

[0044] Example 2 This invention proposes a comprehensive fish disease control system that integrates artificial intelligence and mechanistic models, corresponding to the method in Example 1, such as... Figure 1 As shown, it includes: The intelligent data module is used to collect and preprocess data from the fishpond. The mechanism model module is used to establish the mechanism model PCLake of the fishpond ecosystem. The mechanism model PCLake is calibrated and validated using preprocessed fishpond data. The calibrated mechanism model PCLake is used to simulate ecological processes, analyze the simulation results, and obtain the ecological safety boundaries of each indicator. The artificial intelligence module is used to search for the optimal solution within the ecological security boundary of each indicator given by the mechanistic model PCLake using deep reinforcement learning algorithms, and to provide the final control measures. The equipment control module is used to receive the instructions output by the control measures plan, perform spatiotemporal decomposition, and activate the relevant equipment through the Internet of Things.

[0045] The implementation methods of each module and its function in the system are completely consistent with the steps of the method in Implementation Example 1, so they will not be repeated here.

[0046] Example 3 This invention proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the integrated regulation method for fish diseases that integrates artificial intelligence and mechanistic models as described in Embodiment 1.

[0047] Example 4 This invention proposes a computer-readable storage medium storing a computer program that enables a computer to execute the integrated fish disease control method that combines artificial intelligence and mechanistic models as described in Embodiment 1.

[0048] In the embodiments disclosed in this application, a computer storage medium may be a tangible medium that may contain or store programs for use by or in conjunction with an instruction execution system, apparatus, or device. The computer storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of computer storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, and portable compact disc read-only memory (CD). ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0049] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0050] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A comprehensive regulation method for fish diseases integrating artificial intelligence and mechanistic models, characterized in that, Includes the following steps: Collect and preprocess fishpond data; A mechanistic model PCLake for fishpond ecosystems was established. The mechanistic model PCLake was calibrated and validated using preprocessed fishpond data. Ecological process simulations were performed using the calibrated mechanistic model PCLake. The simulation results were analyzed to obtain the ecological safety boundaries of each indicator. By using deep reinforcement learning algorithms, the optimal solution is searched within the ecological security boundary of each indicator given by the mechanistic model PCLake, and the final control measures are proposed. The system receives instructions from the control measures plan, performs spatiotemporal decomposition, and then uses the decomposed instructions to activate relevant devices via the Internet of Things.

2. The integrated regulation method for fish diseases combining artificial intelligence and mechanistic models as described in claim 1, characterized in that, The mechanistic model PCLake includes an abiotic module, a phytoplankton module, an aquatic plant module, a zooplankton module, a fish module, a benthic animal module, and an auxiliary module. The fish module includes a dynamic feeding submodule and a fish disease submodule.

3. The integrated regulation method for fish diseases combining artificial intelligence and mechanistic models as described in claim 1, characterized in that, The specific process of simulating ecological processes using the PCLake model with a predetermined utilization rate, analyzing the simulation results, and obtaining the ecological security boundaries of each indicator is as follows: Different gradients were set for different indicators within ±50% of the standard value to simulate the process of fishpond water transitioning from clear water to turbid water. The simulation results were analyzed to find the critical point at which the fishpond ecosystem undergoes abrupt change or irreversible degradation. This critical point constitutes the ecological safety range of the parameter. Similarly, the accumulation of ammonia nitrogen and nitrite under different water temperatures and feeding amounts was simulated to find the concentration thresholds for ammonia nitrogen poisoning and nitrite poisoning in fish, thus obtaining a set of quantitative ecological safety boundaries.

4. The integrated regulation method for fish diseases combining artificial intelligence and mechanistic models as described in claim 2, characterized in that, The dynamic feeding submodule is specifically as follows: The dynamic feeding submodule takes the feeding amount as input and outputs the rate of change of organic matter concentration, nitrogen concentration, and phosphorus concentration in the water body. It calculates these rates based on the feeding amount, specifically as follows: The amount of feed in the water can be calculated based on the amount of feed given, expressed by the following formula: In the formula, Let t represent the amount of bait in the water at time t, and n be the total number of times bait is thrown. It is the first i Feeding amount per feeding It is the first i The timing of the second feeding. It is a Dirac function; The formula for calculating the rate of change of organic matter concentration in water is derived based on the principle of decomposition and release of organic matter in feed: In the formula, O ( t This indicates the concentration of organic matter in the water. α This indicates the rate of release of organic matter from the bait. This represents the amount of food in the water at time t. k decay Indicates the decomposition rate; The organic matter concentration in water is obtained based on the formula for calculating the rate of change of organic matter concentration in water. O ( t The rates of change in nitrogen and phosphorus concentrations are obtained based on the organic matter concentration in the water body, using the following formulas: In the formula, Indicates the nitrogen concentration in the water. f N This indicates the mass fraction of nitrogen (N) in the bait. η N This indicates the mineralization efficiency of nitrogen (N) during the decomposition of organic matter. Indicates the phosphorus concentration in the water. f P This indicates the mass fraction of phosphorus (P) in the feed. η P This indicates the mineralization efficiency of phosphorus (P) during the decomposition of organic matter.

5. The integrated regulation method for fish diseases combining artificial intelligence and mechanistic models as described in claim 2, characterized in that, The fish disease submodule is used to dynamically calculate the fish disease risk index based on the concentration of water quality parameters, and to determine the level of fish diseases based on the fish disease risk index. Specifically: The fish disease submodule calls upon the rate of change of water quality parameters from the abiotic module, phytoplankton module, aquatic plant module, zooplankton module, benthic animal module, and dynamic feeding submodule. Based on the rate of change of water quality parameters output by each module, it updates the concentration of each water quality parameter. Based on the concentration of water quality parameters, it calculates the fish disease risk index, expressed by the formula: In the formula, D ( i () is a fish disease risk index, which represents the probability of fish contracting different levels of diseases. It is the first i Concentration of various water quality parameters C i,threshold It is the first i Safety thresholds for various water quality parameters w i These are weighting coefficients; The disease level corresponding to the highest fish disease risk index is determined as the current fish disease level.

6. The integrated regulation method for fish diseases combining artificial intelligence and mechanistic models as described in claim 1, characterized in that, The deep reinforcement learning algorithm employs a proximal policy optimization algorithm, which uses water quality parameters, environmental parameters, meteorological parameters, and fish disease levels to form the state space of the fishpond ecosystem. The combination of control measures serves as the agent's action space, including oxygenation intensity, water exchange volume, feed amount, and medication dosage. A reward function is used to evaluate the reward value of each action; the reward function expression is as follows: In the formula, It's a reward value. α 1 represents the weight of disease control; This indicates the difference in disease severity between affected and healthy fish. α 2 is the economic cost weight; Cost This indicates the expenses incurred in response to this disease. α 3 is the ecological stability weight. This indicates the ecological stability of a fishpond when diseases occur; The specific steps of using deep reinforcement learning algorithms to search for the optimal solution within the ecological security boundary of each indicator given by the PCLake mechanism model are as follows: Based on the current state of the fishpond ecosystem, a candidate regulatory action is generated. The mechanistic model PCLake is then activated to simulate the ecological process. The simulated future water quality change curve is monitored to check whether any indicators exceed the ecological safety boundary. If no boundary is exceeded and the simulation results tend to improve, the candidate regulatory action is considered safe and feasible, and its reward value is added to the experience pool. If a boundary is exceeded, the candidate regulatory action is marked as a dangerous action, the reward value is negative, and the probability of selecting such a regulatory action is reduced in the strategy update.

7. The integrated regulation method for fish diseases combining artificial intelligence and mechanistic models as described in claim 1, characterized in that, The instructions output by the receiving and control measures scheme are decomposed in time and space, and the decomposed instructions are used to activate relevant devices via the Internet of Things. Specifically: The instructions are decomposed into three dimensions: time axis, spatial domain, and measure set. The time axis is divided into three time periods: 0-2h, 2-6h, and 6-24h. The spatial domain is divided into surface, middle, and bottom layers. The surface layer specifically includes surface aerators to promote gas exchange and inject new external water. The middle layer specifically includes diffusers for targeted drug administration to inhibit pathogens. The bottom layer specifically includes submersible aerators to promote gas exchange and extract aquaculture water containing organic matter. The measure set includes aeration rate, water exchange rate, and drug administration. The decomposed instructions are then used to activate relevant equipment via the Internet of Things (IoT).

8. A comprehensive fish disease control system integrating artificial intelligence and mechanistic models, characterized in that, include: The intelligent data module is used to collect and preprocess data from the fishpond. The mechanism model module is used to establish the mechanism model PCLake of the fishpond ecosystem. The mechanism model PCLake is calibrated and validated using preprocessed fishpond data. The calibrated mechanism model PCLake is used to simulate ecological processes, analyze the simulation results, and obtain the ecological safety boundaries of each indicator. The artificial intelligence module is used to search for the optimal solution within the ecological security boundary of each indicator given by the mechanistic model PCLake using deep reinforcement learning algorithms, and to provide the final control measures. The equipment control module is used to receive the instructions output by the control measures plan, perform spatiotemporal decomposition, and activate the relevant equipment through the Internet of Things.

9. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the integrated fish disease control method that integrates artificial intelligence and mechanistic models as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, The computer program enables the computer to execute the integrated fish disease control method that integrates artificial intelligence and mechanistic models as described in any one of claims 1-7.

Images