A feed mandarin fish factory farming water quality intelligent monitoring and closed-loop control method

Abstract

The present application relates to the field of water quality monitoring and regulation, and particularly relates to a method for intelligent monitoring and closed-loop regulation of water quality in feed mandarin fish factory farming. The method includes the following steps: obtaining and normalizing an original data set, coupling and reconstructing the normalized original data set to generate continuous space data; generating water stratification indexes based on the continuous space data; constructing pollution competition ratios, deviation degrees of current temperatures from optimal growth temperatures, buffer capacity inhibition factors, and acid-base risk mapping factors based on the continuous space data, normalizing them respectively, and further constructing risk entropy; constructing behavior responses based on the continuous space data to generate behavior response values; and constructing potential energy functions based on the behavior response values, risk entropy, and water stratification indexes to obtain regulation directions and comprehensive control output quantities. The method solves the problems of multi-source data dispersion, discontinuous spatial distribution, difficulty in quantifying coupling relationships between key parameters, and lack of biological feedback basis for regulation and decision-making.

Description

Technical Field

[0001] This invention relates to the field of water quality monitoring and control, and in particular to a method for intelligent monitoring and closed-loop control of water quality in factory farming of feed mandarin fish. Background Technology

[0002] In the intensive farming of mandarin fish, water quality is a key factor affecting their growth rate, feeding behavior, and survival rate. Especially under high-density farming conditions, the decomposition of uneaten feed, accumulation of excrement, and microbial metabolism lead to dynamic changes in parameters such as dissolved oxygen, ammonia nitrogen, water temperature, and pH. These parameters exhibit clear coupling relationships; for example, decreased dissolved oxygen concentration weakens the buffering capacity against ammonia toxicity, increased water temperature accelerates metabolic processes and amplifies the ammonia accumulation effect, and pH changes further affect the transformation of toxic ammonia forms. Therefore, water quality is not described by a single parameter but is the result of the combined effects of multiple parameters. In existing technologies, intensive farming typically monitors key indicators using a small number of water quality sensors and employs fixed thresholds for judgment. When an indicator exceeds a set range, aeration or water exchange equipment is activated. However, this method relies heavily on single-point sampling data and cannot reflect the spatial distribution differences of the water body. Especially when water stratification or localized pollution exists, single-point data cannot accurately represent the overall water quality status, easily leading to misjudgments or delayed adjustments.

[0003] Furthermore, existing methods typically set separate control thresholds for parameters such as dissolved oxygen and ammonia nitrogen, lacking a unified evaluation model for each parameter. This results in the inability to reflect the coupled influence of multiple factors during the control decision-making process. For example, even when dissolved oxygen is within the normal range but ammonia nitrogen continues to rise, the system may still be classified as safe, thus delaying the timing of control measures. Simultaneously, traditional control strategies often rely on empirical rules or simple logical control, lacking a continuously changing regulatory mechanism, making them prone to over- or under-regulation, affecting water body stability. Current technologies for water quality control usually rely solely on environmental parameters, failing to incorporate biological feedback information such as fish feeding behavior and physiological states into the control system, leading to discrepancies between the control results and actual aquaculture needs. Therefore, there is an urgent need to provide a smart water quality monitoring and control method to address these issues. Summary of the Invention

[0004] This invention provides a method for intelligent monitoring and closed-loop control of water quality in the factory farming of feed mandarin fish, in order to solve the problems of discrete multi-source data, discontinuous spatial distribution, difficulty in quantifying the coupling relationship between key parameters, and lack of biological feedback basis for control decisions.

[0005] The present invention provides a method for intelligent monitoring and closed-loop control of water quality in intensive mandarin fish farming, which specifically includes the following steps:

[0006] The original dataset is acquired and normalized. The normalized dataset is then coupled and reconstructed to generate continuous spatial data. Based on the continuous spatial data, a water stratification index is generated.

[0007] Based on continuous spatial data, the pollution competition ratio, the deviation of the current temperature from the optimal growth temperature, the buffer capacity inhibition factor, and the pH risk mapping factor are constructed and normalized respectively. Based on the normalized pollution competition ratio, the normalized deviation of the current temperature from the optimal growth temperature, the normalized buffer capacity inhibition factor, and the normalized pH risk mapping factor, a risk entropy is constructed. Based on continuous spatial data, a behavioral response is constructed and a behavioral response value is generated. Based on the behavioral response value, the risk entropy, and the water stratification index, a potential energy function is constructed to obtain the regulation direction and the comprehensive control output.

[0008] Preferably, the aquaculture pond is spatially discretely divided into at least two three-dimensional grid units. Sensor nodes are deployed in each grid unit to synchronously collect dissolved oxygen concentration, ammonia nitrogen concentration, water temperature, and pH value, forming an original data set. After normalizing the original data set, a normalized original data set is obtained, including normalized dissolved oxygen concentration, normalized ammonia nitrogen concentration, normalized water temperature, and normalized pH value.

[0009] Preferably, the Euclidean distance between the target point and each sensing node is calculated, and a basic diffusion function is constructed by combining the spatial attenuation coefficient; based on the Euclidean distance between the target point and each sensing node, a stability term is introduced, and a spatial comprehensive weight term is constructed by combining the basic diffusion function and the vertical periodic perturbation term.

[0010] Preferably, the spatial comprehensive weighting term is combined with the normalized dissolved oxygen concentration, normalized ammonia nitrogen concentration, normalized water temperature, and normalized pH value respectively to calculate the reconstructed values ​​of dissolved oxygen concentration, ammonia nitrogen concentration, water temperature, and pH, thus forming continuous spatial data.

[0011] Preferably, the water stratification index is calculated based on the reconstructed dissolved oxygen concentration value.

[0012] Preferably, a pollution competition ratio is constructed based on the reconstructed values ​​of ammonia nitrogen concentration and dissolved oxygen concentration; the deviation of the current temperature from the optimal growth temperature is constructed based on the reconstructed value of water temperature; a buffer capacity inhibition factor is constructed based on the reconstructed value of dissolved oxygen concentration; and an acid-base risk mapping factor is constructed based on the reconstructed value of pH.

[0013] Preferably, the risk entropy change rate is calculated based on the risk entropy; when both the risk entropy and the risk entropy change rate exceed a preset threshold, regulation is required.

[0014] Preferably, feeding behavior data is introduced, and behavioral responses are constructed by combining the reconstructed values ​​of ammonia nitrogen concentration and dissolved oxygen concentration.

[0015] Preferably, the partial derivative of the potential energy function with respect to dissolved oxygen is taken to obtain the control direction; based on the control direction and combined with the reconstructed ammonia nitrogen concentration value, the comprehensive control output is calculated.

[0016] Preferably, after obtaining the comprehensive control output, a parameter correction mechanism is constructed to update the spatial attenuation coefficient.

[0017] The beneficial effects of the technical solution of the present invention are:

[0018] 1. By introducing a basic diffusion function, a stable term, and a vertical periodic perturbation term to construct a spatial comprehensive weight term, the original data is coupled and reconstructed to obtain a continuous spatial distribution of water quality. This process enables the water quality status at any location in the water body to be obtained through calculation, eliminating the spatial blind spot problem caused by traditional single-point monitoring and significantly improving the completeness and precision of water quality perception.

[0019] 2. By constructing pollution competition ratios, deviations of current temperature from optimal growth temperature, buffer capacity inhibition factors, and pH risk mapping factors, multidimensional water quality parameters are uniformly mapped into dimensionless risk contributions. Furthermore, a probability distribution is formed, and risk entropy is calculated based on information entropy theory. This transforms the originally multivariate, nonlinearly coupled water quality state into a single, quantifiable risk indicator. This process not only preserves the coupling relationships between parameters but also reflects the overall degree of disorder through entropy values, thus avoiding misjudgments caused by isolated parameter assessments in traditional threshold methods and improving the accuracy and robustness of water quality state determination.

[0020] 3. By constructing a potential energy function that includes risk entropy, water stratification index, and behavioral response value, multi-source inputs are uniformly transformed into a single potential energy index. By taking the partial derivative of the potential energy function with respect to dissolved oxygen variable, the gradient direction of state change is obtained, transforming the regulation process from discrete threshold triggering to continuous gradient driving. This enables precise determination of the regulation direction, avoids the problems of overshoot or undershoot in regulation, and improves the stability and response accuracy of regulation. Attached Figure Description

[0021] Figure 1 This is a flowchart of a method for intelligent monitoring and closed-loop control of water quality in the factory farming of feed mandarin fish, as described in this invention. Detailed Implementation

[0022] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0024] The following description, in conjunction with the accompanying drawings, details the specific scheme of the intelligent monitoring and closed-loop control method for water quality in the factory farming of feed mandarin fish provided by this invention.

[0025] See attached document Figure 1 The diagram illustrates a flowchart of a method for intelligent monitoring and closed-loop control of water quality in factory-scale aquaculture of mandarin fish according to an embodiment of the present invention. The method includes the following steps:

[0026] S1. Obtain and normalize the original dataset, couple and reconstruct the normalized dataset to generate continuous spatial data; generate water stratification index based on continuous spatial data;

[0027] The aquaculture pond is divided into several sections using existing regular grid partitioning methods, such as equal-interval partitioning, and spatial coordinate discretization. Each three-dimensional mesh cell contains sensor nodes deployed by technicians. And the spatial coordinates of each grid cell, i.e., the sensing node Spatial coordinates are Sensor nodes Equipped with a dissolved oxygen sensor, an electrochemical sensor, a temperature sensor, and a pH sensor, it operates over time. Simultaneous collection of dissolved oxygen concentration ammonia nitrogen concentration Water temperature and pH value This forms the original dataset. The data in the original dataset is normalized using min-max normalization to obtain the normalized original dataset. This includes the normalized dissolved oxygen concentration. Normalized ammonia nitrogen concentration Normalized water temperature and normalized pH value .

[0028] Furthermore, in order to transform discrete sampled data into continuous spatial data, thereby providing a unified input scale for subsequent risk calculation, the normalized original dataset is coupled and reconstructed. The specific implementation process is as follows:

[0029] First, taking any target point as the calculation object, the spatial coordinates of the target point are... Calculate the Euclidean distance between the target point and each sensor node. Euclidean distance serves as the fundamental variable for spatial influence attenuation. Considering the exponential decay characteristic of dissolved oxygen diffusion in water, and the influence of vertical stratification disturbances, a combined weighting function is constructed. Its formation process is as follows: first, define the basic diffusion function... This expresses the physical law that solute diffusion decreases with distance; then, a stability term is introduced. To avoid the weights from amplifying infinitely when the distance approaches zero; subsequently, a vertical periodic perturbation term is superimposed. The simulation of water stratification and oscillation yields the final dissolved oxygen reconstruction expression: , in, It is in time Spatial coordinates The reconstructed dissolved oxygen concentration value at the location is used as the dissolved oxygen input data under a unified spatial coordinate system. Indicates the number of sensor nodes; This represents the spatial attenuation coefficient, used to control the diffusion rate, and is corrected based on a parameter correction mechanism. This represents the vertical disturbance frequency coefficient, reflecting the intensity of water stratification. It is determined based on the specific application scenario; for example, when the water depth is 0.8m to 1.5m, it is taken as... When the water depth is 1.5m to 2.5m, take The water depth is over 2.5m, take ; It is a spatial comprehensive weight term, describing the first... The combined influence of each sensing node on the dissolved oxygen state at the target point.

[0030] Meanwhile, the normalized ammonia nitrogen concentration was introduced into the same coupling reconstruction method. Normalized water temperature and normalized pH value The reconstructed ammonia nitrogen concentration was calculated. Water temperature reconstruction value and pH reconstruction value and reconstructed values ​​with dissolved oxygen concentration Together they constitute continuous spatial data.

[0031] Subsequently, the water stratification index was calculated based on the reconstructed dissolved oxygen concentration values. The calculation process is as follows: the difference in dissolved oxygen concentration between the upper and lower layers at the same horizontal coordinate is selected and normalized, as expressed by the following formula: , in, It is in time Water stratification index; , , These represent specific sampling depths for the upper, middle, and bottom layers of the water body, respectively. They are derived from preset structural parameters for stratified sampling of the water body, and are specifically defined as follows: This represents a depth of 0.1m to 0.3m below the water surface, used to reflect the area affected by air exchange. As a specific example, it can be taken as 0.2m. The midpoint of the water depth, i.e. ,in This refers to the total depth of the water body, for example, when the water depth is 1.5m. ; The depth is defined as a representative range of 0.1m to 0.2m from the bottom of the pool. As a specific embodiment, it can be set as follows: ; It is in time Spatial coordinates Reconstructed dissolved oxygen concentration value at the location; It is in time Spatial coordinates Reconstructed dissolved oxygen concentration value at the location; It is in time Spatial coordinates Reconstructed dissolved oxygen concentration value at the location; It is the difference in dissolved oxygen concentration between the upper and lower layers at the same horizontal coordinate, and is an approximate expression of the vertical concentration gradient; As a constant used to avoid division by zero errors, in a specific embodiment, it can be set to ; Used to normalize the difference in dissolved oxygen concentration between the upper and lower layers at the same horizontal coordinate.

[0032] S2. Based on continuous spatial data, construct the pollution competition ratio, the deviation of the current temperature from the optimal growth temperature, the buffer capacity inhibition factor, and the pH risk mapping factor, and normalize them respectively; based on the normalized pollution competition ratio, the normalized deviation of the current temperature from the optimal growth temperature, the normalized buffer capacity inhibition factor, and the normalized pH risk mapping factor, construct the risk entropy; based on continuous spatial data, construct the behavioral response and generate the behavioral response value; based on the behavioral response value, the risk entropy, and the water stratification index, construct the potential energy function to obtain the regulation direction and the comprehensive control output.

[0033] After obtaining continuous spatial data and water stratification indices, the risk assessment phase begins, mapping multiple physical quantities to a single risk metric. First, the ammonia nitrogen concentration value is reconstructed. Reconstructed value of dissolved oxygen concentration Combining to build a pollution competition ratio : , The numerator represents the source of toxicity, the denominator represents the total environmental capacity, and the result is the proportion of ammonia nitrogen, which reflects the proportion of toxicity, i.e., the pollution competition ratio.

[0034] Furthermore, based on the reconstructed water temperature value, the degree of deviation of the current temperature from the optimal growth temperature is constructed. The calculation method is as follows: within a set time window, specifically in a 5-minute time window, the reconstructed water temperature value is subtracted from the reference temperature value suitable for mandarin fish growth. , forming a sequence Calculate the variance of the sequence as The suitable temperature reference value for mandarin fish growth is the ratio of the suitable temperature baseline value for mandarin fish growth to the maximum allowable temperature for mandarin fish growth, such as 26℃ / 32℃; the suitable temperature baseline value for mandarin fish growth is a fixed reference parameter, such as 26℃.

[0035] Next, perform buffering capacity mapping on dissolved oxygen: , in, This represents the result of buffering capacity mapping for dissolved oxygen, i.e., the buffering capacity inhibition factor. The above formula uses the square root function to reduce the marginal effect of high dissolved oxygen conditions.

[0036] For pH value, considering its periodic fluctuation characteristics, an acid-base risk mapping factor is constructed. :

[0037] By introducing the sine function The pH reconstruction value is mapped to the range of [-1,1] and output as a continuously changing periodic function. The change of pH reconstruction value is transformed into a response signal with fluctuation characteristics. Then, by adding 1 and dividing by 2, the output range is shifted and compressed from [-1,1] to [0,1], making it a standardized risk dimension that describes the intensity ratio of the influence of the current pH state on the stability of the water body.

[0038] Pollution competition ratio The degree of deviation of the current temperature from the optimal growth temperature Buffer capacity inhibitor pH risk mapping factor This forms the basis of the probability distribution, which is then further normalized: , The normalized pollution competition ratio was obtained. The degree of deviation of the normalized current temperature from the optimal growth temperature Normalized buffer capacity inhibition factor and normalized pH risk mapping factor and ensure Then, risk entropy is constructed. :

[0039] Risk Entropy As a quantitative result of disorder, it is directly used as the input variable for the next stage. At the same time, the first-order difference method is used to calculate the risk entropy change rate through forward difference. When both the risk entropy and the risk entropy change rate exceed the thresholds preset by technicians, such as the risk entropy threshold being set to 0.9 and the risk entropy change rate threshold being set to 0.06, it indicates that the water body is in an unstable state and needs to be regulated.

[0040] Furthermore, in order to incorporate biofeedback into regulation, feeding behavior data was introduced to construct behavioral responses: , in, It is a behavioral response value that describes the combined effect of the feeding drive intensity and environmental inhibition effect of fish under current water quality conditions; This refers to the number of feeding behaviors of mandarin fish. It is obtained by continuously collecting videos of fish activity by deploying underwater cameras, such as Hikvision DS-2CD2043G0-I, in the breeding pond. Existing motion analysis methods, such as YOLOv5 or background subtraction, are used to identify the movement trajectory of the fish. At the same time, feeding actions are determined by preset rules, such as rapid forward movement of the fish and sudden changes in local contours. Each behavior that meets the conditions is counted as a feeding event. Finally, all feeding events are accumulated and statistically analyzed within a set time window, such as a time window of 5 minutes. The standardized critical dissolved oxygen concentration for feeding is obtained by standardizing the critical dissolved oxygen concentration for feeding of mandarin fish using a Z-score with a mean of 0 and a standard deviation of 1. The critical dissolved oxygen concentration for feeding of mandarin fish is between 3.5 and 4.5 mg / L. As a specific example, 4.0 mg / L is used. This is the response slope coefficient, describing the sensitivity of feeding behavior to changes in dissolved oxygen. It was obtained through five-fold cross-validation, and the reference value range is [value range missing]. As a specific embodiment, we take 1.5; It is a dissolved oxygen-driven term used to describe the rapid increase in feeding behavior after dissolved oxygen reaches a critical level; This is the feeding intensity term, used to reduce the impact of extreme values; It is an ammonia nitrogen inhibitor, describing the inhibitory effect of ammonia nitrogen on feeding behavior.

[0041] Furthermore, in order to convert the behavioral response value Risk entropy and water stratification index To transform this into a unified control direction and construct a potential energy function, the specific process is as follows: First, the square of the risk entropy is used. Strengthen risk impact; secondly, introduce hierarchical logic functions. To limit the scope of tiered contributions; then add To smooth out behavioral fluctuations; finally, a temperature stability term is introduced. The potential energy function is obtained by combining the results: , in, It is the composite potential function, describing the time... The degree to which the state of the water body deviates from the ideal stable state.

[0042] To achieve closed-loop control, the partial derivative of the potential energy function with respect to the dissolved oxygen variable is taken. The derivation is based on the chain rule. , , By analyzing the dependence on dissolved oxygen, the direction of regulation is obtained:

[0043] Each partial derivative term is expanded layer by layer using the chain rule, ensuring that all preceding variables participate in the calculation. The final control quantity is then formed. , in, Indicates time The comprehensive control output, that is, the quantitative indicator of the intensity of water body regulation; The step size coefficient is used to adjust the response speed and is preset by professional technicians, such as 0.1. The pollution compensation coefficient is based on the reconstructed values ​​of ammonia nitrogen concentrations across all aquaculture areas. mean Sure, ; It is a regulatory driving force based on the potential energy gradient, describing the marginal impact rate of dissolved oxygen changes on the overall risk potential. It is a pollution-driven compensatory control term that describes the contribution of ammonia nitrogen toxicity to the control demand. It is used as an input to the actuator and directly mapped to the actuator's actions.

[0044] Furthermore, a parameter correction mechanism is constructed to eliminate model errors. A correction function is built using the difference between the target dissolved oxygen concentration and the reconstructed actual dissolved oxygen concentration. , in, It is the update result of the next time step, that is, the optimal spatial decay estimate after the current water quality state error correction; The learning rate is a preset control parameter, such as 0.05. The target dissolved oxygen concentration is determined by consulting aquaculture standards. As a specific example, it can be taken as 7. It is the adjusted parameter update increment.

[0045] Finally, the updated results are applied again to the spatial decay coefficient in the original coupling reconstruction formula to achieve adaptive closed loop.

[0046] In summary, a method for intelligent monitoring and closed-loop control of water quality in the factory farming of feed-fed mandarin fish has been developed.

[0047] The order of the embodiments is for illustrative purposes only and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0048] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0049] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for intelligent monitoring and closed-loop control of water quality in factory-scale aquaculture of feed-fed mandarin fish, characterized in that, Includes the following steps: The original data set is acquired and normalized, and then coupled and reconstructed to generate continuous spatial data. Water stratification indices are generated based on continuous spatial data. Based on continuous spatial data, we construct pollution competition ratio, deviation of current temperature from optimal growth temperature, buffer capacity inhibition factor, and pH risk mapping factor, and normalize them respectively. Risk entropy is constructed based on the normalized pollution competition ratio, the normalized deviation of the current temperature from the optimal growth temperature, the normalized buffer capacity inhibition factor, and the normalized pH risk mapping factor. Based on continuous spatial data, construct behavioral responses and generate behavioral response values; Based on behavioral response values, risk entropy, and water stratification index, a potential energy function is constructed to obtain the regulation direction and comprehensive control output.

2. The method for intelligent monitoring and closed-loop control of water quality in factory-scale mandarin fish farming according to claim 1, characterized in that, The aquaculture pond is spatially discretized into at least two three-dimensional grid units. Sensor nodes are deployed in each grid unit to synchronously collect dissolved oxygen concentration, ammonia nitrogen concentration, water temperature, and pH value, forming the original data set. After normalizing the original data set, the normalized original data set is obtained, including the normalized dissolved oxygen concentration, normalized ammonia nitrogen concentration, normalized water temperature, and normalized pH value.

3. The method for intelligent monitoring and closed-loop control of water quality in factory-scale mandarin fish farming according to claim 2, characterized in that, Calculate the Euclidean distance between the target point and each sensing node, and construct the basic diffusion function by combining the spatial attenuation coefficient; Based on the Euclidean distance between the target point and each sensing node, a stability term is introduced, and a spatial comprehensive weight term is constructed by combining the basic diffusion function and the vertical periodic perturbation term.

4. The method for intelligent monitoring and closed-loop control of water quality in factory-scale mandarin fish farming according to claim 3, characterized in that, By combining the spatial comprehensive weighting term with the normalized dissolved oxygen concentration, normalized ammonia nitrogen concentration, normalized water temperature, and normalized pH value respectively, the reconstructed values ​​of dissolved oxygen concentration, ammonia nitrogen concentration, water temperature, and pH are calculated, thus forming continuous spatial data.

5. The method for intelligent monitoring and closed-loop control of water quality in factory-scale mandarin fish farming according to claim 4, characterized in that, The water stratification index is calculated based on the reconstructed dissolved oxygen concentration value.

6. The method for intelligent monitoring and closed-loop control of water quality in factory-scale mandarin fish farming according to claim 4, characterized in that, Based on the reconstructed values ​​of ammonia nitrogen concentration and dissolved oxygen concentration, a pollution competition ratio is constructed; based on the reconstructed value of water temperature, the deviation of the current temperature from the optimal growth temperature is constructed; based on the reconstructed value of dissolved oxygen concentration, a buffering capacity inhibition factor is constructed; based on the reconstructed value of pH, an acidity / alkalinity risk mapping factor is constructed.

7. The method for intelligent monitoring and closed-loop control of water quality in factory-scale mandarin fish farming according to claim 1, characterized in that, Based on risk entropy, the rate of change of risk entropy is calculated; when both risk entropy and the rate of change of risk entropy exceed a preset threshold, regulation is required.

8. The method for intelligent monitoring and closed-loop control of water quality in factory-scale mandarin fish farming according to claim 4, characterized in that, By incorporating feeding behavior data and combining reconstructed values ​​of ammonia nitrogen concentration and dissolved oxygen concentration, a behavioral response is constructed.

9. The method for intelligent monitoring and closed-loop control of water quality in factory-scale mandarin fish farming according to claim 4, characterized in that, The partial derivative of the potential function with respect to dissolved oxygen is used to obtain the control direction; based on the control direction and combined with the reconstructed ammonia nitrogen concentration, the comprehensive control output is calculated.

10. The method for intelligent monitoring and closed-loop control of water quality in factory-scale mandarin fish farming according to claim 3, characterized in that, After obtaining the comprehensive control output, a parameter correction mechanism is constructed to update the spatial attenuation coefficient.

Images