A method and system for coordinated regulation of water bodies for aquatic population reproduction and seedling raising
By extracting features from multi-view underwater camera array data, constructing a biochemical composite load index, generating a dynamic shear stress safety limit, and coordinating the control of aeration equipment and pH compensation, the problems of hysteresis response, mechanical damage, and pH fluctuation in the regulation of aquatic population reproduction and seedling raising were solved, and adaptive water quality regulation was achieved.
Patent Information
- Application Number
- CN202611394340.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-09
- Publication Date
- 2026-11-17
AI Technical Summary
Existing aquatic population breeding and seedling control systems suffer from problems such as delayed passive oxygenation response, mechanical aeration shear force that can easily damage fertilized eggs, pH fluctuations caused by aeration and deaeration, and a lack of adaptive correction mechanisms in fixed parameter models.
By acquiring multi-view underwater camera array data, optical flow motion features and image grayscale features are extracted to construct a biochemical composite load index. Combined with real-time water temperature and fertilized egg hardening process, a dynamic shear stress safety upper limit is generated. This is used to coordinate the control of oxygenation equipment and pH compensation to achieve feedforward control, and closed-loop parameter updates are performed after the process ends.
It achieved accurate prediction of oxygen consumption, avoided mechanical damage to fertilized eggs, stabilized the pH value of the water, adapted to the individual characteristics differences of different breeding batches, and reduced prediction bias in long-term operation.
Smart Images

Figure CN123100535A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of aquaculture environmental control technology, specifically to a method and system for the coordinated regulation of aquatic group breeding and seedling water bodies. Background Technology
[0002] Group breeding and seedling raising are crucial aspects of aquaculture. During the breeding season, the concentrated release of reproductive substances by the cultured population can cause sudden changes in the biochemical oxygen demand of the water. Existing water quality control methods typically rely on real-time measurements from dissolved oxygen sensors to passively control aeration equipment. Due to the physical delay between biochemical reactions and water mixing, by the time the sensor detects a decrease in dissolved oxygen concentration, the water body is often already in a state of high oxygen deficiency. This delayed, passive feedback control is insufficient to meet the instantaneous high oxygen demand during concentrated breeding.
[0003] To quickly replenish dissolved oxygen, conventional control logic typically increases the output power of mechanical turbulent aerators. However, during the hardening stage after fertilized eggs are released, the egg membrane structure is highly sensitive to fluid shear stress. The shear force generated by high-power mechanical aeration can easily cause physical damage to the fertilized eggs, affecting the final hatching rate. Simultaneously, high-intensity physical aeration accelerates the escape of dissolved carbon dioxide from the water, disrupting the original carbonate balance and causing pH fluctuations. Most current control systems execute dissolved oxygen supply and pH adjustment as independent loops, failing to proactively compensate for the deaeration effect of the aeration equipment.
[0004] Furthermore, different batches of aquatic populations exhibit objective physiological differences in swimming aggregation and substance release during the breeding season. Existing control models mostly employ fixed parameters and lack closed-loop correction mechanisms based on actual water quality feedback. When faced with changes in operating conditions across different breeding cycles, the fixed parameter models cannot autonomously correct prediction biases based on historical residuals from previous operations, resulting in limited adaptability of the control system during long-term operation. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides a method and system for the coordinated regulation of aquatic group breeding and seedling raising water bodies. It solves the problems of passive oxygenation response lag, mechanical aeration shear force easily damaging fertilized eggs, aeration deaeration causing pH fluctuations in water bodies, and the lack of adaptive correction mechanism in fixed parameter models in the existing aquatic group breeding and seedling raising regulation.
[0006] To achieve the above objectives, the present invention provides the following technical solution:
[0007] The first aspect of this invention provides a method for the coordinated regulation of water bodies for aquatic population reproduction and seedling raising, comprising the following steps:
[0008] Acquire video stream data of aquaculture water bodies collected by multi-view underwater camera arrays, extract optical flow motion features and image grayscale features from video frame sequences of video stream data, and obtain population movement aggregation index, reproductive material release intensity characterization value and effective optical flow feature point distribution density parameter;
[0009] The initiation of the reproductive substance release process is determined based on the population movement convergence index, and the initiation time is determined. The population movement convergence index and the reproductive substance release intensity characterization value are fused based on the effective optical flow feature point distribution density parameter to obtain the biochemical composite load index. The biochemical composite load index is integrated to obtain the predicted oxygen consumption increment, the predicted peak oxygen consumption increment, and the estimated value of the fertilized egg equivalent average density. The feedforward control time window is determined in combination with the real-time water temperature and the initiation time.
[0010] Based on the estimated equivalent average density of fertilized eggs and the hardening process of fertilized eggs over time, a dynamic shear stress safety upper limit sequence that varies with time is generated.
[0011] Within the feedforward control time window, the total oxygenation power demand is determined based on the predicted oxygen consumption increment, and the input power of the micro-nano bubble generator and the mechanical turbulent aerator is allocated under the constraint of the dynamic shear stress safety upper limit sequence. At the same time, the variable frequency buffer metering pump is controlled to perform feedforward pH compensation based on the operating status of the micro-nano bubble generator and the mechanical turbulent aerator.
[0012] After the feedforward control time window ends, the equivalent biochemical oxygen demand is estimated by combining the actual dissolved oxygen measurement value and the actual oxygenation output of the micro-nano bubble generator and the mechanical turbulent aerator. The oxygen demand prediction residual between the equivalent biochemical oxygen demand and the predicted peak oxygen demand is calculated. Based on the oxygen demand prediction residual, the preset calibration parameters to be calibrated in feature extraction and load prediction are updated in a closed loop.
[0013] This invention constructs an oxygen consumption prediction model by extracting biological movement and substance release features from visual data. Based on the constraint of the output power of mechanical equipment by the biological hardening process, it achieves coordinated control of dissolved oxygen supply, physical shear force limitation and water pH. The prediction parameters are updated in a closed loop using a filtering algorithm.
[0014] Furthermore, the population movement convergence index, reproductive substance release intensity characterization value, and effective optical flow feature point distribution density parameters are obtained, including:
[0015] The global average motion intensity is obtained by calculating the average of the sum of squares of the velocity components of each effective optical flow feature point. The spatial divergence of the optical flow motion vector field is calculated to obtain the optical flow divergence variance. The global average motion intensity and the optical flow divergence variance are linearly combined to obtain the group motion convergence index. The gray-level histogram of the video frame is statistically analyzed to calculate the image gray-level information entropy. After smoothing the image gray-level information entropy sequence, the first-order time derivative is calculated, and the absolute value of the first-order time change is taken as the characterization value of the reproductive material release intensity. The effective optical flow feature point distribution density parameter is obtained by dividing the total number of effective optical flow feature points in the current video frame by the total number of pixels in the image.
[0016] By using the above method, the spatial aggregation and swimming state of the group during mating is quantified by the variance of optical flow divergence and the average motion intensity. At the same time, the changes in water turbidity are captured by the time change rate of gray information entropy, thereby converting the behavior of biological groups and the process of material release into calculable digital features.
[0017] Furthermore, obtaining the biochemical compound load index includes:
[0018] The Sigmoid function is used to establish a continuous mapping relationship between the effective optical flow feature point distribution density parameter and the dynamic feature effectiveness weight. The product of the population movement aggregation index and the dynamic feature effectiveness weight is calculated. The product of the difference between the number 1 and the dynamic feature effectiveness weight, the characterization value of reproductive material release intensity, and the preset feature scale matching coefficient is calculated. The two products are added together to obtain the biochemical composite load index.
[0019] This step utilizes a mapping function to determine the weight of different characteristics in the current water body state, and nonlinearly integrates the dynamic group movement state with the static material release characteristics to generate an evaluation index characterizing the intensity of water body load sources.
[0020] Furthermore, obtaining the predicted oxygen consumption increment, the predicted peak oxygen consumption increment, the estimated equivalent average density of fertilized eggs, and the feedforward control time window includes:
[0021] Starting from the initiation of the reproductive substance release process, the biochemical composite load index is integrated over time. The time integration result is multiplied by the preset oxygen consumption integral mapping coefficient and fertilized egg density integral mapping coefficient to obtain the predicted oxygen consumption increment and the estimated equivalent average density of fertilized eggs updated over time. The maximum recorded predicted oxygen consumption increment is determined as the predicted peak oxygen consumption increment. The real-time water temperature collected by the temperature sensor is read, and the biochemical oxygen consumption response delay time corresponding to the real-time water temperature is calculated based on the Arrhenius relation. The time period from the initiation of the reproductive substance release process to the end of the biochemical oxygen consumption response delay time is determined as the feedforward control time window.
[0022] The above process converts the time integral of the biochemical composite load index into physical quantities of predicted oxygen consumption and fertilized egg density, and calculates the biochemical reaction delay time under actual water temperature conditions by combining thermodynamic relationships, thereby providing a clear range of action for the power allocation of subsequent aeration equipment.
[0023] Furthermore, generating the time-varying dynamic shear stress safety upper limit sequence includes:
[0024] By subtracting the product of the estimated equivalent average density of the fertilized egg and the preset density correction coefficient from the preset basic shear stress tolerance value of the single unit, the initial shear stress tolerance threshold is obtained. The time difference between the current calculation time and the start time is obtained. Based on the exponential asymptotic relationship between the shear stress tolerance of the fertilized egg and the hardening time, combined with the initial shear stress tolerance threshold, the upper limit of conventional shear stress and the hardening rate constant of the egg membrane, the dynamic shear stress safety upper limit corresponding to different time nodes is calculated, thus forming a dynamic shear stress safety upper limit sequence.
[0025] The initial threshold is corrected by introducing the influence of population density on the collision force of individual cells, and a dynamic shear stress tolerance model is established based on the physical process of egg membrane water absorption, expansion and solidification, so as to provide a physical safety boundary for the operation and control of subsequent aeration equipment.
[0026] Furthermore, the input power allocation for the micro / nano bubble generator and the mechanical turbulence aerator includes:
[0027] When the total oxygenation power demand is not greater than the rated upper limit power of the micro-nano bubble generator, the total oxygenation power demand is used as the feedforward input power of the micro-nano bubble generator; when the total oxygenation power demand is greater than the rated upper limit power of the micro-nano bubble generator, the rated upper limit power of the micro-nano bubble generator is used as the feedforward input power, and the difference between the total oxygenation power demand and the rated upper limit power of the micro-nano bubble generator is used as the supplementary input power of the mechanical turbulent flow aerator; using the pre-established inverse mapping relationship between shear stress and power, the maximum allowable input power corresponding to the current dynamic shear stress safety upper limit is determined, and the supplementary input power of the mechanical turbulent flow aerator is limited using the maximum allowable input power.
[0028] This step prioritizes assigning low-shear-stress micro-nano bubble generators to handle basic oxygen supply. When high-power mechanical turbulence aerators are required, dynamic shear force mapping is forcibly applied to prevent the oxygenation process from damaging the fertilized egg structure.
[0029] Furthermore, controlling the variable frequency buffer metering pump to perform feedforward pH compensation includes:
[0030] Calculate the product of the input power of the mechanical turbulent aerator and its corresponding dissipation coefficient, and calculate the product of the input power of the micro / nano bubble generator and its corresponding dissipation coefficient. Summate the two products to obtain the equivalent dissipation rate of dissolved carbon dioxide. Calculate the quotient of the equivalent dissipation rate of dissolved carbon dioxide and the equivalent concentration of pH buffer solution in the storage tank to obtain the buffer injection volume rate of the variable frequency buffer metering pump, and control the operation of the variable frequency buffer metering pump. During the operation of the variable frequency buffer metering pump, read the real-time measurement value of the pH sensor, and stop the injection when the real-time measurement value reaches or falls below the lower limit of the preset allowable pH range.
[0031] The escape rate of dissolved carbon dioxide was calculated based on the degassing characteristics of two different types of aeration equipment. The required buffer injection rate was simultaneously derived and feedforward compensation was performed to offset the fluctuations in water pH caused by aeration operations.
[0032] Furthermore, the calculation of the oxygen consumption prediction residuals includes:
[0033] The baseline dissolved oxygen level is determined based on the steady-state dissolved oxygen data of the aquaculture pond before the reproductive event is triggered. The baseline dissolved oxygen level is then subtracted from the actual dissolved oxygen measurement at the end of the feedforward control time window, and the equivalent dissolved oxygen increment is calculated by converting the total electrical energy actually input by the micro-nano bubble generator and the mechanical turbulent aerator within the feedforward control time window. The equivalent biochemical oxygen consumption is then subtracted from the peak increment of the predicted oxygen consumption to obtain the oxygen consumption prediction residual.
[0034] This step utilizes a conversion model from total input electrical energy to dissolved oxygen, combined with actual sensor observations, to reconstruct the true biochemical oxygen consumption of the water body, thereby obtaining the oxygen consumption residual between the feedforward prediction and the actual water environment.
[0035] Furthermore, the closed-loop update of the preset calibration parameters includes:
[0036] A state vector consisting of feature calibration coefficients and feature scale matching coefficients is constructed; a prior state estimate for the current breeding cycle is obtained based on the state transition matrix, and an observation matrix predetermined based on historical samples is acquired; the Kalman gain is calculated by combining the prior state estimate, the observation matrix, and the preset error covariance; the oxygen consumption prediction residual is used as the observation input, and the prior state estimate is updated posteriorly by combining the Kalman gain; the update result is subjected to a limiting process with a preset allowable range; and the parameters after limiting are used to replace the preset calibration parameters to be calibrated and participate in the feature calculation and load prediction for the next breeding cycle.
[0037] By introducing a state estimation filtering method to handle observation noise and model drift, periodic posterior correction of the mapping parameters is completed, ensuring that the control system can adapt to the individual characteristic differences of aquatic populations in different breeding batches.
[0038] A second aspect of the present invention provides a coordinated regulation system for aquatic population reproduction and seedling raising water bodies, applied to the aforementioned coordinated regulation method for aquatic population reproduction and seedling raising water bodies, comprising:
[0039] The multi-dimensional feature extraction module acquires video stream data of aquaculture water bodies collected by a multi-view underwater camera array, extracts optical flow motion features and image grayscale features from the video frame sequence of the video stream data, and obtains the population movement aggregation index, reproductive material release intensity characterization value and effective optical flow feature point distribution density parameter.
[0040] The feedforward load prediction module determines the start of the reproductive substance release process and the start time based on the population movement convergence index. It integrates the population movement convergence index and the reproductive substance release intensity characterization value based on the effective optical flow feature point distribution density parameter to obtain the biochemical composite load index. It integrates the biochemical composite load index to obtain the predicted oxygen consumption increment, the predicted peak oxygen consumption increment, and the estimated value of the fertilized egg equivalent average density. It then determines the feedforward control time window by combining the real-time water temperature and the start time.
[0041] The oxygenation and pH compensation coordinated control module determines the total oxygenation power demand based on the predicted oxygen consumption increment within the feedforward control time window, and allocates the input power of the micro-nano bubble generator and the mechanical turbulent aerator under the constraint of the dynamic shear stress safety upper limit sequence. At the same time, it controls the variable frequency buffer metering pump to perform feedforward pH compensation based on the operating status of the micro-nano bubble generator and the mechanical turbulent aerator.
[0042] The closed-loop parameter calibration module estimates the equivalent biochemical oxygen demand (BOD) by combining the actual dissolved oxygen measurement value and the actual oxygenation output of the micro-nano bubble generator and the mechanical turbulent aerator after the feedforward control time window ends. It then calculates the oxygen demand prediction residual between the equivalent BOD and the predicted peak oxygen demand and performs closed-loop updates on the preset calibration parameters in feature extraction and load prediction based on the oxygen demand prediction residual.
[0043] This invention provides a method and system for the coordinated regulation of water bodies in aquatic population breeding and seedling raising. It has the following beneficial effects:
[0044] 1. This invention extracts optical flow motion features and image grayscale features from multi-view underwater video frames to calculate the biochemical composite load index and determines the feedforward control time window by combining it with real-time water temperature. This method directly converts the behavior of biological groups into predicted values of oxygen consumption and egg density, and determines the operating range and power requirements of aeration equipment in advance before the actual biochemical oxygen consumption in the water body occurs, thus solving the problem of sudden water hypoxia caused by the concentrated release of reproductive substances by aquaculture groups.
[0045] 2. This invention generates a dynamic shear stress safety upper limit sequence based on the hardening process of the fertilized egg over time, thereby limiting the supplementary input power of the mechanical turbulent aerator and prioritizing the scheduling of micro-nano bubble generators with low shear force. This collaborative control mechanism avoids mechanical damage to the fertilized egg caused by single high-power aeration; simultaneously, based on the operating status of the two types of aeration equipment, the equivalent carbon dioxide emission rate is calculated, and the buffer metering pump is synchronously controlled to perform feedforward compensation, offsetting the pH drift caused by physical aeration.
[0046] 3. After the feedforward control time window ends, this invention derives the equivalent biochemical oxygen consumption by combining the actual dissolved oxygen measurement value and the actual oxygenation output of the equipment, and calculates the oxygen consumption prediction residual between the peak increment of the predicted oxygen consumption. The system uses this residual as the observation input, and combines it with Kalman gain to perform closed-loop posterior update of the calibration parameters to be calibrated in feature extraction and load prediction, enabling the control model to autonomously adapt to the individual characteristic differences of different batches of aquatic populations, and reducing prediction bias during long-term operation. Attached Figure Description
[0047] Figure 1 This is a schematic diagram of the system architecture of the present invention;
[0048] Figure 2 This is a schematic diagram of the method flow of the present invention;
[0049] Figure 3 This is a schematic diagram of the multidimensional feature extraction process of the present invention;
[0050] Figure 4 This is a schematic diagram of the feedforward load prediction process of the present invention;
[0051] Figure 5 This is a schematic diagram of the physical boundary modeling process of the present invention;
[0052] Figure 6 This is a schematic diagram of the decoupling and collaborative control process of the present invention;
[0053] Figure 7 This is a schematic diagram of the closed-loop state calibration process of the present invention;
[0054] Figure 8 This is a schematic diagram illustrating the time-series comparison and evaluation of dissolved oxygen concentration according to the present invention;
[0055] Figure 9 This is a schematic diagram of power decoupling allocation under dynamic shear boundary constraints according to the present invention. Detailed Implementation
[0056] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.
[0057] Reference Figure 1 This invention provides a coordinated regulation system for aquatic population breeding and seedling raising water bodies, including: a multi-dimensional feature extraction module, a feedforward load prediction module, a shear safety boundary calculation module, an oxygenation and pH compensation coordinated control module, and a closed-loop parameter calibration module.
[0058] The aquatic group breeding and seedling raising water body collaborative regulation system of this embodiment is set in the aquaculture pond and includes a sensing end, an execution end and a computing control end.
[0059] The sensing end includes a multi-view underwater camera array and a multi-parameter water quality sensor. The multi-view underwater camera array has at least two underwater cameras arranged orthogonally along their optical axes to acquire video stream data of the aquaculture water body from different perspectives. The multi-parameter water quality sensor includes a dissolved oxygen sensor, a pH sensor, and a temperature sensor, which are used to collect dissolved oxygen concentration, pH value, and water temperature of the aquaculture water body, respectively.
[0060] The actuators include a micro / nano bubble generator, a mechanical turbulence aerator, and a variable frequency buffer metering pump. The micro / nano bubble generator is used to oxygenate the aquaculture water under conditions of low macroscopic hydraulic shear stress; the mechanical turbulence aerator is used to drive the circulation of the aquaculture water and supplement oxygenation through mechanical aeration that generates greater local hydraulic shear stress. The variable frequency buffer metering pump is connected to the storage tank and is used to inject pH buffer solution into the aquaculture water according to control commands.
[0061] The computing control unit includes an edge computing gateway, a processor, and an industrial control bus. The processor executes the computing programs corresponding to each functional module; the sensing and execution ends interact with the computing control unit via the industrial control bus to exchange data and control commands.
[0062] During system operation, the sensing end continuously collects video stream data and water quality parameters of the aquaculture water body and sends the collection results to the computing and control end.
[0063] The multidimensional feature extraction module processes the video stream data, obtains the optical flow motion vector representing the movement state of the aquatic group through optical flow calculation, and calculates the gray-scale information entropy of the water body image based on the gray-scale distribution of the video frames. Thus, it obtains the basic feature parameters used to characterize the movement state of the aquatic group and the changes in the optical state of the water body caused by the release of reproductive substances.
[0064] The feedforward load prediction module receives the basic feature parameters output by the multidimensional feature extraction module. When the population movement convergence index meets the preset reproductive event triggering conditions, the feedforward load prediction module fuses the basic feature parameters, calculates the predicted oxygen consumption increment updated over time, and determines the peak predicted oxygen consumption increment based on the maximum predicted oxygen consumption increment within a reproductive cycle, while simultaneously obtaining the estimated value of the equivalent average density of fertilized eggs.
[0065] The feedforward load prediction module also reads the real-time water temperature collected by the temperature sensor, calculates the biochemical oxygen consumption response delay time corresponding to the biochemical oxygen consumption process after the release of reproductive substances based on the real-time water temperature, and determines the feedforward control time window corresponding to this reproductive event based on the biochemical oxygen consumption response delay time.
[0066] The shear safety boundary calculation module receives the estimated equivalent average density of fertilized eggs from the feedforward load prediction module. Based on the estimated equivalent average density of fertilized eggs and the hydration and hardening characteristics of the egg membrane over time, it determines the initial shear stress tolerance threshold at the beginning of ovulation and gradually relaxes the upper limit of shear stress that can be applied to the aquaculture water as the fertilized eggs harden, thus forming a dynamic shear stress safety upper limit sequence.
[0067] The oxygenation and pH compensation coordinated control module receives the predicted oxygen consumption increment and the dynamic shear stress safety upper limit sequence. Based on the change in the predicted oxygen consumption increment, it determines the current total oxygenation power demand and generates constrained power allocation instructions for the two types of aeration equipment according to the control mode of prioritizing the oxygenation load of micro-nano bubble generators and supplementing the remaining oxygenation load by mechanical turbulent flow aerators constrained by the dynamic shear stress safety upper limit.
[0068] The oxygenation and pH compensation coordinated control module also estimates the equivalent dissolved carbon dioxide emission rate caused by the operation of the micro-nano bubble generator and the mechanical turbulent aerator based on their actual output power, and determines the pH buffer compensation requirement accordingly, generating buffer injection commands for the variable frequency buffer metering pump. The micro-nano bubble generator, mechanical turbulent aerator, and variable frequency buffer metering pump then execute oxygenation and pH buffer injection within the feedforward control time window based on the corresponding commands.
[0069] After the feedforward control time window ends, the closed-loop parameter calibration module reads the actual dissolved oxygen measurement value collected by the dissolved oxygen sensor, and determines the equivalent biochemical oxygen demand corresponding to this breeding cycle by combining the actual oxygenation output of the two types of aeration equipment during this feedforward control period.
[0070] The closed-loop parameter calibration module compares the equivalent biochemical oxygen demand (BOD) with the predicted peak BOD increment for the current reproductive cycle to obtain the BOD prediction residual. This residual is then used to iteratively update the calibration parameters in the multidimensional feature extraction module and the feedforward load prediction module. The updated calibration parameters are then used for feature extraction and load prediction in the next reproductive cycle, thus forming a closed-loop calibration of prediction parameters across reproductive cycles.
[0071] See attached document Figure 2 , Figure 2 This is a flowchart illustrating a method for coordinated regulation of water bodies in aquatic population breeding and seedling raising according to an embodiment of the present invention. The present invention provides a method for coordinated regulation of water bodies in aquatic population breeding and seedling raising, which may include:
[0072] Step S1: Extract multi-dimensional features. Real-time video stream data of the aquaculture water body collected by a multi-view underwater camera array is acquired. After time alignment of the video streams from different perspectives, an optical flow algorithm is used to extract optical flow motion vectors from the video frame sequence. Based on the optical flow motion vectors, a group movement aggregation index is calculated to characterize the activity level and aggregation / dispersion state of the aquaculture group. Simultaneously, the image grayscale information entropy of the video frames is calculated. The intensity of reproductive material release is obtained based on the change in image grayscale information entropy over time, and the distribution density parameter of effective optical flow feature points at the current moment is recorded.
[0073] Step S2, predict feedforward load. Continuously monitor the population movement convergence index. When the population movement convergence index exceeds the preset steady-state baseline and the duration reaches the preset time judgment window, it is determined that the current reproductive material release process has started, and the corresponding start time is recorded.
[0074] The dynamic feature effectiveness weights are determined based on the effective optical flow feature point distribution density parameters. These weights are then used to weight and fuse the population movement aggregation index and the reproductive material release intensity characterization value to obtain the biochemical composite load index. The biochemical composite load index is integrated over time from the start-up moment to obtain the predicted oxygen consumption increment and the estimated equivalent average density of fertilized eggs updated over time. The maximum predicted oxygen consumption increment recorded within the feedforward control time window of a reproductive cycle is determined as the peak predicted oxygen consumption increment for that cycle.
[0075] The system reads the real-time water temperature collected by the temperature sensor, calculates the biochemical oxygen demand response delay time corresponding to the release of reproductive substances based on the Arrhenius relation, and determines the feedforward control time window based on the biochemical oxygen demand response delay time.
[0076] Step S3: Calculate the shear safety boundary. Obtain the estimated equivalent average density of fertilized eggs obtained in Step S2. Based on the estimated equivalent average density of fertilized eggs and the physical characteristics of the egg membrane gradually hydrating and hardening after fertilization, determine the initial shear stress tolerance threshold corresponding to the early stage of ovulation. According to the relationship between the shear stress tolerance of fertilized eggs and the increase in hardening time, the upper limit of the shear stress that can be applied to the aquaculture water is gradually relaxed according to an exponential asymptotic law, generating a dynamic shear stress safety upper limit sequence that changes with time. This sequence is used to limit the hydraulic shear stress that the mechanical turbulent aerator can generate at different time points.
[0077] Step S4: Perform coordinated control of oxygenation and pH compensation. Within the feedforward control time window, determine the current total oxygenation power demand based on the rate of change of the predicted oxygen consumption increment continuously updated over time in step S2. The micro-nano bubble generator is given priority in handling the oxygenation load; when the available output of the micro-nano bubble generator is insufficient to meet the total oxygenation power demand, under the dynamic shear stress safety upper limit constraint generated in step S3, the remaining oxygenation load is supplemented by the mechanical turbulent flow aerator, thereby generating constrained power allocation commands corresponding to the micro-nano bubble generator and the mechanical turbulent flow aerator.
[0078] The equivalent rate of dissolved carbon dioxide dissipation caused by the aeration process is estimated based on the output power of the two types of aeration equipment, and the pH buffer compensation amount is determined accordingly. A variable frequency buffer metering pump is then used to inject pH buffer into the aquaculture water. During buffer injection, the real-time measurement value of the pH sensor is read, and the buffer injection action is limited according to the preset pH safety range to avoid feedforward compensation causing the water pH to exceed the allowable range.
[0079] Step S5: Perform closed-loop parameter calibration. After the feedforward control time window ends, read the actual dissolved oxygen measurement value collected by the dissolved oxygen sensor, and estimate the equivalent biochemical oxygen demand corresponding to this breeding cycle by combining it with the actual oxygenation output of the micro-nano bubble generator and the mechanical turbulent aerator during this feedforward control period.
[0080] The equivalent biochemical oxygen demand is compared with the peak increment of the predicted oxygen demand for this cycle obtained in step S2 to obtain the oxygen demand prediction residual. Using the oxygen demand prediction residual as the observation input, the feature calibration coefficients in step S1 and the feature scale matching coefficients in step S2 are iteratively updated through a Kalman filter, and the updated calibration parameters are used for multidimensional feature extraction and feedforward load prediction in the next reproductive cycle.
[0081] Reference Figure 3 The specific implementation method for step S1, which extracts multidimensional features, is as follows.
[0082] The multi-dimensional feature extraction module establishes a data communication link with the sensing end to receive video stream data of the aquaculture water body acquired by the multi-view underwater camera array. In this array, the optical axes of at least two underwater cameras are arranged orthogonally. For each video stream acquired from different perspectives, the multi-dimensional feature extraction module performs time alignment according to the video frame timestamps and performs optical flow feature extraction and image grayscale feature extraction on the video frames from each perspective.
[0083] The same feature calculation method is used for each viewpoint. The following uses a video frame from any viewpoint as an example to illustrate the calculation process of single-viewpoint features. After the calculation of similar features for each viewpoint is completed, the similar feature parameters at the same time are weighted and fused to form the output feature parameters corresponding to the current time. The fusion weights for each viewpoint are predetermined during the system initialization and calibration phase based on the effective observation area of the corresponding camera, the stability of effective optical flow feature points, and the occlusion situation, and remain unchanged within the same breeding cycle to reduce the impact of occlusion or local water turbidity on the feature extraction results of a single viewpoint.
[0084] Step S101: Extract optical flow motion features and calculate the group motion convergence index.
[0085] The multidimensional feature extraction module uses an optical flow algorithm to process temporally consecutive adjacent video frames. This embodiment employs the Lucas-Kanade optical flow algorithm to determine the displacement of effective feature points based on the spatiotemporal changes in image brightness within adjacent video frames, and obtains the optical flow motion vector of each effective feature point in the image coordinate system. The same optical flow calculation method is used for each viewpoint.
[0086] The detection of optical flow feature points and the calculation of image gradients in the Lucas-Kanade optical flow algorithm can be achieved using existing computer vision processing methods.
[0087] After obtaining the optical flow motion vector field corresponding to the current video frame, the multi-dimensional feature extraction module counts the total number of effective optical flow feature points and calculates the average value of the sum of squares of the velocity components of each effective optical flow feature point to obtain the global average motion intensity. Global average motion intensity characterizes the overall activity level of the aquaculture population from the current perspective; this parameter increases accordingly as the population's swimming speed or activity frequency increases. The formula for calculating global average kinetic energy is:
[0088] ;
[0089] In the formula, The global average motion intensity; This represents the total number of valid optical flow feature points within the current video frame. For the first The velocity components of each feature point in the horizontal direction of the image coordinate system; For the first The velocity components of each feature point in the vertical direction of the image coordinate system.
[0090] The multidimensional feature extraction module further calculates the spatial divergence of the optical flow motion vector field and statistically analyzes the dispersion of the divergence values corresponding to each effective spatial location to obtain the optical flow divergence variance. The local divergence and convergence states of the optical flow field reflect the divergence and convergence trends of target motion in the image, respectively. The optical flow divergence variance characterizes the degree of difference in these divergence or convergence states between different spatial locations. An increase in the optical flow divergence variance indicates a corresponding increase in the spatial divergence and convergence differences of group motion within the current field of view. The formula for calculating the divergence covariance is:
[0091] ;
[0092] In the formula, The variance of optical flow divergence; Variance calculation operator; This represents the optical flow vector field. This is the function for calculating the spatial divergence of the optical flow vector field.
[0093] The multidimensional feature extraction module linearly combines the global average motion intensity with the variance of optical flow divergence to obtain the population movement convergence index. The population movement convergence index is used to comprehensively characterize the activity level and spatial convergence changes of aquaculture populations. The formula for calculating this index is:
[0094] ;
[0095] In the formula, The aggregation and dispersion index of group movements; and These are the characteristic calibration coefficients corresponding to the global average motion intensity and the variance of optical flow divergence, respectively; The global average motion intensity; This represents the variance of optical flow divergence. The characteristic calibration coefficients can be calibrated during the system initialization phase based on historical samples corresponding to the target aquaculture pond, camera installation location, and camera imaging parameters, and determined through regression fitting.
[0096] The multidimensional feature extraction module also considers the total number of effective optical flow feature points within the current video frame. With the total number of pixels in the image Calculate the effective optical flow feature point distribution density parameters of the current video frame. ,Right now The effective optical flow feature point distribution density parameter is used to characterize the proportion of effective image information available for optical flow calculation in the current video frame, and serves as an input parameter for subsequently determining the reliability of optical flow feature extraction. For a multi-view underwater camera array, after obtaining the corresponding effective optical flow feature point distribution density parameter for each viewpoint, the effective optical flow feature point distribution density parameter for the current moment is formed according to the aforementioned viewpoint fusion weights.
[0097] Step S102: Extract the image grayscale information entropy and obtain the characterization value of reproductive material release intensity.
[0098] During the release of reproductive materials, egg cells, sperm, and accompanying suspended matter enter the aquaculture water, altering the scattering and transmission states of the water between the camera and the target, and potentially causing some optical flow feature points to become unstable for detection. To obtain water optical change information that complements the optical flow motion features, the multi-dimensional feature extraction module simultaneously performs grayscale statistical processing on the original video frames from each perspective.
[0099] The multi-dimensional feature extraction module statistically analyzes the distribution probability of pixels at each gray level in the current video frame and calculates the image gray-level information entropy based on the gray-level histogram. Under conditions where camera exposure parameters and the observed background are basically stable, changes in the content of suspended matter and reproductive substances in the water will cause corresponding changes in the image gray-level distribution and identifiable details. Therefore, this embodiment does not directly use the information entropy value at a single moment as the reproductive substance content, but instead uses the change in image gray-level information entropy over time to characterize the changes in the optical state of the water body caused by the release of reproductive substances. The formula for calculating the global information entropy of the image is:
[0100] ;
[0101] In the formula, The image grayscale information entropy of the current video frame; The pixel grayscale value of the image; grayscale The probability distribution of occurrence in the entire current video frame; It is a logarithmic function with base 2.
[0102] The multidimensional feature extraction module generates an image grayscale entropy sequence in chronological order. To reduce the impact of local illumination fluctuations, short-term occlusion by a single organism, and random noise in video frames on the calculation results of temporal changes, a pre-defined moving average window is first used to smooth the image grayscale entropy sequence, and then the first-order time derivative is calculated on the smoothed entropy sequence. The length of the moving average window can be preset according to the camera sampling frame rate and the duration of the short-term image fluctuations that need to be filtered out.
[0103] The multidimensional feature extraction module takes the absolute value of the first-order time change as the characterization value of the reproductive material release intensity. The reproductive material release intensity characterization value is used to represent the rate of change in the optical state of the water body during the release of reproductive material. It is a characteristic parameter obtained based on video images and is not a direct measurement result of the actual mass or volume release rate of reproductive material. The formula for calculating the equivalent reproductive material release rate is:
[0104] ;
[0105] In the formula, This is a characterization value for the intensity of reproductive material release; This represents the grayscale entropy of a smoothed image relative to the time variable. The change in quantity.
[0106] For a multi-view underwater camera array, the corresponding image grayscale information entropy and reproductive material release intensity characterization value are calculated for each view, and the fusion result at the current moment is obtained according to the aforementioned view fusion weight.
[0107] After completing steps S101 and S102, the multidimensional feature extraction module will use the population movement convergence and divergence index obtained at the current moment. Effective optical flow feature point distribution density parameters and the characterization value of reproductive material release intensity Write the data to the shared storage area of the computing control terminal for the feedforward load forecasting module to read and execute subsequent load forecasting processing.
[0108] See attached document Figure 4 The specific implementation method for step S2, predicting the feedforward load, is as follows.
[0109] The feedforward load forecasting module reads the population movement convergence / disconvergence index from the shared storage area of the calculation control terminal. Effective optical flow feature point distribution density parameters and the characterization value of reproductive material release intensity Based on the aforementioned basic characteristic parameters, reproductive event determination and load prediction are performed.
[0110] The feedforward load prediction module pre-stores the steady-state baseline corresponding to the population movement convergence index. The steady-state baseline is determined based on the historical population movement convergence index collected when the aquaculture system is in a non-breeding state, and is used to characterize the movement level of the target aquaculture population under normal aquaculture conditions.
[0111] During operation, the feedforward load prediction module continuously compares the real-time population aggregation index with the steady-state baseline. When the population aggregation index exceeds the steady-state baseline and the duration reaches a preset time window, the release of reproductive material is determined to have started, and the corresponding system time is recorded as the start time. If the duration exceeding the steady-state baseline does not reach the time determination window, the reproduction event determination is not triggered, and the duration statistics are repeated.
[0112] The time determination window can be pre-calibrated based on the duration characteristics of the group movement behavior during the breeding season of the target aquatic species through historical breeding samples, so as to reduce false triggering caused by short-term stress swimming, local chasing or other transient movement changes.
[0113] Step S201: Determine the dynamic feature validity weights.
[0114] In the initial stage of reproductive material release, the optical state of the aquaculture water changes little, and the number of effective optical flow feature points is usually high in video frames. However, as the number of eggs, sperm, and accompanying suspended matter increases in the water, image scattering and occlusion levels change accordingly, and the number of effective optical flow feature points may decrease. To adjust the contribution ratio of different features in load calculation based on the availability of optical flow features in the current image, the feedforward load prediction module uses the effective optical flow feature point distribution density parameter... Determine the effectiveness weight of dynamic features .
[0115] This embodiment uses the Sigmoid function to establish the effective optical flow feature point distribution density parameters. With dynamic feature effectiveness weight The continuous mapping relationship between them allows the weights to be continuously adjusted as the feature point density changes, reducing the output abrupt changes that occur when switching with a fixed threshold. The specific calculation formula is as follows:
[0116] ;
[0117] In the formula, For dynamic feature validity weights; The effective optical flow feature point distribution density parameter corresponding to the current moment; The critical threshold for the effective optical flow feature point density; This refers to the weighted transition slope coefficient; This represents the natural exponential function.
[0118] in, The calibration can be pre-defined based on the average density of effective optical flow feature points of the target camera under normal imaging conditions and the attenuation of feature points under conditions of water turbidity and obstruction. In one embodiment, It can be set to 30% to 50% of the average density of effective optical flow feature points in the unobstructed state. Used for adjustment exist Nearby The rate of change can be calibrated based on the correspondence between the change in the density of effective feature points and the stability of features in historical video samples.
[0119] when Higher than hour, Increasing this value gives the group movement convergence / dispersion index a higher weight in subsequent fusion calculations; when When it decreases, The corresponding decrease leads to a corresponding increase in the contribution ratio of the reproductive material release intensity characterization value in the fusion calculation.
[0120] Step S202: Calculate the biochemical composite load index and obtain load prediction parameters.
[0121] The feedforward load forecasting module uses the dynamic characteristic validity weights obtained in step S201. The aggregation and dispersion index of group movement Characterization value of reproductive substance release intensity Weighted fusion was performed to obtain the biochemical composite load index. The fusion calculation formula is as follows:
[0122] ;
[0123] In the formula, The biochemical composite load index; The aggregation and dispersion index of group movements; This is a characterization value for the intensity of reproductive material release; The feature scale matching coefficient; This represents the dynamic feature validity weight.
[0124] Feature Scale Matching Coefficient This is used to adjust the numerical scale difference between group motion features and image grayscale change features, so that the two types of features can be fused according to the dynamic feature effectiveness weight. The statistical distribution and variation range of the two types of characteristic parameters can be determined by calibrating the historical breeding video samples collected under the target breeding environment.
[0125] Biochemical composite load index This represents the overall load intensity corresponding to the reproductive substance release behavior at the current moment. Since the release of reproductive substances is continuous, the feedforward load prediction module performs time integration on the biochemical composite load index from the start time t0 to characterize the cumulative load change since the start of the reproductive substance release process.
[0126] The feedforward load prediction module uses pre-calibrated integral gain parameters to convert the above integral results into predicted increases in oxygen consumption and estimated values for the equivalent average density of fertilized eggs, respectively. The specific integral mapping formula is as follows:
[0127] ;
[0128] ;
[0129] In the formula, For automatic startup time From the current computation time The cumulative predicted oxygen consumption increment. The feedforward load prediction module continuously updates during the feedforward control phase. And record the largest number that has appeared in the current breeding cycle. At the end of the feedforward control time window, the maximum value recorded within the cycle is determined as the predicted peak oxygen consumption increment for this cycle.
[0130] This is an estimated value of the equivalent average density of fertilized eggs, used to characterize the overall density level of fertilized eggs within the target aquaculture water body, and does not represent the point-by-point three-dimensional density distribution corresponding to each spatial coordinate position.
[0131] This is the current computation time, and it continues to increase as the system runs during the feedforward control phase. For the time variable of the infinitesimal integral, The integral mapping coefficient for oxygen consumption. The integral mapping coefficient for fertilized egg density.
[0132] It can be calibrated based on the statistical relationship between the integral results of the biochemical compound load index during the historical reproductive cycle of the target aquatic species and the corresponding actual oxygen consumption; It can be calibrated based on the relationship between the integral results of the biochemical compound load index in the historical reproductive cycle and the corresponding oviposition or fertilized egg statistics.
[0133] In calibration At that time, the effective water volume of the target aquaculture pond and the effective observation range of the multi-view underwater camera array were included as fixed engineering conditions in the calibration sample, so that... This corresponds to the equivalent average density of fertilized eggs under these aquaculture conditions. For changes in pond volume, camera arrangement, or target aquatic species, historical samples under the corresponding conditions can be collected again. Perform calibration.
[0134] Step S203: Calculate the biochemical oxygen demand response delay time based on the Arrhenius relation.
[0135] After reproductive material is released into the aquaculture water, the organic matter and nitrogenous substances within it undergo biochemical processes such as decomposition, transformation, and nitrification, creating an additional oxygen-consuming load. There is a time delay between the start of the reproductive material release and the point at which dissolved oxygen changes stably reflect its oxygen-consuming effect, and the reaction rate is affected by water temperature.
[0136] The feedforward load prediction module reads the real-time water temperature collected by the temperature sensor. Based on the Arrhenius relation, the biochemical oxygen demand response delay time corresponding to this reproductive event was calculated. The biochemical oxygen demand response delay time is used to characterize the predicted time span from the initiation of the reproductive substance release process to the entry of its main oxygen demand effect into a stage that can be stably characterized by dissolved oxygen sensors.
[0137] The system is based on the biochemical oxygen demand response delay time. Determine the feedforward control time window corresponding to this reproduction event, and start from the start time. Incremental control for this breeding event is executed within the feedforward control time window. The specific derivation formula is as follows:
[0138] ;
[0139] In the formula, The time delay is used to determine the biochemical oxygen demand response time and is also used as a time parameter to determine the feedforward control time window. It is a pre-exponential factor for biochemical reactions; It represents the apparent activation energy of a biochemical reaction; It is the ideal gas constant; Represents the natural exponential function; The real-time water temperature is collected by a temperature sensor, and its unit of measurement is Kelvin.
[0140] For different target aquatic species, the composition of microorganisms in the aquaculture water, and operating conditions, the biochemical oxygen demand response can be determined based on the corresponding aquaculture environment. and The calibration results are pre-stored in the parameter storage area of the computing control terminal.
[0141] After completing the above processing, the feedforward load prediction module will update the predicted oxygen consumption increment over time. The data is written in real-time to the shared storage area of the computing control terminal for the oxygenation and pH compensation co-control module to calculate the current oxygenation power requirement; the equivalent average density of fertilized eggs is estimated. Provided to the shear safety boundary calculation module to determine the upper limit of dynamic shear stress safety; and includes the biochemical oxygen demand response delay time. The corresponding feedforward control time window is provided for subsequent control and closed-loop parameter calibration processes. After the feedforward control time window ends, the peak increment of predicted oxygen consumption recorded in this cycle is saved for the closed-loop parameter calibration module to calculate the oxygen consumption prediction residual.
[0142] Reference Figure 5 The specific implementation method for establishing the physical boundary in step S3 is as follows.
[0143] The shear safety boundary calculation module reads the estimated value of the equivalent average density of fertilized eggs output by the feedforward load prediction module via the industrial control bus. Based on the estimated equivalent average density of fertilized eggs and the hardening process of the egg membrane after fertilization, the shear stress limit conditions that the mechanical turbulent aerator should meet within the feedforward control time window are determined.
[0144] This embodiment focuses on fertilized eggs whose membranes undergo a gradual hydration and hardening process after expulsion. When fertilized eggs are first expelled, their membranes have not yet reached a stable hardened state and are relatively sensitive to the hydraulic shear stress generated by the aeration equipment. As the hydration and hardening process progresses, their shear stress tolerance gradually increases. Therefore, after the release of reproductive material is initiated, the hydraulic shear stress generated by the mechanical turbulent aerator is limited according to the current hardening stage of the fertilized egg.
[0145] Step S301: Determine the initial shear stress tolerance threshold.
[0146] The shear safety boundary calculation module estimates the value based on the equivalent average density of fertilized eggs. Determine the initial shear stress tolerance threshold at the beginning of ovulation. The estimated equivalent average density of fertilized eggs is used to characterize the overall density level of fertilized eggs within the target aquaculture water body. Under the condition that the effective water volume of the aquaculture pond and the effective range of the aeration equipment are determined, An increase indicates a corresponding increase in the number of fertilized eggs per unit water volume, and a corresponding increase in the probability of the fertilized eggs coming into contact or colliding with each other under the influence of water flow disturbance. Based on this, this embodiment utilizes... The baseline shear stress tolerance value under low-density fertilized egg conditions is modified so that the initial allowable shear stress decreases as the equivalent average density of the fertilized egg increases. The formula for calculating the initial shear stress tolerance threshold is configured as follows:
[0147] ;
[0148] In the formula, This is the initial shear stress tolerance threshold; The shear stress tolerance value of the fertilized egg under preset low-density conditions and the initial hydration stage; This is the density correction factor; The estimated value of the equivalent average density of fertilized eggs.
[0149] The hydraulic shear tolerance of the target aquatic species can be determined by conducting a hydraulic shear tolerance test on fertilized eggs in the initial hydration stage after discharge. During the test, the shear stress of the water flow is gradually changed under the condition that the density of the fertilized eggs is within a preset low density range, and the allowable basic shear stress range is determined according to the integrity of the egg membrane or the normal development of the embryo.
[0150] Density correction factor This method, used to reflect the degree of correction of allowable shear stress by changes in the equivalent average density of fertilized eggs, can be calibrated by setting different fertilized egg densities and recording the corresponding safe shear stress ranges, under the conditions of a defined effective water volume in the aquaculture pond, the installation location of the mechanical turbulent aerator, and the range of its main water flow disturbances. The calibrated values are... and The parameter storage area is pre-stored in the computing control terminal.
[0151] Considering Calculated using density correction relation, when When the calculated results exceed the commonly used range of historical calibration samples, they may be lower than the range of shear stress with practical physical significance. Therefore, the shear safety boundary calculation module sets a preset lower limit for the shear stress threshold. . The lower limit of the safe tolerance range, determined based on the shear tolerance test results of the target aquatic species during the initial hydration stage, was selected and verified through testing. When the calculated... Below At that time, with This serves as the initial value for subsequent calculations of the safety upper limit of dynamic shear stress, thereby avoiding negative or physically meaningless abnormal results in the case of linear extrapolation of the density correction relationship.
[0152] Step S302: Generate a dynamic shear stress safety upper limit sequence.
[0153] During the hydration and hardening process following fertilization, the mechanical stability of the egg membrane increases over time. Correspondingly, the upper limit of the shear stress that a mechanical turbulent aerator can apply to the aquaculture water can be increased from the initial stage of fertilization. Gradually increase, and approach the upper limit of conventional shear stress corresponding to the completion of hardening of the fertilized egg membrane. .
[0154] The shear safety boundary calculation module uses the start time of the propagation material release process recorded in step S2. As a time reference, calculate the current computation time. Relative to startup time Time elapsed Based on the exponential asymptotic relationship between the shear stress tolerance of the fertilized egg and the hardening time, the upper limit of the dynamic shear stress safety at the current moment is calculated. The specific mathematical formula is as follows:
[0155] ;
[0156] In the formula, For the time elapsed The corresponding upper limit of dynamic shear stress safety; The current calculation time With startup time The time difference between them, i.e. ; This refers to the upper limit of shear stress that can be applied under normal seedling conditions after the egg membrane of the fertilized egg has reached a stable and hardened state. The rate constant for egg membrane hardening; It is a natural constant; This is the initial shear stress tolerance threshold.
[0157] The tolerance of fertilized eggs after hydration and hardening can be determined by conducting tolerance tests under different water flow shear stress conditions. The shear stress tolerance level can be measured at different time points after fertilization, and the above-mentioned relationship can be fitted based on the obtained time-to-tolerance level data. For different aquatic species or different water temperature conditions, corresponding calibrations can be performed, and the corresponding values can be... and The parameter storage area is located in the computing control terminal.
[0158] Within the feedforward control time window, the shear safety boundary calculation module repeatedly calculates according to the control update cycle preset by the calculation control terminal. According to each time point This constitutes a dynamic shear stress safety upper limit sequence. The control update cycle corresponds to the power command update cycle of the subsequent mechanical turbulence aerator, enabling the aeration and pH compensation coordinated control module to read the corresponding shear stress safety upper limit at the current moment each time a power command for the mechanical turbulence aerator is generated.
[0159] The shear safety boundary calculation module writes the current dynamic shear stress safety upper limit and its time-updated data into the shared storage area of the calculation control terminal for the aeration and pH compensation co-control module to read. Based on the dynamic shear stress safety upper limit, the aeration and pH compensation co-control module determines the maximum allowable output power of the mechanical turbulence aerator, thus limiting the operating power of the mechanical turbulence aerator within the feedforward control time window to the shear stress tolerance conditions of the fertilized egg.
[0160] Reference Figure 6 The specific implementation of step S4, decoupling and collaborative control, is as follows.
[0161] In this embodiment, the oxygenation and pH compensation coordinated control module performs incremental control for the current reproductive event within the feedforward control time window. The oxygenation and pH compensation coordinated control module reads the real-time updated predicted oxygen consumption increment from the feedforward load prediction module via the industrial control bus. It also reads the current dynamic shear stress safety upper limit provided by the shear safety boundary calculation module to determine the control commands for the micro-nano bubble generator, mechanical turbulent aerator, and variable frequency buffer metering pump.
[0162] The control commands generated for this breeding event are incremental control commands added on top of the basic water quality maintenance operation conditions of the aquaculture pond. After the feedforward control time window ends, the incremental control for this breeding event is stopped, and the execution end resumes operation according to the original basic water quality maintenance conditions.
[0163] The oxygenation and pH compensation coordinated control module controls oxygen consumption based on the predicted increase. The rate of change of the current additional oxygen load over time is determined, and combined with the effective water volume of the aquaculture pond and the oxygenation efficiency parameters of the aeration equipment, the total oxygenation power requirement at the current moment is determined. The effective water volume of the aquaculture pond is determined during system initialization based on the actual operating water level of the pond. The oxygenation efficiency parameter of the aeration equipment can be calibrated by running the corresponding aeration equipment under known water volume and water quality conditions, measuring the dissolved oxygen increment caused by a unit input power, and storing the calibration results in the calculation and control terminal.
[0164] Step S401: Allocate the input power of the two types of aeration equipment under the dynamic shear stress safety upper limit constraint.
[0165] Micro-nano bubble generators supply oxygen to aquaculture water through gas-liquid mixing and the release of microbubbles, generating relatively low macroscopic hydraulic shear stress during operation. Mechanical turbulent flow aerators, on the other hand, create water circulation and gas-liquid exchange through mechanical drive, resulting in relatively high local hydraulic shear stress during oxygenation. Based on these operational characteristics of the two types of aeration equipment, micro-nano bubble generators are prioritized for oxygenation during the feedforward control time window. When their output capacity is insufficient, mechanical turbulent flow aerators supplement the remaining oxygenation demand. Simultaneously, the maximum allowable output of the mechanical turbulent flow aerators is limited by a dynamic shear stress safety upper limit.
[0166] The oxygenation and pH compensation coordinated control module first determines the total oxygenation power requirement. The feedforward input power of the micro / nano bubble generator is determined by its rated upper limit power. The specific allocation formula is configured as follows:
[0167] ;
[0168] In the formula, This is the feedforward input power for the micro / nano bubble generator; This refers to the total oxygenation power requirement parameter; It is a minimum value function; This is the rated upper limit power of the micro / nano bubble generator. Not greater than At that time, the micro-nano bubble generator will handle all the incremental oxygenation needs corresponding to this reproductive event; when Greater than At that time, the micro-nano bubble generator operates at its maximum allowable power limit, and the remaining oxygenation demand is handled by the mechanical turbulent aerator.
[0169] For mechanical turbulent aerators, the oxygenation and pH compensation coordinated control module is based on... and The remaining power demand is used to calculate the supplementary input power, and the supplementary input power is limited by the current dynamic shear stress safety limit. To ensure that the power command of the mechanical turbulent aerator corresponds to the hydraulic shear stress that the fertilized egg can withstand, the calculation control terminal pre-stores the inverse mapping relationship between the shear stress and power of the mechanical turbulent aerator. The specific allocation formula introduces a lower limit constraint and is configured as follows:
[0170] ;
[0171] In the formula, Provides supplementary input power for mechanical turbulent aerators; It is a minimum value function; It is a function for maximizing the value; This refers to the total oxygenation power requirement parameter; This is the feedforward input power for the micro / nano bubble generator; This represents the upper limit of the dynamic shear stress safety at the current moment. This is a shear stress and power inverse mapping function used to determine the maximum allowable input power for a mechanical turbulence aerator based on the upper limit of allowable shear stress.
[0172] The inverse mapping relationship between shear stress and power can be obtained through equipment calibration during the system installation and commissioning phase. Specifically, after determining the installation location and operating water level of the target aquaculture pond, the mechanical turbulent flow aerator is operated at different power levels or frequency converters to measure the water flow shear stress in its main operating area. A correspondence is then established between the representative shear stress value corresponding to each operating state and the equipment input power. The representative shear stress value can be determined based on the larger shear stress measured in the main area where the fertilized eggs are located, so that subsequent power limits cover the high-shear areas within the corresponding areas. After forming a fitting curve or lookup table based on the calibration data, the corresponding maximum allowable input power can be obtained by reversing the given dynamic shear stress safety upper limit.
[0173] When the total oxygenation power demand exceeds the combined oxygenation capacity of the micro-nano bubble generator and the mechanical turbulence aerator (limited by shear stress), both types of aeration equipment operate at the power allowed under the current constraints, and the output power of the mechanical turbulence aerator is no longer increased. This ensures that the incremental oxygenation control corresponding to this reproductive event conforms to the shear stress tolerance condition of the fertilized egg.
[0174] Step S402: Perform synchronous feedforward pH compensation based on the operating status of the aeration equipment.
[0175] When micro-nano bubble generators and mechanical turbulent aerators are running, they alter the mass transfer state at the gas-liquid interface while transferring oxygen to the water, causing some dissolved carbon dioxide in the aquaculture water to escape into the gas phase. This reduction in dissolved carbon dioxide alters the carbonate balance in the water and may cause an increase in pH. To mitigate the collateral impact of the oxygenation process on the pH of the aquaculture water, an oxygenation and pH compensation co-control module estimates the equivalent dissolved carbon dioxide escape rate caused by the operating power of the two types of aeration equipment. The evaluation calculation formula is configured as follows:
[0176] ;
[0177] In the formula, This represents the equivalent rate of dissolved carbon dioxide dissipation at the current moment. This is the dispersion coefficient corresponding to the mechanical turbulent aerator; This represents the dissipation coefficient corresponding to the micro / nano bubble generator; Provides supplementary input power for mechanical turbulent aerators; This is the feedforward input power for the micro / nano bubble generator.
[0178] and These can be determined through equipment calibration. During the calibration process, under the condition that the water volume, initial water quality, and environmental conditions are within the preset range, the corresponding aeration equipment is operated individually at different constant power levels. The change in dissolved carbon dioxide concentration in the water over time is continuously measured. The corresponding emission coefficient is determined based on the change in carbon dioxide emission under each operating power, and the calibration results are written into the parameter storage area of the calculation control terminal.
[0179] The oxygenation and pH compensation co-control module determines the pH buffer injection requirement based on the equivalent escape rate of dissolved carbon dioxide and generates feedforward control commands for the variable frequency buffer metering pump. In this embodiment, to address the dissolved carbon dioxide escape and corresponding pH rise caused by aeration, a pH buffer is pre-configured in the storage tank to suppress this pH rise. The volumetric rate formula for the metering pump to inject the acid-base buffer into the water is configured as follows:
[0180] ;
[0181] In the formula, The buffer injection volume rate of the variable frequency buffer metering pump; This represents the equivalent concentration of the pH buffer solution in the storage tank. This represents the equivalent rate of dissolved carbon dioxide dissipation at the current moment.
[0182] The equivalent escape rate of dissolved carbon dioxide is characterized by equivalent amount per unit time corresponding to the pH buffer compensation requirement, and the equivalent concentration of the pH buffer is characterized by equivalent amount per unit volume. Both use the same equivalent metric standard, so the ratio corresponds to the buffer injection volume rate.
[0183] In one embodiment, the pH buffer solution may be a weakly acidic buffer system adapted to the target culture medium and the water quality conditions. The pH buffer concentration is determined based on the actual prepared concentration of the buffer solution and is pre-stored as a known parameter in the calculation control terminal. The pH buffer solution referred to in this article is used to compensate for the pH rise caused by aeration and does not mean that independently switchable acid and alkali solutions are simultaneously installed in the same storage tank.
[0184] During feedforward injection by the variable frequency buffer metering pump, the computing control terminal synchronously reads the real-time measurement value of the pH sensor and uses the preset allowable pH range as a safety limit for buffer injection. The allowable pH range can be preset according to the suitable water quality range for the seedling stage of the target aquatic species. When the real-time pH has reached or fallen below the lower limit of the allowable range, the buffer injection in the direction of decreasing pH is stopped; when the real-time pH meets the preset compensation condition, the variable frequency buffer metering pump runs according to the current buffer injection command to limit the reverse pH shift caused by feedforward compensation.
[0185] The oxygenation and pH compensation coordinated control module is generated according to the same control update cycle. , and The corresponding control commands are then converted into the device control command format used by the industrial control bus and sent to the micro-nano bubble generator, mechanical turbulence aerator, and variable frequency buffer metering pump, respectively. This ensures that oxygenation control and pH buffer compensation for this breeding event are executed within the same control cycle, and that appropriate buffer compensation is applied in advance before significant pH changes are caused by dissolved carbon dioxide escape.
[0186] After the feedforward control time window ends, the incremental oxygenation command and feedforward pH compensation command generated in this step for this reproductive event are stopped. The actual output power of the two types of aeration equipment during the feedforward control period is recorded over time in the calculation control terminal, so that the closed-loop parameter calibration module can estimate the equivalent biochemical oxygen demand of this reproductive cycle in subsequent steps.
[0187] Reference Figure 7 The specific implementation method for step S5, which involves performing closed-loop parameter calibration, is as follows.
[0188] The computational control unit monitors the feedforward control time window of the current reproductive cycle based on the system's runtime. This is at the self-startup time. When the running time reaches the feedforward control time window determined in step S203, the incremental oxygenation and feedforward pH compensation for this reproductive event are stopped, and the closed-loop parameter calibration process begins. At this point, based on the calculation results of the aforementioned biochemical oxygen consumption response delay time, the main oxygen consumption effect caused by the release of reproductive substances has entered a stage that can be measured by the dissolved oxygen sensor.
[0189] The closed-loop parameter calibration module reads the actual water quality measurement results of the current breeding cycle and the actual equipment operation data recorded during the feedforward control period. It compares the actual oxygen consumption results with the feedforward load prediction results and updates the calibration parameters based on the deviation between the two. The parameter update is used for feature extraction and load prediction in the next breeding cycle. After the feedforward control time window ends, the actuator continues to operate under the basic water quality maintenance conditions.
[0190] Step S501: Estimate the equivalent biochemical oxygen demand and calculate the oxygen demand prediction residual.
[0191] The closed-loop parameter calibration module reads the actual dissolved oxygen measurement value collected by the dissolved oxygen sensor at the end of the current feedforward control time window. It also reads the actual output power of the micro-nano bubble generator and the mechanical turbulence aerator recorded by the calculation control terminal during the execution of step S4.
[0192] When determining the oxygen consumption for this period, in addition to considering the changes in dissolved oxygen concentration in the water within the feedforward control time window, the equivalent oxygenation actually provided to the aquaculture water by the two types of aeration equipment during this period is also taken into account in order to estimate the equivalent biochemical oxygen consumption during this period.
[0193] in, The dissolved oxygen data is determined based on data collected before the breeding event was triggered, when the aquaculture pond was under stable basic operating conditions. In practice, a preset steady-state data collection interval can be selected before the breeding event is triggered, and the representative dissolved oxygen value within this interval can be used as the baseline dissolved oxygen for the current breeding cycle. The same steady-state data collection rules are used for each breeding cycle to reduce the impact of differences in initial dissolved oxygen states between different cycles on the oxygen consumption prediction residual.
[0194] It should be noted that the equivalent biochemical oxygen demand is a characterization result of dissolved oxygen change and artificial oxygenation within the feedforward control time window. It is used to compare with the predicted oxygen demand under the same conditions and is not limited to the absolute oxygen demand obtained by measuring various biochemical oxygen demand sources in aquaculture water.
[0195] The closed-loop parameter calibration module compares the equivalent biochemical oxygen demand (BOD) with the predicted peak increment of OD recorded by the feedforward load prediction module for this reproductive cycle to obtain the first... The predicted residual for oxygen consumption over a reproductive cycle. The specific calculation formula is configured as follows:
[0196] ;
[0197] In the formula, For the first Oxygen consumption prediction residuals for each reproductive cycle; The baseline dissolved oxygen level was determined based on steady-state operating data prior to the triggering of this reproductive cycle. This is the actual dissolved oxygen measurement value at the end of the feedforward control time window of this cycle.
[0198] The input electrical energy oxygenation coefficient is used to convert the actual electrical energy input of the two types of aeration equipment into the equivalent dissolved oxygen increment under the effective water volume conditions of the target aquaculture pond. During the system installation and calibration phase, the aeration equipment can be operated under the effective water volume and preset water quality conditions of the target aquaculture pond. The input electrical energy and corresponding dissolved oxygen increment of the equipment can be recorded and stored in the computing and control terminal.
[0199] For the system in The actual total oxygenation output power at any given time is specifically the sum of the input power actually issued and executed by the micro-nano bubble generator and the mechanical turbulent aerator after the equipment's rated power limit and dynamic shear stress safety upper limit constraint, and is the total oxygenation power requirement before the amplitude limiting allocation in step S401. .
[0200] and The first The start time of each reproduction cycle and the end time of the feedforward control time window. For the feedforward load forecasting module in the first The predicted peak oxygen consumption increment recorded within each reproductive cycle.
[0201] Oxygen consumption prediction residual This reflects the deviation between the predicted oxygen consumption and the equivalent oxygen consumption estimated based on actual dissolved oxygen changes and actual oxygenation output under the current basic operating conditions, and serves as the observation input for subsequent calibration parameter updates.
[0202] Step S502: Update the calibration parameters based on the Kalman filter.
[0203] During continuous aquaculture operations, the presence of deposits on the underwater camera lens surface, changes in camera imaging status, changes in the optical properties of the water background, and differences in the body size and activity characteristics of different breeding batches of aquaculture groups may all cause deviations between the parameters pre-calibrated in steps S1 and S2 and the current actual aquaculture conditions.
[0204] The calibration parameters to be calibrated are the characteristic calibration coefficients in the population movement convergence index and the characteristic scale matching coefficients in the biochemical composite load index.
[0205] During the specific calibration, multiple historical breeding cycles were selected, and the time integral results of the biochemical composite load index for each breeding cycle were obtained. The corresponding actual oxygen consumption was obtained by determining the oxygen consumption results based on dissolved oxygen changes and actual oxygenation. At the same time, the corresponding egg production or fertilized egg statistics were obtained. Based on the statistical relationship between the time integral results and the corresponding actual quantities, the integral mapping coefficients of oxygen consumption and fertilized egg density were determined, and the determined mapping coefficients were stored in the calculation control terminal.
[0206] The closed-loop parameter calibration module uses the oxygen consumption prediction residuals obtained from each breeding cycle as observation information to calibrate the characteristic calibration coefficients in the population movement convergence and divergence index. , and the characteristic scale matching coefficient in the biochemical composite load index Perform cross-cycle recursive updates.
[0207] The closed-loop parameter calibration module is constructed from , and The state vector The state prior estimate for the current cycle is obtained based on the parameter estimation results of the previous breeding cycle and a preset state transition matrix. In embodiments where parameter changes are considered to be slow changes between adjacent breeding cycles, the state transition matrix can be preset according to the assumption that the parameters remain constant or drift slowly.
[0208] Here, the state vector subscript Used to indicate the first State estimation results for each reproductive cycle; feature calibration coefficients The subscript in the name is used to distinguish the characteristic coefficient and does not represent the reproductive cycle number.
[0209] After each reproductive cycle is completed, the closed-loop parameter calibration module calculates the oxygen consumption prediction residual obtained for that cycle. As an observation input, the observation matrix in the Kalman filter is used to describe... , and The extent to which changes affect the predicted oxygen consumption results.
[0210] The observation matrix can be determined using historical breeding cycle samples during the system initialization and calibration phase. Specifically, while keeping other calculation conditions constant, the parameters to be calibrated are changed within their allowable ranges. , or Following the processing steps S1 and S2, the corresponding predicted oxygen consumption is recalculated, and the observation sensitivity of each parameter is determined based on the changes in the predicted oxygen consumption before and after the parameter changes. The observation sensitivity of each parameter forms the observation matrix for subsequent Kalman recursion and is stored in the computation control terminal.
[0211] Therefore, this embodiment does not directly solve for the three parameters to be calibrated based on the oxygen consumption prediction residual of a single breeding cycle. Instead, it performs recursive estimation by combining the state prior estimate of each parameter, the error covariance, the pre-calibrated observation sensitivity, and the oxygen consumption prediction residual obtained from consecutive breeding cycles.
[0212] The closed-loop parameter calibration module calculates the Kalman gain corresponding to the current breeding cycle based on the state prior estimate, error covariance, and observation matrix, and uses oxygen consumption to predict the residual. The prior state estimate is updated posteriorly. The state update formula is configured as follows:
[0213] ;
[0214] In the formula, This is the updated system state vector; This is the prior state estimation vector derived from the previous cycle; The Kalman gain matrix calculated for the current period; For the first The oxygen consumption prediction residuals for each breeding cycle. The recursive processing of the state prior covariance, observation noise covariance, and posterior covariance in the Kalman filter can be performed according to the conventional matrix recursive method of linear Kalman filtering; the observation relationship of the aquaculture system is determined by the aforementioned parameter observation sensitivity calibration results.
[0215] To limit the impact of a single abnormal measurement on the calibration parameters, the closed-loop parameter calibration module sets allowable ranges for each state parameter after the Kalman update and performs amplitude limiting on update results that exceed the allowable ranges. The specific amplitude limiting formula is configured as follows:
[0216] ;
[0217] In the formula, For the updated state vector, the th Each parameter component; and These are the preset lower and upper limits for this parameter, respectively. It is a function for maximizing the value; It is a minimum value function.
[0218] The allowable range of each parameter can be determined based on the valid value range of the corresponding parameter in historical calibration samples, and verified in conjunction with the stability of the prediction results within this value range and the subsequent executable power range of the oxygenation equipment. The allowable parameter range is written to non-volatile memory during the system initialization phase.
[0219] After the amplitude limiting process is completed, the closed-loop parameter calibration module will update the parameters. and Write the updated parameters to the parameter storage area of the multidimensional feature extraction module. Write the parameters to the parameter storage area of the feedforward load forecasting module to replace the corresponding old parameters and participate in the feature calculation and load forecasting of the next breeding cycle.
[0220] Through the above-mentioned update process of continuous breeding cycles, the prediction parameters can be periodically calibrated based on the actual oxygen consumption measurement results, so as to reduce the prediction deviation caused by changes in camera imaging status, water optical background and aquaculture group characteristics over long-term operation.
[0221] To further illustrate the practical application effects and technical feasibility of the present invention, the following detailed application example of the aquatic group breeding and seedling water body coordinated regulation system of the present invention is presented in conjunction with a specific numerical calculation scenario.
[0222] In a specific application embodiment, the aquatic group breeding and seedling raising water body coordinated regulation system of the present invention is configured in an indoor factory-style recirculating aquaculture pond for pearl grouper with a standard water volume of 50 cubic meters. During the initial steady-state operation phase monitored by the system, the water temperature is maintained at 28 degrees Celsius, the basic dissolved oxygen concentration is stable at 6.5 mg / L, the basic pH value is 8.0, the rated upper limit power of the micro-nano bubble generator is set to 1000 watts, and the maximum physical output power of the mechanical turbulent aerator is 2000 watts.
[0223] During system operation, the video stream processing results from the multi-view underwater camera array showed changes in the swimming behavior of the target aquatic group. The multi-dimensional feature extraction module, through optical flow analysis, found that the global average motion intensity rapidly increased from 0.15 in the normal swimming state to 0.65, and the optical flow divergence variance simultaneously increased to 0.25, clearly indicating that the group exhibited high-frequency chasing and local aggregation behaviors. At the same time, the calculation results based on the grayscale information entropy of the video images showed that the optical state of the water body changed rapidly, and the characterization value of reproductive material release intensity reached 0.45.
[0224] After receiving the aforementioned basic characteristic parameters, the feedforward load prediction module determines that the population movement convergence index has exceeded the preset steady-state baseline of 0.30 for 5 consecutive minutes. Based on this, it confirms that the reproductive and defecation event has officially started. The system uses the effective optical flow feature point distribution density parameter at the current moment to calculate the dynamic feature effectiveness weight and integrates it to generate the biochemical composite load index. By performing an integral operation on the time dimension of this index, the system predicts that this defecation event will subsequently trigger a peak increase in expected oxygen consumption of 2.5 mg / L, and estimates that the equivalent average density of fertilized eggs will reach 800 eggs / L. Combined with the real-time water temperature of 28 degrees Celsius, the system deduces from the Arrhenius response relationship that the delay time from the start of defecation to the manifestation of major biochemical oxygen consumption is 45 minutes. This 45-minute time period is defined by the system as the feedforward control time window.
[0225] Within the aforementioned 45-minute feedforward control time window, the shear safety boundary calculation module, based on the estimated density of 800 fertilized eggs per liter, calculates the lower limit of the shear stress tolerance of the water flow at the initial stage of ovulation as 0.15 Pascals. As time progresses and based on the physical characteristics of the hydration and hardening of the fertilized egg membrane, this safety upper limit smoothly widens to 0.40 Pascals within 45 minutes following an exponential asymptotic law.
[0226] The oxygenation and pH compensation co-control module calculates the current total oxygenation power required as 1500 watts based on the expected peak increase in oxygen consumption. According to the decoupled control logic, the micro-nano bubble generator is prioritized to operate at full load to handle 1000 watts of output power. The remaining 500 watts of power shortfall needs to be supplemented by the mechanical turbulent flow aerator. However, after comparing the current dynamic shear limit of 0.15 Pascals, the system forcibly implements a safety limit on the mechanical turbulent flow aerator, limiting its current allowable actual output power to 300 watts. Based on this, the system issues a command to inject a total of 1300 watts of physical oxygenation power. At the same time, the system evaluates the equivalent dissolved carbon dioxide emission rate driven by the two devices operating at this power and simultaneously issues a feedforward compensation command to the variable frequency buffer metering pump to inject a weakly acidic buffer solution of a specific equivalent concentration into the water at a volume rate of 50 ml per minute, in advance suppressing the upward trend of water pH.
[0227] After the feedforward control time window has run for 45 minutes, the system switches to closed-loop calibration mode. At this time, the dissolved oxygen sensor measures a water concentration of 6.3 mg / L. The system calculates the actual concentration decrease with the equivalent dissolved oxygen injected by the two devices during the feedforward period, restoring the true equivalent biochemical oxygen demand to 2.3 mg / L. The system then compares this true oxygen demand with the previously predicted peak value of 2.5 mg / L to obtain a residual oxygen demand prediction of 0.2 mg / L. The closed-loop parameter calibration module inputs this 0.2 mg / L residual into the Kalman filter to fine-tune and update the population feature scale matching coefficients in the subsequent identification algorithm, so as to ensure the prediction accuracy of the next breeding cycle.
[0228] To verify the comprehensive regulatory effectiveness of the present invention, the system developers designed and implemented specific comparative verification experiments in accordance with the rigorous single variable principle.
[0229] The experimental subjects were selected from the same batch of pearl giant grouper parent fish of the same size and in the breeding period. The experimental environment was set up as two 50-cubic-meter breeding ponds with the same physical size and infrastructure. One pond was set as the experimental group breeding pond for deploying the collaborative control system of this invention, and the other pond was set as the control group breeding pond for deploying the traditional proportional-integral-derivative feedback control system based on the threshold set by the dissolved oxygen sensor.
[0230] The specific experimental steps included equipment initialization and water quality stabilization to ensure that the initial dissolved oxygen, water temperature, and pH benchmarks of the two groups of water bodies were highly consistent. Subsequently, the spontaneous reproductive and defecation behaviors of the two parent groups were induced by artificially controlling the light and water flow. During the 120 minutes of the reproductive event and subsequent biochemical aerobic fermentation, the dissolved oxygen change curves and pH fluctuation extremes of the two groups of water bodies were recorded throughout the process. Fertilized egg samples were collected 12 hours after the end of the experiment for microscopic examination of survival rate and egg membrane integrity.
[0231] The experimental results showed differences. In terms of water quality stability, the experimental group, relying on feedforward load prediction, started oxygenation supplementation in advance before oxygen consumption occurred, so that the maximum drop in dissolved oxygen concentration was only 0.3 mg / L. In contrast, the control group was limited by sensor feedback lag, which caused the highest level of aeration to be triggered only after the large-scale outbreak of biochemical oxygen consumption. This led to a deep drop in dissolved oxygen concentration of 1.5 mg / L and triggered hypoxia and surfacing stress response in the aquaculture population. In terms of pH maintenance, the experimental group's synchronous feedforward buffer compensation mechanism kept the pH of the water within a narrow safe range of 8.0 to 8.1. In contrast, the control group's violent aeration in the lag period caused a large amount of carbon dioxide to escape from the water, and its pH peak rose out of control to 8.5, which also increased the toxicity of free ammonia in the water.
[0232] Regarding core breeding indicators, the experimental group, by strictly implementing dynamic fluid shear force limiting control based on the physical hardening process of fertilized eggs, avoided physical tearing of the newly hatched soft-shelled eggs by high-intensity mechanical water flow. Microscopic examination showed that the fertilized egg integrity rate reached 89%, and the subsequent hatching rate remained above 85%. In contrast, the feedback control system of the control group unconditionally drove the mechanical turbulent aerator to full power when low dissolved oxygen was detected. The resulting huge local shear stress caused the egg membrane of a large number of fertilized eggs in the vulnerable stage to rupture or the internal structure to be damaged. Ultimately, the fertilized egg integrity rate was reduced to 68%, and the final hatching rate was only 65%. The above experimental verification fully demonstrates that the present invention has the superiority in ensuring water quality stability and improving the survival rate of high-value aquatic seedlings.
[0233] Figure 8 The graph shows the relationship between system operating time and dissolved oxygen concentration in the water. The solid line with circular markings represents the trajectory of the system of the present invention, while the dashed line with square markings represents the trajectory of the traditional feedback control system. The horizontal axis represents the system operating time from 0 to 120 minutes, and the vertical axis represents the dissolved oxygen concentration in the water from 4.5 to 7.0 mg / L. After the reproductive and defecation event is triggered at the 20-minute mark, the feedforward aeration of the present invention causes a slight increase in dissolved oxygen, which is then stabilized by closed-loop fine-tuning after 45 minutes. In contrast, the traditional system causes a sharp parabolic drop in dissolved oxygen after the event and triggers delayed aeration at 45 minutes, resulting in severe fluctuations in water quality. This clearly reflects the superiority of feedforward synergistic regulation in eliminating environmental lag effects.
[0234] Figure 9In (a), the horizontal axis represents the progress of the feedforward time window from 0 to 60 minutes, and the vertical axis represents the equipment input drive power from 0 to 2000 watts. The solid line in the figure represents the constant output of the 1000-watt micro-nano bubble generator, the dotted line represents the theoretical total oxygenation power requirement, and the gray filled area represents the supplementary output of the mechanical turbulence aerator constrained by the boundary. In (b), the vertical axis represents the mechanical shear force of the water flow from 0 to 0.6 Pascals. The dashed line with triangular markings shows the safe upper limit of the time-varying dynamic fluid shear force, which is exponentially relaxed over time based on the hydration and hardening characteristics of the fertilized egg membrane. The discrete data points with diamond markings show the discrete evaluation shear stress actually applied by the equipment after being forcibly truncated by this upper limit. The two figures together fully illustrate how the power allocation algorithm strictly obeys the time-varying physical damage boundary constraints.
[0235] While specific embodiments of the present invention have been described above, those skilled in the art should understand that these specific embodiments are merely illustrative. Those skilled in the art can omit, substitute, and modify the details of the above methods and systems in various ways without departing from the principles and essence of the present invention. For example, combining the above method steps to perform substantially the same function and achieve substantially the same result using substantially the same method falls within the scope of the present invention.
Claims
1. A method for synergistically regulating a water body for breeding and rearing of an aquatic population, characterized in that, Includes the following steps: Acquire video stream data of aquaculture water bodies collected by multi-view underwater camera array, extract optical flow motion features and image grayscale features from the video frame sequence of the video stream data, and obtain the population movement aggregation index, reproductive material release intensity characterization value and effective optical flow feature point distribution density parameter. The initiation of the reproductive substance release process is determined based on the population movement convergence index, and the initiation time is determined. The population movement convergence index and the reproductive substance release intensity characterization value are fused based on the effective optical flow feature point distribution density parameter to obtain the biochemical composite load index. The biochemical composite load index is integrated to obtain the predicted oxygen consumption increment, the predicted peak oxygen consumption increment, and the estimated value of the fertilized egg equivalent average density. The feedforward control time window is determined in combination with the real-time water temperature and the initiation time. Based on the estimated equivalent average density of the fertilized egg and the hardening process of the fertilized egg over time, a dynamic shear stress safety upper limit sequence that varies with time is generated. Within the feedforward control time window, the total oxygenation power demand is determined based on the predicted oxygen consumption increment, and the input power of the micro-nano bubble generator and the mechanical turbulent aerator is allocated under the constraint of the dynamic shear stress safety upper limit sequence. At the same time, the variable frequency buffer metering pump is controlled to perform feedforward pH compensation based on the operating status of the micro-nano bubble generator and the mechanical turbulent aerator. After the feedforward control time window ends, the equivalent biochemical oxygen demand is estimated by combining the actual dissolved oxygen measurement value and the actual oxygenation output of the micro-nano bubble generator and the mechanical turbulent aerator. The oxygen demand prediction residual between the equivalent biochemical oxygen demand and the predicted peak oxygen demand is calculated. Based on the oxygen demand prediction residual, the preset calibration parameters to be calibrated in feature extraction and load prediction are updated in a closed loop.
2. The method according to claim 1, wherein the method is characterized by, The parameters for obtaining the population movement convergence index, the intensity of reproductive substance release, and the distribution density of effective optical flow feature points include: The global average motion intensity is obtained by calculating the average of the sum of squares of the velocity components of each effective optical flow feature point, and the optical flow divergence variance is obtained by calculating the spatial divergence of the optical flow motion vector field. The global average motion intensity and the optical flow divergence variance are linearly combined to obtain the group motion convergence index. The gray-level information entropy of the image is calculated by statistically analyzing the gray-level histogram of the video frames. After smoothing the image gray-level information entropy sequence, the first-order time derivative is performed, and the absolute value of the first-order time change is taken as the characterization value of the release intensity of the reproductive material. The effective optical flow feature point distribution density parameter is obtained by dividing the total number of effective optical flow feature points in the current video frame by the total number of pixels in the image.
3. The method according to claim 1, wherein the method is characterized by, The obtained biochemical composite load index includes: The Sigmoid function is used to establish a continuous mapping relationship between the effective optical flow feature point distribution density parameter and the dynamic feature effectiveness weight; Calculate the product of the population movement aggregation index and the dynamic feature effectiveness weight, calculate the product of the difference between the number 1 and the dynamic feature effectiveness weight and the product of the reproductive substance release intensity characterization value and the preset feature scale matching coefficient, and sum the two products to obtain the biochemical composite load index.
4. The method according to claim 1, wherein the method is characterized by, The process of obtaining the predicted oxygen consumption increment, the predicted peak oxygen consumption increment, the estimated value of the equivalent average density of fertilized eggs, and the feedforward control time window includes: From the start of the reproductive material release process, the biochemical composite load index is integrated over time. The time integration result is multiplied by the preset oxygen consumption integral mapping coefficient and the fertilized egg density integral mapping coefficient to obtain the predicted oxygen consumption increment and the estimated equivalent average density of the fertilized eggs updated over time. The maximum predicted oxygen consumption increment recorded is determined as the predicted oxygen consumption peak increment. Read the real-time water temperature collected by the temperature sensor, and calculate the biochemical oxygen demand response delay time corresponding to the real-time water temperature based on the Arrhenius relation; The time period from the start of the reproductive substance release process to the end of the biochemical oxygen demand response delay time is defined as the feedforward control time window.
5. The method according to claim 1, wherein the method is characterized by, The generation of the time-varying dynamic shear stress safety upper limit sequence includes: The initial shear stress tolerance threshold is obtained by subtracting the product of the estimated equivalent average density of the fertilized egg and the preset density correction coefficient from the preset basic shear stress tolerance value of the single unit. The time difference between the current operation time and the start time is obtained. Based on the exponential asymptotic relationship between the fertilized egg's shear stress tolerance and hardening time, and combined with the initial shear stress tolerance threshold, the conventional shear stress upper limit, and the egg membrane hardening rate constant, the dynamic shear stress safety upper limit corresponding to different time nodes is calculated, thus forming the dynamic shear stress safety upper limit sequence.
6. The method for coordinated regulation of aquatic population breeding and seedling raising water bodies according to claim 1, characterized in that, The input power for distributing the micro / nano bubble generator and the mechanical turbulence aerator includes: When the total oxygenation power requirement is not greater than the rated upper limit power of the micro-nano bubble generator, the total oxygenation power requirement is used as the feedforward input power of the micro-nano bubble generator. When the total oxygenation power demand exceeds the rated upper limit power of the micro-nano bubble generator, the rated upper limit power of the micro-nano bubble generator is used as the feedforward input power, and the difference between the total oxygenation power demand and the rated upper limit power of the micro-nano bubble generator is used as the supplementary input power of the mechanical turbulent aerator. By utilizing the pre-established inverse mapping relationship between shear stress and power, the maximum allowable input power corresponding to the current dynamic shear stress safety limit is determined, and the supplementary input power of the mechanical turbulent aerator is limited by the maximum allowable input power.
7. The method for coordinated regulation of aquatic population breeding and seedling raising water bodies according to claim 1, characterized in that, The control of the variable frequency buffer metering pump to perform feedforward pH compensation includes: Calculate the product of the input power of the mechanical turbulent aerator and the corresponding dispersion coefficient, calculate the product of the input power of the micro-nano bubble generator and the corresponding dispersion coefficient, add the two products together, and obtain the equivalent dispersion rate of dissolved carbon dioxide. The equivalent escape rate of dissolved carbon dioxide is calculated as the quotient of the equivalent concentration of pH buffer solution in the storage tank to obtain the buffer injection volume rate of the variable frequency buffer metering pump, and the operation of the variable frequency buffer metering pump is controlled. During the operation of the variable frequency buffer metering pump, the real-time measurement value of the pH sensor is read, and the injection is stopped when the real-time measurement value reaches or falls below the lower limit of the preset allowable pH range.
8. The method according to claim 1, wherein the method is characterized by, The calculation of the oxygen consumption prediction residual includes: The baseline dissolved oxygen level is determined based on the steady-state dissolved oxygen data of the aquaculture pond before the reproductive event is triggered. The equivalent biochemical oxygen demand is obtained by subtracting the actual dissolved oxygen measurement at the end of the feedforward control time window from the basic dissolved oxygen benchmark, and adding the equivalent dissolved oxygen increment calculated by the total electrical energy actually input by the micro-nano bubble generator and the mechanical turbulent aerator during the feedforward control time window. The oxygen consumption prediction residual is obtained by subtracting the predicted peak oxygen consumption increment from the equivalent biochemical oxygen consumption.
9. The method according to claim 1, wherein the method is characterized by, The closed-loop update of the preset calibration parameters includes: Construct a state vector consisting of feature calibration coefficients and feature scale matching coefficients, which are the calibration parameters to be calibrated. The state prior estimate of the current breeding cycle is obtained based on the state transition matrix, and the observation matrix is obtained in advance based on historical samples; The Kalman gain is calculated by combining the state prior estimate, the observation matrix, and the preset error covariance. The oxygen consumption prediction residual is used as the observation input, and the state prior estimate is updated posteriorly by combining the Kalman gain. The update result is subjected to amplitude limiting processing by a preset allowable range. The parameters after amplitude limiting are used to replace the preset calibration parameters to be calibrated and participate in the feature calculation and load prediction of the next breeding cycle.
10. A water body coordinated regulation system for breeding and rearing of aquatic population, characterized in that, The method for coordinated regulation of aquatic population breeding and seedling raising water bodies according to any one of claims 1-9 includes: The multi-dimensional feature extraction module acquires video stream data of aquaculture water bodies collected by a multi-view underwater camera array, extracts optical flow motion features and image grayscale features from the video frame sequence of the video stream data, and obtains the population movement aggregation index, reproductive material release intensity characterization value and effective optical flow feature point distribution density parameter. The feedforward load prediction module determines the start of the reproductive substance release process and the start time based on the population movement convergence index. It integrates the population movement convergence index and the reproductive substance release intensity characterization value based on the effective optical flow feature point distribution density parameter to obtain the biochemical composite load index. It integrates the biochemical composite load index to obtain the predicted oxygen consumption increment, the predicted peak oxygen consumption increment, and the estimated value of the fertilized egg equivalent average density. It then determines the feedforward control time window by combining the real-time water temperature and the start time. The shear safety boundary calculation module generates a dynamic shear stress safety upper limit sequence that varies with time, based on the estimated equivalent average density of the fertilized egg and the hardening process of the fertilized egg over time. The oxygenation and pH compensation coordinated control module determines the total oxygenation power demand based on the predicted oxygen consumption increment within the feedforward control time window, and allocates the input power of the micro-nano bubble generator and the mechanical turbulent aerator under the constraint of the dynamic shear stress safety upper limit sequence. At the same time, it controls the variable frequency buffer metering pump to perform feedforward pH compensation based on the operating status of the micro-nano bubble generator and the mechanical turbulent aerator. After the feedforward control time window ends, the closed-loop parameter calibration module estimates the equivalent biochemical oxygen demand by combining the actual dissolved oxygen measurement value and the actual oxygenation output of the micro-nano bubble generator and the mechanical turbulent aerator. It then calculates the oxygen demand prediction residual between the equivalent biochemical oxygen demand and the predicted peak oxygen demand increment, and performs closed-loop updates on the preset calibration parameters in feature extraction and load prediction based on the oxygen demand prediction residual.
