Dissolved oxygen, temperature and salinity synergistic control method for fresh aquatic product transportation process

Abstract

This invention relates to the field of cloud-edge collaborative technology, specifically to a method for coordinated regulation of dissolved oxygen, temperature, and salinity in the transportation of live aquatic products. The edge-side controller collects dissolved oxygen, temperature, and salinity values, as well as signals from aerators, compressors, and salt pumps. It inputs these values ​​into an edge-side predictive compensation model to calculate parameter change gradients and outputs an advance compensation adjustment. The cloud server acquires historical transportation curves and stress mortality indicators to construct a dynamic benchmark collaborative constraint surface. At the current node, it extracts dissolved oxygen, temperature, and salinity thresholds as benchmark input parameters and sends them to the edge-side controller. The edge-side controller calculates the difference between the benchmark input parameters and the advance compensation adjustment, generating a collaborative control command. This method offsets the physical time delay of thermodynamic transfer and salt diffusion in water, avoids overshoot and oscillation, and overcomes the benchmark mismatch caused by fixed thresholds deviating from the actual survival benchmark.

Description

Technical Field

[0001] This invention relates to the field of cloud-edge collaborative technology, specifically to a method for the coordinated regulation of dissolved oxygen, temperature, and salinity in the transportation of live aquatic products. Background Technology

[0002] Current water quality control schemes for transporting live aquatic products typically employ closed-loop control with preset fixed thresholds. Dissolved oxygen, temperature, and salinity sensors are installed in the transport tank, and the controller reads the sensor values ​​at fixed sampling intervals. When the detected dissolved oxygen level is below a preset first fixed threshold, the controller outputs a start signal to activate the aerator; when the detected temperature level is above a preset second fixed threshold, the controller outputs a frequency adjustment signal to reduce the operating power of the refrigeration compressor; and when the detected salinity level deviates from a preset third fixed threshold, the controller outputs an opening signal to activate the salt pump. Throughout the transport process, the thresholds for the three parameters remain constant, and the controller relies solely on the difference between the real-time sensor feedback values ​​and the fixed thresholds for closed-loop adjustment.

[0003] The aforementioned fixed-threshold closed-loop control method suffers from a reference mismatch due to adjustment lag when dealing with multi-parameter physical coupling. Water thermodynamic transfer and salinity diffusion exhibit physical inertia; disturbances in aerated water flow affect temperature stratification; and salinity changes alter gas solubility curves. Relying solely on real-time sensor feedback for closed-loop regulation cannot compensate for the physical time delay between actuator action and sensor response. Furthermore, stress responses in aquatic organisms at different transportation stages cause dynamic changes in basal metabolic rate, causing the consistently unchanged fixed threshold to deviate from the actual survival baseline required by the aquatic organisms at the current stage. This results in a reference mismatch in the control commands generated by the controller based on the fixed threshold and the delayed feedback values, leading to overshooting and oscillations in water quality parameters. Summary of the Invention

[0004] The purpose of this invention is to provide a method for coordinated regulation of dissolved oxygen, temperature and salinity in the transportation of live aquatic products, which can solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A method for coordinated control of dissolved oxygen, temperature, and salinity in the transportation of live aquatic products includes: an edge-side controller that collects dissolved oxygen, temperature, and salinity values ​​of the transport water in real time, and simultaneously extracts aerator start / stop signals, refrigeration compressor operating power, and salt pump opening signals; the edge-side controller inputs the dissolved oxygen, temperature, salinity, aerator start / stop signals, refrigeration compressor operating power, and salt pump opening signals into a preset edge-advanced prediction and compensation model, calculates the parameter change gradient based on water thermodynamics and gas solubility physical laws, and outputs an advanced compensation adjustment amount; a cloud server obtains historical transportation curves and stress mortality indicators of multiple batches of similar aquatic products, constructs a dynamic benchmark coordinated constraint surface with time and transportation mileage as independent variables, extracts dissolved oxygen thresholds, temperature thresholds, and salinity thresholds from the dynamic benchmark coordinated constraint surface at the current transportation node as benchmark input parameters, and sends them to the edge-side controller; the edge-side controller calculates the difference between the benchmark input parameters and the advanced compensation adjustment amount, and generates coordinated control commands for the aerator, refrigeration compressor, and salt pump based on the difference.

[0007] Preferably, the edge-side controller collects dissolved oxygen, temperature, and salinity values ​​of the transport water in real time, and simultaneously extracts the start / stop signals of the aerator, the operating power of the refrigeration compressor, and the opening signal of the salt pump. This includes: the edge-side controller reads the electrical signals of the dissolved oxygen sensor, temperature sensor, and conductivity sensor at a fixed sampling period, and simultaneously reads the relay status word connected to the aerator, the inverter power register value connected to the refrigeration compressor, and the pulse width modulation duty cycle value connected to the salt pump through the underlying driver register; the edge-side controller converts the dissolved oxygen sensor signal, the temperature sensor signal, and the conductivity sensor signal into engineering physical quantities, maps the relay status word, the inverter power register value, and the pulse width modulation duty cycle value into a unified actuator action feature vector, and aligns and merges the engineering physical quantities and the actuator action feature vector according to the timestamp to generate an edge input data matrix.

[0008] Preferably, the calculation of parameter change gradients and output of advance compensation adjustment amounts based on the thermodynamics of water and the physical laws of gas solubility includes: the edge advance prediction compensation model internally stores the coefficient matrix of the water heat transfer differential equation and the partial derivative matrix of gas solubility; the edge advance prediction compensation model inputs the edge input data matrix into the coefficient matrix of the water heat transfer differential equation to solve the current temperature gradient and salinity diffusion gradient, inputs the current temperature gradient and the salinity diffusion gradient into the partial derivative matrix of gas solubility to calculate the dissolved oxygen equilibrium point shift rate; the edge advance prediction compensation model multiplies the dissolved oxygen equilibrium point shift rate, the current temperature gradient, and the salinity diffusion gradient by the corresponding preset prediction time step to generate the dissolved oxygen compensation component, temperature compensation component, and salinity compensation component in the advance compensation adjustment amount.

[0009] Preferably, the cloud server acquires historical transportation curves and stress mortality indicators of multiple batches of similar aquatic products, and constructs a dynamic benchmark collaborative constraint surface with time and transportation mileage as independent variables. This includes: the cloud server extracting historical dissolved oxygen curves, historical temperature curves, and historical salinity curves containing time and mileage labels from the database, and extracting stress mortality indicators for the time periods corresponding to the historical dissolved oxygen curves, historical temperature curves, and historical salinity curves; the cloud server sorting and filtering the extracted historical curve data according to the stress mortality indicators in ascending order, and selecting a preset proportion of historical curve data at the top as a benchmark sample set; and the cloud server performing three-dimensional spline interpolation calculations on the benchmark sample set to generate the dynamic benchmark collaborative constraint surface with the time and mileage labels as grid nodes.

[0010] Preferably, extracting dissolved oxygen threshold, temperature threshold, and salinity threshold from the dynamic benchmark collaborative constraint surface at the current transportation node as benchmark input parameters and sending them to the edge controller includes: the cloud server receiving the current GPS coordinates and current system time reported by the edge controller at preset intervals; the cloud server calculating the cumulative transportation mileage based on the current GPS coordinates, combining the cumulative transportation mileage and the current system time into two-dimensional query coordinates; the cloud server inputting the two-dimensional query coordinates into the dynamic benchmark collaborative constraint surface, calculating the weighted average of four grid nodes within the grid interval where the two-dimensional query coordinates are located using a bilinear interpolation algorithm, splitting the weighted average into the dissolved oxygen threshold, the temperature threshold, and the salinity threshold, and encapsulating the dissolved oxygen threshold, the temperature threshold, and the salinity threshold into a benchmark data frame for sending.

[0011] Preferably, the edge-side controller calculates the difference between the reference input parameter and the advance compensation adjustment amount, and generates a coordinated control command for the aerator, the refrigeration compressor, and the salt pump based on the difference. This includes: the edge-side controller subtracting the dissolved oxygen compensation component from the dissolved oxygen threshold in the reference input parameter to obtain a dissolved oxygen deviation value; subtracting the temperature compensation component from the temperature threshold to obtain a temperature deviation value; and subtracting the salinity compensation component from the salinity threshold to obtain a salinity deviation value. The edge-side controller inputs the dissolved oxygen deviation value into a preset aerator mapping function to calculate the aerator target speed; inputs the temperature deviation value into a preset compressor mapping function to calculate the compressor target frequency; inputs the salinity deviation value into a preset salt pump mapping function to calculate the salt pump target duty cycle; and combines the aerator target speed, the compressor target frequency, and the salt pump target duty cycle into the coordinated control command.

[0012] Preferably, aligning and merging the engineering physical quantities and the actuator motion feature vectors according to timestamps to generate an edge input data matrix includes: when the sampling timestamps of the dissolved oxygen sensor electrical signal, the temperature sensor electrical signal, and the conductivity sensor electrical signal are detected to be inconsistent, the edge-side controller uses the sampling timestamp of the temperature sensor electrical signal as the reference timestamp, performs cubic spline interpolation calculations on the dissolved oxygen sensor electrical signal and the conductivity sensor electrical signal to fill in the missing time point data; the edge-side controller uses a sliding window to extract the engineering physical quantities and the actuator motion feature vectors within a preset time period before the current reference timestamp, performs Kalman filtering calculations on the data within the sliding window, and stacks the filtered multidimensional vectors column-wise to generate the edge input data matrix.

[0013] Preferably, the edge-advanced prediction compensation model internally stores a coefficient matrix of the water body heat transfer differential equation, including: at each transportation start-up stage, the edge-advanced prediction compensation model records the current temperature gradient and the salinity diffusion gradient for multiple consecutive time steps, and inputs the recorded gradient sequence into a recursive least squares estimator; the recursive least squares estimator calculates the actual heat transfer coefficient and the actual salinity diffusion coefficient of the current water body based on the gradient sequence and the actuator action feature vector in the edge input data matrix; the edge-advanced prediction compensation model replaces the initial heat transfer coefficient in the coefficient matrix of the water body heat transfer differential equation with the actual heat transfer coefficient, and replaces the initial salinity diffusion coefficient in the coefficient matrix of the water body heat transfer differential equation with the actual salinity diffusion coefficient.

[0014] Preferably, the extraction of stress mortality indicators for the time period corresponding to the historical dissolved oxygen curve, the historical temperature curve, and the historical salinity curve includes: the cloud server extracting the external environmental temperature curve and the vehicle vertical acceleration curve from the same historical batch as the historical dissolved oxygen curve; the cloud server calculating the integral of the absolute value of the difference between the external environmental temperature curve and the historical temperature curve, and calculating the root mean square value of the vehicle vertical acceleration curve; the cloud server inputting the integral of the absolute value of the difference and the root mean square value as correction factors into a preset mortality correction function to scale the original stress mortality indicators, and using the scaled value as the finally extracted stress mortality indicators.

[0015] Preferably, the dissolved oxygen deviation value is input into a preset aerator mapping function to calculate the target speed of the aerator, including: the aerator mapping function is internally equipped with a proportional-integral controller and an anti-integral saturation logic module; the proportional-integral controller receives the dissolved oxygen deviation value and performs cumulative calculation; when the aerator start / stop signal is detected to be in a closed state, the anti-integral saturation logic module cuts off the integral accumulation channel of the proportional-integral controller and locks the historical integral value; when the aerator start / stop signal is detected to be in a closed state, the anti-integral saturation logic module connects the integral accumulation channel of the proportional-integral controller; the proportional-integral controller calculates an intermediate output based on the currently unlocked accumulated value and the dissolved oxygen deviation value, limits the intermediate output between a preset upper limit value and a preset lower limit value of the physical speed, and outputs the target speed of the aerator.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0017] 1. This invention uses an edge-side controller to input real-time collected water quality values ​​and actuator signals into an edge-advanced prediction and compensation model to calculate the parameter change gradient and output the advanced compensation adjustment amount. Simultaneously, a cloud server constructs a dynamic benchmark collaborative constraint surface based on historical transport curves from multiple batches and stress mortality indicators, extracts benchmark input parameters, and issues them out. The edge-side controller generates collaborative control commands based on the difference between the benchmark input parameters and the advanced compensation adjustment amount. This technique utilizes the edge-side advanced prediction and compensation model to proactively calculate the parameter change gradient and output the advanced compensation adjustment amount in advance, offsetting the physical time delay caused by water thermodynamic transfer and salt diffusion, and eliminating overshoot and oscillations caused by sensor feedback lag. By using the dynamic benchmark collaborative constraint surface constructed in the cloud based on historical data, the issued benchmark input parameters are dynamically adjusted according to changes in aquatic metabolic rate during transport, overcoming the benchmark mismatch defect caused by fixed thresholds deviating from actual survival benchmarks.

[0018] 2. This invention eliminates the impact of inconsistent sampling times and noise interference on the input data matrix by aligning and merging sensor electrical signals and actuator action feature vectors according to timestamps and using a sliding window Kalman filter; it updates the heat transfer coefficient and salinity diffusion coefficient based on gradient sequence and action feature vector using a recursive least squares estimator, making the internal matrix parameters of the edge prediction model fit the current actual water state; it scales the original stress mortality index by calculating correction factors based on the external ambient temperature curve and vehicle vertical acceleration curve, eliminating the deviation of external physical interference from the construction of the reference surface; and it prevents abnormal accumulation of the controller's integral term during actuator shutdown by setting an anti-integral saturation logic module in the aerator mapping function to cut off the integral accumulation channel and lock the integral historical value when the aerator is shut down. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the overall execution process of the coordinated regulation of dissolved oxygen, temperature, and salinity during the transportation of fresh aquatic products according to the present invention.

[0020] Figure 2 This is a flowchart of the edge-side data acquisition and input data matrix generation process of the present invention;

[0021] Figure 3 This is a flowchart of the edge advance prediction compensation model calculation of the present invention;

[0022] Figure 4 This is a flowchart of the cloud-based dynamic benchmark collaborative constraint surface construction process of the present invention;

[0023] Figure 5 This is a flowchart of the cloud-based baseline parameter extraction and distribution process of the present invention;

[0024] Figure 6 This is a flowchart illustrating the collaborative control command generation and aerator control process of the present invention. Detailed Implementation

[0025] refer to Figure 1In one embodiment, the method for coordinated control of dissolved oxygen, temperature, and salinity during the transportation of live aquatic products relies on an edge controller deployed on the aquatic product transport vehicle and a remote cloud server. The edge controller is electrically connected to dissolved oxygen sensors, temperature sensors, and conductivity sensors deployed inside the transport water tank, and simultaneously establishes communication connections with the drive relays of the aerator, the frequency converter drive module of the refrigeration compressor, and the pulse width modulation drive module of the salting pump. The cloud server and the edge controller establish a two-way data interaction channel through a wireless mobile communication network. The edge controller collects the dissolved oxygen, temperature, and salinity values ​​of the transport water in real time, and simultaneously extracts the start / stop signals of the aerator, the operating power of the refrigeration compressor, and the opening signal of the salting pump. Specifically, the edge controller reads the electrical signals output by the dissolved oxygen sensor, the temperature sensor, and the conductivity sensor at a fixed sampling period. The value of the fixed sampling period matches the response bandwidth of the sensors. At the same time, it reads the status word of the relay connected to the aerator, the value of the power register of the frequency converter connected to the refrigeration compressor, and the pulse width modulation duty cycle value of the salting pump through the underlying drive register. The edge-side controller converts the electrical signals from the dissolved oxygen sensor, temperature sensor, and conductivity sensor into corresponding engineering physical quantities—dissolved oxygen value, temperature value, and salinity value—using a preset calibration conversion curve. The salinity value is calculated from the conductivity measurement value combined with the current temperature value using a preset salinity conversion model. The edge-side controller also maps the relay status words, inverter power register values, and pulse width modulation duty cycle values ​​into a unified actuator action feature vector. The relay status word is mapped to the binary value of the aerator start / stop signal, the inverter power register value is mapped to the real-time operating power value of the refrigeration compressor, and the pulse width modulation duty cycle value is mapped to the real-time opening value of the salt pump. Finally, the edge-side controller aligns and merges the converted engineering physical quantities and the mapped actuator action feature vectors according to the timestamps on the same time axis to generate an edge input data matrix.

[0026] Table 1. Dimensions and Parameter Definitions of the Edge Input Data Matrix

[0027] The table above defines the column dimensions of the edge input data matrix and the physical meaning of the corresponding parameters. The row dimensions of the matrix are the continuous sampling points sorted by timestamp. Each row of data corresponds to all the acquisition and mapping parameters at the same sampling time, ensuring that the data input to the subsequent model has time synchronization and dimensional uniformity.

[0028] refer to Figure 3The edge-side controller inputs the generated edge input data matrix into a preset edge advance prediction compensation model. Based on the thermodynamics of water and the physical laws of gas solubility, it calculates the parameter change gradient and outputs the advance compensation adjustment amount. The edge advance prediction compensation model is pre-deployed in the local storage unit of the edge-side controller, storing the coefficient matrix of the water heat transfer differential equation and the partial derivative matrix of gas solubility. The edge advance prediction compensation model inputs the engineering physical quantities and actuator action feature vectors from the edge input data matrix into the coefficient matrix of the water heat transfer differential equation. Through numerical solution, it obtains the temperature gradient and salinity diffusion gradient of the transported water at the current moment. Then, it inputs the obtained current temperature gradient and salinity diffusion gradient into the partial derivative matrix of gas solubility to calculate the dissolved oxygen equilibrium point shift rate under the current water environment. Based on the calculated dissolved oxygen equilibrium point shift rate, current temperature gradient, and salinity diffusion gradient, the edge advance prediction compensation model multiplies them by the corresponding preset prediction time step to generate the dissolved oxygen compensation component, temperature compensation component, and salinity compensation component in the advance compensation adjustment amount. These three compensation components combine to form the complete advance compensation adjustment amount. The predicted time step value is matched with the thermodynamic inertia of water, the salt diffusion rate, and the response delay of the actuator to cover the entire time delay from the actuator action to the sensor detection of parameter changes.

[0029] The cloud server acquires historical transportation curves and stress mortality indicators for multiple batches of similar aquatic products, constructing a dynamic benchmark collaborative constraint surface with time and transportation mileage as independent variables. Specifically, the cloud server's local database stores historical transportation data for the same type of live aquatic products, covering the entire transportation cycle and including water quality parameter curves, actuator action data, transportation trajectory data, and transportation result data for all time periods. The cloud server extracts historical dissolved oxygen curves, historical temperature curves, and historical salinity curves with time and mileage labels from the database, and simultaneously extracts stress mortality indicators for the transportation time periods corresponding to the aforementioned historical dissolved oxygen, temperature, and salinity curves. The cloud server sorts and filters all extracted historical curve data according to the stress mortality indicator in ascending order, selecting a predetermined proportion of the top-ranked historical curve data as the benchmark sample set. The value of the predetermined proportion is determined based on the total amount and dispersion of historical data. The filtering process removes abnormal historical data where the stress mortality indicator exceeds the predetermined range. The cloud server performs three-dimensional spline interpolation calculations on multiple sets of historical curve data in the benchmark sample set to generate a dynamic benchmark co-constraint surface with time labels and mileage labels as two-dimensional grid nodes. The output value of the dynamic benchmark co-constraint surface is the co-constraint parameters of dissolved oxygen, temperature and salinity at the corresponding time and mileage nodes. The grid resolution of the surface is determined based on the sampling density of historical data and the rate of change of parameters during transportation.

[0030] The cloud server extracts dissolved oxygen, temperature, and salinity thresholds from the dynamic benchmark collaborative constraint surface at the current transportation node as benchmark input parameters and sends them to the edge controller. Specifically, the cloud server receives the current GPS coordinates and current system time reported by the edge controller at preset intervals. The reporting interval is an integer multiple of the edge controller's sampling period to ensure time synchronization between the reported data and the data collected by the edge controller. Based on the received current GPS coordinates and the starting coordinates of this transportation, the cloud server calculates the cumulative transportation mileage for this transportation and combines the calculated cumulative transportation mileage with the received current system time to form two-dimensional query coordinates. The cloud server inputs the two-dimensional query coordinates into the pre-generated dynamic benchmark collaborative constraint surface and calculates the weighted average of the four adjacent grid nodes within the grid interval containing the two-dimensional query coordinates using a bilinear interpolation algorithm. The weights of the bilinear interpolation are determined by the relative distances between the two-dimensional query coordinates and the four adjacent grid nodes. The cloud server breaks down the calculated weighted average into dissolved oxygen threshold, temperature threshold, and salinity threshold for the corresponding dimensions. It then encapsulates these thresholds into a standard format reference data frame, which is sent to the edge controller via a wireless mobile communication network. The reference data frame contains the corresponding data's time and mileage tags for data verification and synchronization by the edge controller.

[0031] The weighted average calculation process for bilinear interpolation is achieved using the following formula:

[0032]

[0033] in, This is the weighted average output value corresponding to the two-dimensional query coordinates. This refers to the cumulative transportation mileage value in the two-dimensional query coordinates. This refers to the system time value in the two-dimensional query coordinates. , These represent the upper and lower boundary values ​​of the mileage axis for the grid interval where the two-dimensional query coordinates are located. , These represent the upper and lower boundary values ​​of the time axis for the grid interval where the two-dimensional query coordinates are located. , , , These are the surface output values ​​corresponding to the four adjacent grid nodes.

[0034] The edge-side controller calculates the difference between the reference input parameters and the advance compensation adjustment, and generates coordinated control commands for the aerator, refrigeration compressor, and salt pump based on the difference. Specifically, the edge-side controller subtracts the dissolved oxygen compensation component from the dissolved oxygen threshold in the received reference input parameters to obtain the dissolved oxygen deviation value; subtracts the temperature compensation component from the temperature compensation component in the advance compensation adjustment to obtain the temperature deviation value; and subtracts the salinity compensation component from the salinity threshold in the reference input parameters to obtain the salinity deviation value. The edge-side controller inputs the dissolved oxygen deviation value into a preset aerator mapping function to calculate the aerator target speed, inputs the temperature deviation value into a preset compressor mapping function to calculate the compressor target frequency, and inputs the salinity deviation value into a preset salt pump mapping function to calculate the salt pump target duty cycle. The mapping functions for the aerator, compressor, and salt pump are all pre-stored in the local storage unit of the edge controller. The input to the mapping function is the deviation value of the corresponding parameter, and the output is the control target value of the corresponding actuator. The mapping relationship of the function is obtained by calibrating the physical characteristics of the actuator and the response characteristics of the water parameters. The edge controller combines the calculated target speed of the aerator, the target frequency of the compressor, and the target duty cycle of the salt pump into a coordinated control command, which is output to the corresponding aerator drive relay, the refrigeration compressor inverter drive module, and the salt pump pulse width modulation drive module, respectively, to complete the coordinated regulation of dissolved oxygen, temperature, and salinity of the transport water.

[0035] This embodiment offsets the adjustment delay caused by the physical inertia of water parameter changes through real-time data acquisition and advanced prediction compensation at the edge. By constructing and distributing the cloud-based dynamic benchmark collaborative constraint surface, it realizes the parameter constraint benchmark that is dynamically adjusted with the transportation process, thus avoiding the benchmark mismatch problem caused by fixed thresholds.

[0036] refer to Figure 2In a preferred embodiment, the edge-side controller reads the dissolved oxygen sensor, temperature sensor, and conductivity sensor electrical signals at a fixed sampling period. Simultaneously, it reads the relay status word connected to the aerator, the inverter power register value connected to the refrigeration compressor, and the pulse width modulation duty cycle value connected to the salt pump through the underlying driver register. The edge-side controller converts the dissolved oxygen sensor, temperature sensor, and conductivity sensor electrical signals into engineering physical quantities. It maps the relay status word, inverter power register value, and pulse width modulation duty cycle value into a unified actuator action feature vector. The engineering physical quantities and actuator action feature vectors are aligned and merged according to timestamps to generate an edge input data matrix. Specifically, the edge-side controller marks each set of read sensor electrical signals and register values ​​with a corresponding sampling timestamp. The sampling timestamp is generated by the edge-side controller's local hardware clock to ensure the consistency of the time base. When the sampling timestamps of the dissolved oxygen sensor, temperature sensor, and conductivity sensor electrical signals are found to be inconsistent, the edge controller uses the sampling timestamp of the temperature sensor electrical signal as the reference timestamp and performs cubic spline interpolation calculations on the dissolved oxygen sensor and conductivity sensor electrical signals to fill in the missing time points. The sampling timestamp of the temperature sensor is chosen as the reference because the changing inertia of the temperature parameter matches the sensor's response characteristics better, which can reduce the errors caused by interpolation calculations.

[0037] The calculation process for cubic spline interpolation is achieved through the following basis function formula:

[0038]

[0039] in, For the first Cubic spline interpolation function within a sampling interval The reference timestamp to be interpolated. For the first The timestamp of each original sampling point , , , The first The interpolation polynomial coefficients within a sampling interval are obtained by solving for the constraints of continuous function values, continuous first derivatives, and continuous second derivatives at adjacent sampling points.

[0040] After completing timestamp alignment and interpolation, the edge controller uses a sliding window to extract the engineering physical quantities and actuator action feature vectors within a preset time period prior to the current reference timestamp. The duration of the sliding window matches the predicted time step, and the sliding step size of the window is consistent with the fixed sampling period. The edge controller performs Kalman filtering on the data within the sliding window to eliminate sensor measurement noise and random interference from actuator actions. The state variables of the Kalman filter include six dimensions: dissolved oxygen value, temperature value, salinity value, aerator start / stop status, refrigeration compressor operating power, and salt pump opening degree, corresponding to the core parameter dimensions of the edge input data matrix.

[0041] The Kalman filter calculation process is implemented through the following two core formulas:

[0042] State prediction formula:

[0043]

[0044] State update formula:

[0045]

[0046] in, For the first The prior state estimate at time t. For the first The posterior state estimate at time t. Here is the state transition matrix. To control the input matrix, For the first Control input at any time, For the first The posterior state estimate at time t. For the first Kalman gain at time step For the first The measured value at time, This is the measurement matrix.

[0047] The edge controller stacks the filtered multidimensional vectors column by column to generate an edge input data matrix. Each column of the matrix corresponds to a parameter dimension, and each row corresponds to a sampling timestamp, ensuring that the data in the matrix has time synchronization and noise suppression characteristics.

[0048] Table 2. Definition of Kalman Filter State Variables and Covariance Matrix Parameters

[0049] The table above defines the dimensions, physical meanings, and setting rules of the core parameters in the Kalman filtering process, ensuring that the filtering process can adapt to the parameter variation characteristics of the transport water body and suppress measurement noise and random interference.

[0050] This embodiment solves the problem of asynchronous sampling by multiple sensors by using timestamp alignment and cubic spline interpolation. It also suppresses interference from sensor measurement noise and actuator movements by using sliding window Kalman filtering, thereby improving the accuracy and stability of the edge input data matrix and providing reliable input data for subsequent advance prediction compensation.

[0051] In a preferred embodiment, the edge-side controller inputs the edge input data matrix into a preset edge advance prediction compensation model, calculates the parameter change gradient based on the physical laws of water thermodynamics and gas solubility, and outputs the advance compensation adjustment amount. The edge advance prediction compensation model internally stores the coefficient matrix of the water heat transfer differential equation and the partial derivative matrix of gas solubility. The model's calculation process is entirely based on the physical laws of water thermodynamics and gas solubility, without any black-box computational logic. Specifically, the edge advance prediction compensation model inputs the engineering physical quantities and actuator action feature vectors from the edge input data matrix into the coefficient matrix of the water heat transfer differential equation, and obtains the temperature gradient and salinity diffusion gradient of the transported water at the current moment through numerical solution.

[0052] The expression for the one-dimensional unsteady heat transfer differential equation in water is:

[0053]

[0054] in, For the density of the transported water, The specific heat capacity at constant pressure for transporting water. For water temperature, For time, The heat transfer coefficient of the water body, The coordinates are along the depth direction of the water body. The heat flux density generated by the operation of the refrigeration compressor, The heat flux density generated by the aeration from the aerator. The temperature gradient is the first partial derivative of temperature with respect to time. The above differential equation is obtained by numerically solving it.

[0055] The expression for the salinity diffusion control equation is:

[0056]

[0057] in, For water salinity, The salt diffusion coefficient in water. The salinity mass source term is the salinity produced by the salt pump operation, and the salinity diffusion gradient is the first-order partial derivative of salinity with respect to time. The above governing equations were obtained by numerically solving them.

[0058] The edge-advancing prediction compensation model inputs the solved current temperature gradient and salinity diffusion gradient into the gas solubility partial derivative matrix to calculate the dissolved oxygen equilibrium point shift rate under the current aquatic environment. Dissolved oxygen solubility is calculated based on Henry's Law, and the expression is:

[0059]

[0060] in, This is the concentration at the oxygen solubility equilibrium point. Here, is the Henry coefficient, and is the temperature. With salinity A bivariate function, This represents the partial pressure of oxygen at the water interface.

[0061] The formula for calculating the dissolved oxygen equilibrium point shift rate is:

[0062]

[0063] in, This represents the rate of shift of the dissolved oxygen equilibrium point. This is the partial derivative of the Henry coefficient with respect to temperature. The partial derivative of the Henry coefficient with respect to salinity is given by the two partial derivatives, which together form the partial derivative matrix of gas solubility. This matrix is ​​pre-stored in the internal storage unit of the edge advance prediction compensation model.

[0064] The edge advance prediction compensation model calculates the dissolved oxygen equilibrium point shift rate, the current temperature gradient, and the salinity diffusion gradient, and multiplies them by the corresponding preset prediction time step to generate the dissolved oxygen compensation component, temperature compensation component, and salinity compensation component in the advance compensation adjustment amount. The three compensation components are combined to form the complete advance compensation adjustment amount.

[0065] Furthermore, the edge-advanced prediction compensation model records the current temperature gradient and salinity diffusion gradient for multiple consecutive time steps during each transport initiation phase, and inputs the recorded gradient sequence into the recursive least squares estimator. The recursive least squares estimator calculates the actual heat transfer coefficient and actual salinity diffusion coefficient of the water body based on the gradient sequence and the actuator action feature vector in the edge input data matrix. The parameter update formula for the recursive least squares method is:

[0066]

[0067] in, For the first The constantly updated parameter estimates include the actual heat transfer coefficient and actual salinity diffusion coefficient of the current water body. For the first Parameter estimates at time 10:00 For the first Gain matrix at time step For the first Gradient observations at time t, For the first The regression vector at time step is constructed from the actuator action feature vector in the edge input data matrix.

[0068] The edge-advanced prediction and compensation model replaces the initial heat transfer coefficient in the coefficient matrix of the water body heat transfer differential equation with the calculated actual heat transfer coefficient, and replaces the initial salt diffusion coefficient in the coefficient matrix with the actual salt diffusion coefficient, thus completing the online adaptive update of the model coefficient matrix. The coefficient matrix update process is only performed during the transportation initiation phase. After the update is completed, the model parameters remain fixed until the end of this transportation, avoiding the increase in computational load caused by frequent updates during transportation.

[0069] Table 3. Comparison of updated parameters for the coefficient matrix of the differential equation for heat transfer in water.

[0070]

[0071] The table above defines the initial values, update basis, and triggering conditions of the core parameters in the coefficient matrix of the water heat transfer differential equation, ensuring that the coefficient matrix can adapt to the actual physical characteristics of the current transport water body and improve the accuracy of solving the differential equation.

[0072] This embodiment constructs a prediction model based on differential equations of water thermodynamics and gas solubility physical laws, realizing the advanced prediction of water parameter change trends. The model coefficients are updated online adaptively through recursive least squares method, so that the model parameters fit the actual physical characteristics of the current transport water body, improving the calculation accuracy of the advanced compensation adjustment amount, and further offsetting the adjustment delay caused by physical inertia.

[0073] refer to Figure 4 In a preferred embodiment, the cloud server acquires historical transportation curves and stress mortality indicators for multiple batches of similar aquatic products, and constructs a dynamic benchmark collaborative constraint surface with time and transportation mileage as independent variables. Specifically, the cloud server extracts historical dissolved oxygen curves, historical temperature curves, and historical salinity curves containing time and mileage labels from the database, and extracts stress mortality indicators for the time periods corresponding to the historical dissolved oxygen curves, historical temperature curves, and historical salinity curves. The cloud server extracts the external environmental temperature curve and vehicle vertical acceleration curve for the same historical batch as the historical dissolved oxygen curve. The external environmental temperature curve is collected by the onboard environmental sensors of the transport vehicle, and the vehicle vertical acceleration curve is collected by the onboard inertial measurement unit of the transport vehicle. The time labels of the two sets of curves are completely aligned with the time labels of the historical water quality parameter curves.

[0074] The cloud server calculates the integral of the absolute value of the difference between the external ambient temperature curve and the historical temperature curve. The calculation formula is as follows:

[0075]

[0076] in, The result is the integral of the absolute value of the difference. , These are the start and end times for the corresponding transportation period. This is the external ambient temperature curve. This is a historical water temperature curve.

[0077] The cloud server calculates the root mean square value of the vehicle's vertical acceleration curve using the following formula:

[0078]

[0079] in, This is the root mean square value of vertical acceleration. This represents the number of acceleration sampling points within the corresponding time period. For the first The vertical acceleration values ​​at each sampling point.

[0080] The cloud server uses the integral of the absolute value of the difference and the root mean square value as correction factors input into a preset mortality rate correction function to scale the original stress mortality rate index. The scaled value is then used as the final extracted stress mortality rate index. The formula for calculating the mortality rate correction function is:

[0081]

[0082] in, The final extracted stress mortality rate indicator, As a raw indicator of stress mortality, , These are the weighting coefficients corresponding to the two correction factors. The constant correction term is obtained by regression fitting of the weighting coefficients and the constant correction term using historical transportation data.

[0083] The cloud server sorts and filters the extracted historical curve data according to the final extracted stress mortality rate index in ascending order, and selects the historical curve data of the top-ranked preset proportion as the benchmark sample set. Three-dimensional spline interpolation calculation is performed on the benchmark sample set to generate a dynamic benchmark collaborative constraint surface with time label and mileage label as grid nodes.

[0084] refer to Figure 5The cloud server extracts dissolved oxygen, temperature, and salinity thresholds from the dynamic benchmark collaborative constraint surface at the current transportation node as benchmark input parameters and sends them to the edge controller. Specifically, the cloud server receives the current GPS coordinates and current system time reported by the edge controller at preset intervals, calculates the cumulative transportation mileage based on the current GPS coordinates, and combines the cumulative transportation mileage and current system time into two-dimensional query coordinates. The cloud server inputs the two-dimensional query coordinates into the dynamic benchmark collaborative constraint surface, calculates the weighted average of four grid nodes within the grid interval where the two-dimensional query coordinates are located using a bilinear interpolation algorithm, decomposes the weighted average into dissolved oxygen, temperature, and salinity thresholds, and encapsulates the three thresholds into a benchmark data frame and sends it to the edge controller.

[0085] refer to Figure 6 The edge-side controller calculates the difference between the baseline input parameters and the advance compensation adjustment, and generates coordinated control commands for the aerator, refrigeration compressor, and salt pump based on the difference. Specifically, the edge-side controller subtracts the dissolved oxygen compensation component from the dissolved oxygen threshold in the baseline input parameters to obtain the dissolved oxygen deviation value, subtracts the temperature compensation component from the temperature threshold to obtain the temperature deviation value, and subtracts the salinity compensation component from the salinity threshold to obtain the salinity deviation value. The edge-side controller inputs the dissolved oxygen deviation value into a preset aerator mapping function to calculate the aerator target speed, inputs the temperature deviation value into a preset compressor mapping function to calculate the compressor target frequency, and inputs the salinity deviation value into a preset salt pump mapping function to calculate the salt pump target duty cycle. The target speed of the aerator, the target frequency of the compressor, and the target duty cycle of the salt pump are combined into a coordinated control command.

[0086] The aerator's mapping function internally incorporates a proportional-integral (PI) controller and an anti-integral saturation logic module. The PI controller receives dissolved oxygen deviation values, performs cumulative calculations, and its output calculation formula is as follows:

[0087]

[0088] in, This is the intermediate output of the proportional-integral controller. This is the proportionality coefficient. The integral coefficient is... This is the dissolved oxygen deviation value. This is the integral summation term for the dissolved oxygen deviation value.

[0089] When the aerator start / stop signal is detected as being off, the anti-integral saturation logic module cuts off the integral accumulation channel of the proportional-integral controller and locks the historical integral value. The integral accumulation term remains unchanged at the time of locking, preventing abnormal accumulation of the integral term due to persistent deviation. When the aerator start / stop signal is detected as being on, the anti-integral saturation logic module connects the integral accumulation channel of the proportional-integral controller. The proportional-integral controller calculates the intermediate output based on the currently unlocked accumulated value and the dissolved oxygen deviation value, limiting the intermediate output between the preset upper and lower limits of the physical speed, and outputs the target speed of the aerator. The upper and lower limits of the physical speed are determined by the rated operating parameters of the aerator, ensuring that the output target speed is within the safe operating range of the aerator.

[0090] Table 4. Definition of Logic Parameters for Proportional-Integral Controller and Anti-Integral Saturation Controller of Aerator

[0091]

[0092] The table above defines the physical meaning and setting rules of the core parameters of the proportional-integral controller and the anti-integral saturation logic in the aerator mapping function, ensuring that the controller can stably respond to dissolved oxygen deviation values, while avoiding control overshoot problems caused by integral saturation.

[0093] Both the compressor mapping function and the salt pump mapping function adopt a proportional-integral controller structure with output limiting. The parameters of the mapping function are calibrated based on the physical characteristics of the corresponding actuator and the response characteristics of the water parameters, ensuring that the control process of the three actuators is coordinated and avoiding coupling interference caused by the adjustment of a single parameter to the other two parameters.

[0094] This embodiment improves the accuracy of historical sample screening by introducing correction factors for external environment and transportation vibration, making the constructed dynamic benchmark collaborative constraint surface more in line with the actual survival needs of aquatic product transportation. By constructing the actuator mapping function through a proportional-integral controller with anti-integral saturation logic, the problem of integral accumulation during actuator downtime is avoided, thus improving the stability and reliability of collaborative control commands.

Claims

1. A method for coordinated regulation of dissolved oxygen, temperature, and salinity in the transportation of live aquatic products, characterized in that, include: The edge-side controller collects the dissolved oxygen, temperature, and salinity values ​​of the transport water in real time, and simultaneously extracts the start / stop signals of the aerator, the operating power of the refrigeration compressor, and the opening signal of the salt pump. The edge-side controller inputs the dissolved oxygen value, the temperature value, the salinity value, the aerator start / stop signal, the refrigeration compressor operating power, and the salt pump opening signal into a preset edge advance prediction compensation model. Based on the thermodynamics of water and the physical laws of gas solubility, it calculates the parameter change gradient and outputs the advance compensation adjustment amount. The cloud server obtains historical transportation curves and stress mortality indicators of multiple batches of similar aquatic products, constructs a dynamic benchmark collaborative constraint surface with time and transportation mileage as independent variables, and extracts dissolved oxygen threshold, temperature threshold and salinity threshold from the dynamic benchmark collaborative constraint surface at the current transportation node as benchmark input parameters and sends them to the edge controller. The edge-side controller calculates the difference between the reference input parameter and the advance compensation adjustment amount, and generates coordinated control commands for the aerator, the refrigeration compressor and the salt pump based on the difference.

2. The method for coordinated regulation of dissolved oxygen, temperature, and salinity in the transportation of live aquatic products according to claim 1, characterized in that, The edge-side controller collects dissolved oxygen, temperature, and salinity values ​​of the transport water in real time, and simultaneously extracts the start / stop signals of the aerator, the operating power of the refrigeration compressor, and the opening signal of the salt pump. This includes: the edge-side controller reads the electrical signals of the dissolved oxygen sensor, the temperature sensor, and the conductivity sensor at a fixed sampling period, and at the same time reads the relay status word connected to the aerator, the inverter power register value connected to the refrigeration compressor, and the pulse width modulation duty cycle value connected to the salt pump through the underlying drive register. The edge controller converts the dissolved oxygen sensor electrical signal, the temperature sensor electrical signal, and the conductivity sensor electrical signal into engineering physical quantities. It maps the relay status word, the inverter power register value, and the pulse width modulation duty cycle value into a unified actuator action feature vector. The engineering physical quantities and the actuator action feature vector are aligned and merged according to timestamps to generate an edge input data matrix.

3. The method for coordinated regulation of dissolved oxygen, temperature, and salinity in the transportation of live aquatic products according to claim 1, characterized in that, The parameter change gradient is calculated based on the thermodynamics of water and the physical laws of gas solubility, and the advance compensation adjustment amount is output. This includes: the edge advance prediction compensation model internally stores the coefficient matrix of the differential equation of water heat transfer and the partial derivative matrix of gas solubility. The edge advance prediction compensation model inputs the edge input data matrix into the coefficient matrix of the water body heat transfer differential equation to solve the current temperature gradient and salinity diffusion gradient. It then inputs the current temperature gradient and salinity diffusion gradient into the gas solubility partial derivative matrix to calculate the dissolved oxygen equilibrium point shift rate. The edge advance prediction compensation model generates the dissolved oxygen compensation component, temperature compensation component, and salinity compensation component in the advance compensation adjustment amount by multiplying the dissolved oxygen balance point offset rate, the current temperature gradient, and the salinity diffusion gradient by the corresponding preset prediction time step.

4. The method for coordinated regulation of dissolved oxygen, temperature, and salinity in the transportation of live aquatic products according to claim 1, characterized in that, The cloud server acquires historical transportation curves and stress mortality indicators of multiple batches of similar aquatic products, and constructs a dynamic benchmark collaborative constraint surface with time and transportation mileage as independent variables. This includes: the cloud server extracting historical dissolved oxygen curves, historical temperature curves, and historical salinity curves containing time and mileage labels from the database, and extracting stress mortality indicators for the time periods corresponding to the historical dissolved oxygen curves, historical temperature curves, and historical salinity curves. The cloud server sorts and filters the extracted historical curve data according to the stress mortality rate index from smallest to largest, and selects the top-ranked historical curve data as a baseline sample set. The cloud server performs three-dimensional spline interpolation calculations on the benchmark sample set to generate the dynamic benchmark co-constrained surface with the time label and the mileage label as grid nodes.

5. The method for coordinated regulation of dissolved oxygen, temperature, and salinity in the transportation of live aquatic products according to claim 1, characterized in that, Extracting dissolved oxygen threshold, temperature threshold and salinity threshold from the dynamic benchmark collaborative constraint surface at the current transportation node as benchmark input parameters and sending them to the edge controller includes: the cloud server receiving the current GPS coordinates and current system time reported by the edge controller at preset intervals; The cloud server calculates the cumulative transportation mileage based on the current GPS coordinates and combines the cumulative transportation mileage with the current system time into a two-dimensional query coordinate. The cloud server inputs the two-dimensional query coordinates into the dynamic benchmark collaborative constraint surface, calculates the weighted average of four grid nodes within the grid interval where the two-dimensional query coordinates are located using a bilinear interpolation algorithm, splits the weighted average into the dissolved oxygen threshold, the temperature threshold, and the salinity threshold, and encapsulates the dissolved oxygen threshold, the temperature threshold, and the salinity threshold into a benchmark data frame for distribution.

6. The method for coordinated regulation of dissolved oxygen, temperature, and salinity in the transportation of live aquatic products according to claim 1, characterized in that, The edge-side controller calculates the difference between the reference input parameter and the advance compensation adjustment amount, and generates a coordinated control command for the aerator, the refrigeration compressor and the salt pump based on the difference. The command includes: the edge-side controller subtracts the dissolved oxygen compensation component in the advance compensation adjustment amount from the dissolved oxygen threshold in the reference input parameter to obtain a dissolved oxygen deviation value, subtracts the temperature compensation component from the temperature threshold to obtain a temperature deviation value, and subtracts the salinity compensation component from the salinity threshold to obtain a salinity deviation value. The edge-side controller inputs the dissolved oxygen deviation value into a preset aerator mapping function to calculate the target speed of the aerator, inputs the temperature deviation value into a preset compressor mapping function to calculate the target frequency of the compressor, inputs the salinity deviation value into a preset salt pump mapping function to calculate the target duty cycle of the salt pump, and combines the target speed of the aerator, the target frequency of the compressor, and the target duty cycle of the salt pump into the coordinated control command.

7. The method for coordinated regulation of dissolved oxygen, temperature, and salinity in the transportation of live aquatic products according to claim 2, characterized in that, The engineering physical quantities and the actuator action feature vectors are aligned and merged according to timestamps to generate an edge input data matrix. This includes: when the sampling timestamps of the dissolved oxygen sensor electrical signal, the temperature sensor electrical signal and the conductivity sensor electrical signal are found to be inconsistent, the edge-side controller uses the sampling timestamp of the temperature sensor electrical signal as the reference timestamp to perform cubic spline interpolation calculations on the dissolved oxygen sensor electrical signal and the conductivity sensor electrical signal to fill in the missing time point data. The edge controller uses a sliding window to extract the engineering physical quantities and the actuator action feature vectors within a preset time period before the current reference timestamp. It performs Kalman filtering calculations on the data within the sliding window and stacks the filtered multidimensional vectors column-wise to generate the edge input data matrix.

8. The method for coordinated regulation of dissolved oxygen, temperature, and salinity in the transportation of live aquatic products according to claim 3, characterized in that, The edge advance prediction compensation model internally stores the coefficient matrix of the water heat transfer differential equation, including: the edge advance prediction compensation model records the current temperature gradient and the salinity diffusion gradient for multiple consecutive time steps in each transportation start-up stage, and inputs the recorded gradient sequence into the recursive least squares estimator. The recursive least squares estimator calculates the actual heat transfer coefficient and actual salinity diffusion coefficient of the current water body based on the gradient sequence and the actuator action feature vector in the edge input data matrix. The edge-advanced prediction compensation model replaces the initial heat transfer coefficient in the coefficient matrix of the water body heat transfer differential equation with the actual heat transfer coefficient, and replaces the initial salt diffusion coefficient in the coefficient matrix of the water body heat transfer differential equation with the actual salt diffusion coefficient.

9. The method for coordinated regulation of dissolved oxygen, temperature, and salinity in the transportation of live aquatic products according to claim 4, characterized in that, Extracting stress mortality indicators corresponding to the historical dissolved oxygen curve, the historical temperature curve, and the historical salinity curve for the same time period, including: the cloud server extracting the external environmental temperature curve and the vehicle vertical acceleration curve of the same historical batch as the historical dissolved oxygen curve; The cloud server calculates the integral of the absolute value of the difference between the external ambient temperature curve and the historical temperature curve, and calculates the root mean square value of the vehicle's vertical acceleration curve. The cloud server uses the integral of the absolute value of the difference and the root mean square value as correction factors input into a preset mortality rate correction function to scale the original stress mortality rate index, and uses the scaled value as the final extracted stress mortality rate index.

10. The method for coordinated regulation of dissolved oxygen, temperature, and salinity in the transportation of live aquatic products according to claim 6, characterized in that, The dissolved oxygen deviation value is input into a preset aerator mapping function to calculate the target speed of the aerator, including: the aerator mapping function is equipped with a proportional-integral controller and an anti-integral saturation logic module. The proportional-integral controller receives the dissolved oxygen deviation value and performs cumulative calculation. When the start / stop signal of the aerator is detected to be in the off state, the anti-integral saturation logic module cuts off the integral accumulation channel of the proportional-integral controller and locks the historical integral value. When the start / stop signal of the aerator is detected to be in the on state, the anti-integral saturation logic module connects the integral accumulation channel of the proportional-integral controller. The proportional-integral controller calculates the intermediate output based on the currently unlocked accumulation value and the dissolved oxygen deviation value, limits the intermediate output between the preset upper limit and lower limit of the physical speed, and outputs the target speed of the aerator.

Images