Ultralow-temperature preservation method for aquatic products

Abstract

The invention discloses an aquatic product ultralow-temperature fresh-keeping method which comprises the following steps: collecting data of sensors uniformly distributed and mounted in a freezing chamber, and processing the collected sensor data; collecting a large amount of historical data to train an oxidation risk prediction model, and performing oxidation risk real-time prediction through the oxidation risk prediction model; according to the predicted oxidation risk result, electromagnetic nozzles installed on the top and the side wall of the freezing chamber are controlled to adjust the oxygen concentration; an oxygen absorbent is integrated in an aquatic product package, and a microporous valve is arranged on the surface of the package; phase-change materials are embedded in the wall of the freezing chamber and a goods shelf, and the distribution uniformity of a temperature field in the freezing chamber is monitored in real time through an infrared thermal imager; and after each batch of aquatic products are frozen, detecting the surface peroxide value and thiobarbituric acid value of the aquatic products by using infrared spectroscopy, quantifying the oxidation degree, and feeding back fresh-keeping effect detection data to the deep learning model as a new training sample to update network parameters.

Description

Technical Field

[0001] The present invention relates to the technical field of aquatic product fresh-keeping processing, and in particular to a cryogenic fresh-keeping method for aquatic products. Background Art

[0002] Due to their high protein and polyunsaturated fatty acid content, aquatic products are extremely prone to lipid oxidation, microbial spoilage, and enzymatic reactions during storage, resulting in quality deterioration. Cryogenic freezing (usually -40°C to -60°C) extends the shelf life by inhibiting biological activities, but the problem of local lipid oxidation caused by oxygen exposure on the surface of high-fat aquatic products (such as salmon and tuna) is particularly prominent. Research shows that uneven distribution of cold air flow at the initial stage of freezing leads to an increase in local oxygen concentration, accelerating lipid peroxidation reactions, causing surface discoloration, off-flavors, and nutrient loss, directly affecting the commercial value of the product. Therefore, achieving dynamic monitoring of oxygen concentration, precise control of cold air flow, and coordinated optimization of multiple parameters has become the core challenge in improving cryogenic fresh-keeping effects.

[0003] Existing cryogenic fresh-keeping technologies still face the problem of local lipid oxidation caused by oxygen exposure during the freezing process of high-fat aquatic products (such as salmon and tuna). Although existing patents have reduced the oxidation risk through zonal monitoring and dynamic adjustment of cold air flow, there is still room for optimization in terms of real-time performance, self-adaptability, and coordinated control of multiple parameters. Therefore, a more efficient cryogenic fresh-keeping method for aquatic products is needed. Summary of the Invention

[0004] To solve the above-mentioned existing technical problems, the present invention provides a cryogenic fresh-keeping method for aquatic products.

[0005] The technical solution of the present invention is realized as follows:

[0006] A cryogenic fresh-keeping method for aquatic products, characterized by comprising the following steps:

[0007] S1: Collect data from sensors evenly distributed in the freezer and process the collected sensor data;

[0008] S2: Collect a large amount of historical data to train an oxidation risk prediction model and perform real-time prediction of oxidation risk through the oxidation risk prediction model;

[0009] S3: According to the predicted oxidation risk results, control the electromagnetic nozzles installed on the top and side walls of the freezer to adjust the oxygen concentration;

[0010] S4: Integrate an oxygen absorber inside the aquatic product package and set a microporous valve on the package surface;

[0011] S5: Embed phase change materials in the freezer walls and shelves, and use an infrared thermal imager to monitor the uniformity of the temperature field distribution in the freezer in real time;

[0012] S6: After freezing each batch of aquatic products, use infrared spectroscopy to detect the peroxide value and thiobarbituric acid value on the surface of the aquatic products, quantify the degree of oxidation, and feed back the detection data of the preservation effect to the deep learning model to update the network parameters as new training samples.

[0013] Preferably, the step S1 further includes:

[0014] S101: Uniformly install nano oxygen sensors, fiber optic temperature sensors and 3D air flow velocity sensors in the freezer to cover all areas of the freezer, including corners, edges and the central position, so as to comprehensively monitor the environmental parameters;

[0015] Adopt a grid layout according to the freezer structure and air flow distribution characteristics, and appropriately increase the sensor density above, below and around the product placement area to ensure accurate acquisition of the environmental data around the product;

[0016] S102: Real-time collect parameter data such as oxygen concentration, temperature, humidity, cold air flow velocity and direction, and the fat thickness on the surface of the aquatic products through each sensor, and transmit the data to the data processing center to construct a dynamic database. The data transmission can adopt a LoRa wireless communication module to ensure stable communication in a low-temperature environment;

[0017] Use data fusion algorithms such as the Kalman filtering method to process multi-source heterogeneous data, and fuse the data of nano oxygen sensors at different positions to improve the accuracy and reliability of the data. The calculation formula is:

[0018] X_k^' = (X_{k - 1}+Z_k) / 2;

[0019] Among them, X_k^' is the estimated value of the oxygen concentration after fusion, X_{k - 1} is the estimated value at the previous moment, and Z_k is the oxygen concentration value measured by the sensor at the current moment;

[0020] For the temperature and humidity parameters, use the weighted average method to fuse the data, and organize and store the fused data to form a complete dynamic database.

[0021] Preferably, the step S2 further includes:

[0022] S201: Collect a large amount of historical data, including the oxygen concentration change rate, cold air flow uniformity index, product fat content parameters and corresponding oxidation risk level data, as training samples;

[0023] Construct an oxidation risk prediction model using a convolutional neural network. Extract features and learn patterns from the input data through convolutional layers, pooling layers, and fully connected layers, and automatically identify complex associations in the data. During model training, normalize, denoise, and select features for the data, including normalizing the oxygen concentration change rate data. The formula is:

[0024] X_{norm} = (X - X_{min}) / (X_{max} - X_{min});

[0025] Map the original data X to the interval [0, 1], where X_{min} and X_{max} are the minimum and maximum values of the data respectively;

[0026] S202: Input the real-time multi-parameter data into the trained CNN model, update the oxidation risk level every 5 seconds, and divide it into low, medium, and high risk areas according to the level;

[0027] According to the prediction result, dynamically adjust the weight coefficients (ω1, ω2) through a reinforcement learning algorithm to meet the fresh-keeping requirements of different aquatic product types. Among them, the weight coefficient update formula is:

[0028]

[0029] Among them, ω is the weight coefficient, η is the learning rate, r_k is the current reward value, γ is the discount factor, and V(s_k) is the value function of the state s_k, is the gradient of the value function with respect to the weight coefficient.

[0030] Preferably, step S3 further includes:

[0031] S301: Install a micro electromagnetic nozzle array on the top and side walls of the freezer. Each nozzle is equipped with an independent control unit and adopts a rotatable design. It is roughly positioned according to the freezer layout during initial installation and can dynamically adjust the direction and angle during operation;

[0032] According to the AI-predicted oxidation risk result, control the on / off, cold air flow direction, and speed of the micro electromagnetic nozzle. The control signal is adjusted by the control system through the drive circuit to control the nozzle solenoid valve and the fan. Specifically, the nozzles near the high-risk area will increase the cold air flow speed, adjust the direction, and reduce the oxygen concentration;

[0033] S302: Use CFD simulation software to set the boundary and initial conditions such as the size, shape, product placement position, and cold air inlet and outlet positions of the freezer, solve the Navier-Stokes equation, simulate the air flow distribution in the freezer, and obtain the velocity field, pressure field, and temperature field distributions;

[0034] Generate an optimal air flow distribution template based on the simulation results, determine the cold air velocity and direction parameters in each area, optimize the nozzle layout and control parameters, reduce the oxygen accumulation in the dead corner area, and achieve the best air flow distribution effect.

[0035] Preferably, the step S4 further includes:

[0036] S401: Integrate an oxygen absorber (such as iron powder, sulfite, etc.) inside the aquatic product packaging, absorb oxygen through a chemical reaction, and reduce the initial oxygen concentration inside the packaging to below 5%. The chemical reaction formula is as follows:

[0037] 4Fe + 3O2 → 2Fe2O3;

[0038] At the same time, integrate a CO2 slow-release film to slowly release carbon dioxide gas, adjust the gas composition inside the packaging, and inhibit the growth of microorganisms and oxidation reactions;

[0039] S402: Set a microporous valve on the packaging surface, and its opening and closing states are automatically adjusted according to the oxygen concentration in the external freezer. When the external oxygen concentration increases, the microporous valve opens appropriately to accelerate the gas exchange rate between the inside and the outside. Conversely, it closes partially to slow down the exchange rate;

[0040] The gas exchange rate can be expressed as:

[0041] Q = kAΔP;

[0042] where Q is the gas exchange rate, k is the gas permeability coefficient, A is the effective area of the microporous valve, and ΔP is the air pressure difference between the inside and outside of the packaging, realizing the "internal and external collaborative oxygen inhibition" mechanism and reducing the oxidation risk inside the aquatic product packaging.

[0043] Preferably, the step S5 further includes;

[0044] S501: Embed a paraffin-based phase change material, namely PCM, with a phase change temperature of -45°C in the freezer wall and shelves. The paraffin is encapsulated in an aluminum foil bag in a packaged form and fixed in the corresponding position;

[0045] Design the thickness and distribution of the phase change material according to the heat load and temperature fluctuation of the freezer. Determine the dosage and distribution by calculating the maximum heat load of the freezer and the latent heat of phase change of the phase change material. The formula is:

[0046] m = Q / L;

[0047] where m is the mass of the phase change material, Q is the heat load of the freezer, and L is the latent heat of phase change of the phase change material to effectively absorb the energy of temperature fluctuation;

[0048] S502: Use an infrared thermal imager to monitor the uniformity of the temperature field distribution in the freezer in real time. Install it on the top and side of the freezer, scan the inner wall and the surface of the shelves to obtain a temperature distribution image, and evaluate the uniformity of the temperature field using the standard deviation.

[0049] When the temperature field is uneven and the temperature fluctuation exceeds the set threshold, automatically trigger the PCM charge / discharge cooling cycle to maintain the stability of the freezer temperature.

[0050] Preferably, the step S6 further includes:

[0051] S601: After each batch of freezing is completed, use near-infrared spectroscopy to detect the peroxide value and thiobarbituric acid value on the surface of aquatic products to quantify the degree of oxidation. The calculation formula for the peroxide value is:

[0052] PV = a + bλ1 + cλ2 + … + nλ_n;

[0053] where PV is the peroxide value, a, b, c…n are calibration model coefficients, and λ1, λ2…λ_n are absorbance values at the characteristic wavelengths of the near-infrared spectrum.

[0054] Feed the fresh-keeping effect detection data back to the deep learning model, update the network parameters as new training samples, use the backpropagation algorithm to update the parameters of the CNN model, and adjust the model weights and biases according to the error between the predicted value and the actual detected value to minimize the error function, which is calculated by the mean square error. The formula is:

[0055] MSE = (1 / n)∑(Y_pred - Y_true)^2;

[0056] where MSE is the mean square error, n is the number of samples, Y_pred is the predicted value, and Y_true is the actual value. Through the "monitoring - adjustment - verification - optimization" closed-loop mechanism, continuously improve the performance and effect of the fresh-keeping method.

[0057] The beneficial effects of the present invention are:

[0058] 1. The present invention adopts a deep learning model combined with a reinforcement learning algorithm to dynamically update the oxidation risk level every 5 seconds, and dynamically adjusts the prediction weights according to real-time data, enabling the model to accurately adapt to the fresh-keeping requirements of different aquatic products such as high-fat fish and shellfish. Moreover, the active packaging integrates an oxygen absorber and a CO2 slow-release film to reduce the initial oxygen concentration; the microporous valve automatically adjusts the gas exchange rate according to the external oxygen concentration, and cooperates with the intelligent air flow control system to reduce oxygen contact.

[0059] 2. The optimal airflow distribution template based on CFD simulation of the present invention guides the precise adjustment of the cold air flow direction and speed of the nozzle array, reduces the energy loss of the cold air flow in unnecessary areas, dynamically adjusts the cold air flow according to the oxidation risk, and avoids excessive cooling. At the same time, PCM is embedded in the freezer wall and shelves, with a large phase change latent heat, which can absorb and release heat to stabilize the temperature and reduce the frequent start and stop of the refrigeration unit.

[0060] 3. The present invention uses a network composed of nano oxygen sensors, fiber optic temperature sensors and 3D air flow velocity sensors to monitor environmental parameters in real time in all directions. The data fusion algorithm integrates multi-source data, constructs a dynamic database, and provides a comprehensive and accurate basis for intelligent decision-making. Moreover, the preservation effect is detected by near-infrared spectroscopy to quantify the degree of oxidation, and the data is fed back to the deep learning model to update the network parameters, realizing a closed loop of monitoring - adjustment - verification - optimization.

[0061] 4. The solution of the present invention supports the adaptive parameter switching of multiple categories of aquatic products such as fish and crustaceans, without the need for manual resetting.

[0062] 5. The improvement of the preservation effect of the present invention extends the shelf life of aquatic products, reduces storage and transportation losses, improves the economic benefits of enterprises, and the high-quality preservation maintains the excellent quality of aquatic products and enhances the market competitiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic diagram of the working process of a method for ultra-low temperature preservation of aquatic products according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0064] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0065] As Figure 1 shown, the present invention provides a method for ultra-low temperature preservation of aquatic products, including the following steps:

[0066] S1: Collect the data of the sensors evenly distributed and installed in the freezer, and process the collected sensor data;

[0067] S2: Collect a large amount of historical data to train an oxidation risk prediction model, and perform real-time prediction of the oxidation risk through the oxidation risk prediction model;

[0068] S3: According to the predicted oxidation risk result, control the electromagnetic nozzles installed on the top and side walls of the freezer to adjust the oxygen concentration;

[0069] S4: Integrate an oxygen absorber inside the aquatic product packaging and set up a microporous valve on the packaging surface;

[0070] S5: Embed phase change materials in the freezer wall and shelves, and use an infrared thermal imager to monitor the uniformity of the temperature field distribution inside the freezer in real time;

[0071] S6: After each batch of aquatic products is frozen, use infrared spectroscopy to detect the peroxide value and thiobarbituric acid value on the surface of the aquatic products, quantify the degree of oxidation, and feed back the preservation effect detection data to the deep learning model as new training samples to update the network parameters.

[0072] Furthermore, the step S1 further includes:

[0073] S101: Uniformly install nano oxygen sensors, fiber optic temperature sensors and 3D air flow velocity sensors inside the freezer to cover all areas of the freezer, including corners, edges and the center position, so as to comprehensively monitor environmental parameters;

[0074] Adopt a grid layout according to the freezer structure and air flow distribution characteristics, and appropriately increase the sensor density above, below and around the product placement area to ensure accurate acquisition of environmental data around the product;

[0075] S102: Real-time collect parameter data such as oxygen concentration, temperature, humidity, cold air flow velocity and direction, and the fat thickness on the surface of aquatic products through each sensor, and transmit the data to the data processing center to construct a dynamic database. The data transmission can use a LoRa wireless communication module to ensure stable communication in a low-temperature environment;

[0076] Use data fusion algorithms such as the Kalman filter method to process multi-source heterogeneous data, and fuse the data of nano oxygen sensors at different positions to improve the accuracy and reliability of the data. The calculation formula is:

[0077] X_k^'=(X_{k - 1}+Z_k) / 2;

[0078] Among them, X_k^' is the estimated value of the fused oxygen concentration, X_{k - 1} is the estimated value at the previous moment, and Z_k is the oxygen concentration value measured by the sensor at the current moment;

[0079] For temperature and humidity parameters, use the weighted average method to fuse the data, and organize and store the fused data to form a complete dynamic database.

[0080] Furthermore, the step S2 further includes:

[0081] S201: Collect a large amount of historical data, including environmental and product parameters such as the oxygen concentration change rate, cold air flow uniformity index, and product fat content during the ultra-low temperature preservation of different types of aquatic products, as well as the corresponding oxidation risk level data, including oxidation degree indicators such as peroxide value and thiobarbituric acid value obtained through manual detection and historical records;

[0082] Sort and label the collected data, clarify the input parameters of each data sample and the corresponding oxidation risk output results, and form a training data set and a validation data set. Generally, the training data set accounts for a relatively large proportion and is used for updating model parameters, while the validation data set is used to evaluate the performance of the model during training to prevent overfitting;

[0083] Prevent overfitting;

[0084] Remove noise, errors, and missing values from the data. For some obviously abnormal oxygen concentration measurement values, including values beyond the sensor range, they need to be corrected and deleted to ensure the quality of the data;

[0085] Normalize each parameter data so that it is within the same numerical range to improve the training efficiency and accuracy of the model. For oxygen concentration data, it can be normalized to the [-1, 1] interval. The formula is:

[0086] X_{norm} = \frac{2(X - X_{min})}{X_{max} - X_{min}} - 1;

[0087] Among them, X_{norm} is the normalized data, X is the original data, and X_{min} and X_{max} are the minimum and maximum values of the data respectively.

[0088] S202: Input the multi-parameter data during the preservation of aquatic products collected in real time, including the oxygen concentration change rate, cold air flow uniformity index, and product fat content, into the trained CNN model. These data need to go through the same preprocessing steps as the training data, including data cleaning and normalization, to ensure data consistency and model prediction accuracy;

[0089] The CNN model performs forward propagation calculations based on the characteristics of the input data and outputs the corresponding oxidation risk prediction results. For example, for high-fat fish, the model will combine its fat content, current oxygen concentration, and cold air flow situation parameters to predict which of the low, medium, and high oxidation risk levels it is in. The CNN model will update the prediction results every 5 seconds to promptly reflect the changes in the oxidation risk of aquatic products;

[0090] According to the prediction results of the model, relevant parameters in the preservation system are dynamically adjusted through a reinforcement learning algorithm, including the direction and speed of the cold air flow, the gas composition in the modified atmosphere packaging, etc., to reduce the oxidation risk of aquatic products. In this process, the reinforcement learning algorithm automatically updates the weight coefficients according to the error between the prediction results and the actual oxidation risk. The weight coefficient update formula is as follows:

[0091]

[0092] where ω is the weight coefficient, η is the learning rate, r_k is the current reward value, γ is the discount factor, and V(s_k) is the value function of state s_k. is the gradient of the value function with respect to the weight coefficient.

[0093] Furthermore, step S3 further includes:

[0094] S301: According to the layout and size of the freezer, determine suitable positions on the top and side walls to install a micro electromagnetic nozzle array. The installation positions should ensure that the air flow from the nozzles can cover all areas in the freezer, especially pay attention to the product placement area and the corners and edges where oxygen accumulation is likely to occur;

[0095] Each nozzle is equipped with an independent control unit to be able to individually control its on / off state, the direction and speed of the cold air flow. The nozzles are designed to be rotatable. During the initial installation, they are roughly positioned according to the freezer layout, and during subsequent operation, the direction and angle of the nozzles are dynamically adjusted through the control unit;

[0096] According to the oxidation risk results predicted by AI, the control system sends instructions to the nozzle control unit. For areas predicted to have a high oxidation risk, the control system increases the cold air flow speed of the nozzles near this area. The adjustment formula is:

[0097] v new =v base +k v ·R;

[0098] where v new is the new speed, v base is the base speed, k v is the adjustment coefficient, and R is the risk level (R = 2 for high risk);

[0099] At the same time, adjust the nozzle direction so that the air flow directly acts on the risk area to reduce the oxygen concentration. The direction adjustment angle θ is calculated according to the relative position of the risk area, and the formula is:

[0100] θ=arctan(x / y);

[0101] where x and y are the relative coordinate differences;

[0102] For low-risk areas, the cold air flow rate is appropriately reduced to avoid energy waste;

[0103] The adjustment range of the cold air flow rate of the nozzle is 0.5 - 3 m / s, and the adjustment range of the direction is ±15°. The specific adjustment amount is dynamically calculated according to the risk level and the actual position of the product. For example, when the oxidation risk level of a certain area is high, the cold air flow rate of the nozzle can be adjusted according to the following formula:

[0104] v_{new}=v_{base}+k_v\cdotR;

[0105] Where, (v_{new}) is the adjusted cold air flow rate, (v_{base}) is the basic cold air flow rate of the nozzle, (k_v) is the speed adjustment coefficient, and R is the oxidation risk level (which can take 0, 1, 2 corresponding to low, medium, and high risks respectively);

[0106] The direction adjustment angle (theta) is calculated according to the position of the risk area relative to the position of the nozzle. The formula is:

[0107] theta=\arctan\left(frac{y}{x}\right);

[0108] Where, x and y are the horizontal and vertical coordinate differences of the risk area relative to the position of the nozzle respectively.

[0109] S302: Use CFD simulation software to establish a three-dimensional model of the freezer, set boundary conditions and initial conditions. The boundary conditions include the size and shape of the freezer, the placement position of the product, and the positions of the cold air inlet and outlet; the initial conditions include the initial temperature, oxygen concentration, and air flow rate parameters in the freezer;

[0110] Solve the Navier-Stokes equations through CFD simulation to obtain the velocity field, pressure field, and temperature field distributions of the air flow in the freezer. According to the simulation results, analyze the flow situation of the air flow in the freezer, and determine the dead corner areas where oxygen accumulation is likely to occur and the positions where the air flow distribution is uneven;

[0111] According to the CFD simulation results, generate an optimal air flow distribution template. This template defines the cold air flow rate and direction parameters in different areas of the freezer to ensure that the air flow can be evenly distributed throughout the freezer and reduce oxygen accumulation in the dead corner areas. For example, in the template, a higher cold air flow rate is set directly above the product placement area to form a downward blowing air flow to inhibit the rise of oxygen; in the corner areas of the freezer, the nozzle direction is adjusted so that the air flow can form a cycle to avoid oxygen accumulation;

[0112] Match the optimal air flow distribution template with the actual installation positions of the nozzle array to optimize the nozzle layout. If it is found that the air flow distribution in some areas is still not ideal, the installation positions of the nozzles can be appropriately adjusted and the number of nozzles can be increased;

[0113] Determine the basic cold air flow velocity and initial direction of each nozzle according to the velocity and direction parameters in the optimal air flow distribution template. These parameters will be used as the basic values for nozzle control and will be dynamically adjusted according to the AI prediction results during the actual operation process;

[0114] The air flow uniformity index is used to evaluate the optimization effect of the air flow distribution, and its calculation formula is:

[0115] UI = \frac{\sum_{i = 1}^{n}|\mathbf{v}_i - \mathbf{v}_{avg}|}{n\cdot\mathbf{v}_{avg}};

[0116] where UI is the air flow uniformity index, \(\mathbf{v}_i\) i is the air flow velocity vector at the \(i\)-th measurement point, \(\mathbf{v}_{avg}\) avg is the average velocity vector, and \(n\) is the number of measurement points. The optimization goal is to minimize UI; by optimizing the nozzle layout and control parameters, the air flow uniformity index is minimized, thereby achieving the best air flow distribution effect.

[0117] Furthermore, the step S4 further includes:

[0118] S401: Integrate an oxygen absorber and a CO2 slow-release film inside the aquatic product packaging. The oxygen absorber uses iron powder and sulfite, and its chemical reaction formula is:

[0119] 4Fe + 3O2 → 2Fe2O3;

[0120] The CO2 slow-release film is made of a polymer material containing carbon dioxide. By controlling the pore size and thickness of the polymer, the slow release of carbon dioxide is achieved, the gas composition inside the packaging is adjusted, and the growth of microorganisms and oxidation reactions are inhibited;

[0121] Put the processed aquatic products into the packaging integrated with the oxygen absorber and the CO2 slow-release film, and perform sealed packaging. In the initial stage after sealing, the oxygen absorber starts to absorb the oxygen inside the packaging, gradually reducing the oxygen concentration inside the packaging; at the same time, the CO2 slow-release film starts to release carbon dioxide gas, gradually increasing the carbon dioxide concentration inside the packaging;

[0122] S402: Set a microporous valve on the packaging surface, whose opening and closing state is automatically adjusted according to the oxygen concentration in the external freezer. The opening and closing of the microporous valve are controlled by the pressure difference between the external oxygen concentration and the oxygen concentration inside the packaging. When the external oxygen concentration increases, the pressure difference causes the microporous valve to open appropriately, accelerating the gas exchange rate between the inside and the outside, and keeping the oxygen concentration inside the packaging at a low level. Conversely, when the external oxygen concentration decreases, the microporous valve closes partially, slowing down the gas exchange rate and maintaining the stability of the carbon dioxide concentration inside the packaging. The gas exchange rate calculation formula is:

[0123] Q = kAΔP where Q is the gas exchange rate, k is the gas permeability coefficient, A is the effective area of the microporous valve, and ΔP is the pressure difference between the inside and the outside of the packaging;

[0124] The pressure difference ΔP can be calculated from the pressure difference of gas molecule movement caused by the oxygen concentration difference. The formula is:

[0125] ΔP = P_{out}-P_{in};

[0126] where P_{out} is the external oxygen partial pressure and P_{in} is the oxygen partial pressure inside the packaging;

[0127] In one embodiment, by setting a micro oxygen and carbon dioxide sensor inside the packaging to monitor the gas composition inside the packaging in real time, and based on the monitoring data, dynamically adjusting the opening and closing degree of the microporous valve to achieve the dynamic balance of the gas composition inside the packaging. For example, when it is monitored that the oxygen concentration inside the packaging has an upward trend, the opening of the microporous valve is appropriately increased to accelerate the gas exchange rate and reduce the oxygen concentration. Conversely, when the oxygen concentration is lower than the set value, the opening of the microporous valve is reduced to slow down the gas exchange rate and maintain the stability of the oxygen concentration.

[0128] Further, the step S5 further includes:

[0129] S501: Use a paraffin-based phase change material, i.e., PCM, whose phase change temperature is -45°C and the phase change latent heat is L. Encapsulate the paraffin in an aluminum foil bag to form an encapsulated PCM unit, ensuring that the PCM does not leak or contaminate aquatic products during the phase change process;

[0130] Calculate the mass m of the phase change material according to the heat load Q of the freezer and the phase change latent heat L. The formula is:

[0131] m = frac{Q}{L};

[0132] S502: Use an infrared thermal imager to monitor the temperature field distribution inside the freezer in real time. The infrared thermal imager is installed on the top and side of the freezer. By scanning the inner wall of the freezer, the shelves and the surface of the aquatic products, obtain the temperature distribution image;

[0133] The uniformity of the temperature field is evaluated using the standard deviation index, and the calculation formula is:

[0134] σ=sqrt{frac{1}{n}sum_{i=1}^{n}(T_i-bar{T})^2};

[0135] Where σ is the standard deviation, T_i is the temperature at the i-th measurement point, T is the average temperature of all measurement points, and n is the number of measurement points. The smaller the standard deviation, the more uniform the temperature field;

[0136] When the temperature field monitoring result shows that the temperature fluctuation exceeds the set threshold of ±2°C, the charging / discharging cycle of the phase change material is automatically triggered; the phase change material absorbs heat and undergoes a phase change when the temperature rises, changing from a solid state to a liquid state; when the temperature drops, it releases heat and undergoes a phase change, changing from a liquid state to a solid state, thus maintaining the stability of the freezer temperature;

[0137] The phase change process of the phase change material follows the law of conservation of energy, and the calculation formula for the absorbed / released heat is:

[0138] Q=m cdot L;

[0139] Where Q is the heat absorbed / released during the phase change process, m is the mass of the phase change material, and L is the latent heat of phase change;

[0140] Furthermore, during the process of absorbing and releasing heat, the phase change material will have a fine-tuning effect on the temperature in the freezer, thereby indirectly affecting the change rate of the gas composition in the modified atmosphere packaging. Temperature changes will affect the movement speed and diffusion rate of gas molecules. By maintaining the temperature stability, the phase change material helps to keep the gas composition in the modified atmosphere packaging within the set range and extend the shelf life of aquatic products.

[0141] Further, the step S6 further includes:

[0142] S601: After each batch of aquatic products is cryogenically frozen, use a near-infrared spectrometer, i.e., NIRS, to scan and detect the surface of the aquatic products. NIRS obtains spectral characteristic information related to the oxidation degree of the aquatic products by measuring the near-infrared spectral data reflected by the surface of the aquatic products;

[0143] According to the established calibration model, calculate the peroxide value, i.e., PV, and the thiobarbituric acid value, i.e., TBA, of the surface of the aquatic products to quantify the oxidation degree. The calibration model is established based on a large number of samples with known PV and TBA values and their corresponding near-infrared spectral data, and chemometric methods such as partial least squares are used for modeling. The calculation formulas for PV and TBA are respectively:

[0144] PV=a+bλ1+cλ2+…+nλ_n;

[0145] TBA = d + eλ1 + fλ2 + … + pλ_n;

[0146] Wherein, PV is the peroxide value, TBA is the thiobarbituric acid value, a, b, c…n and d, e, f…p are the coefficients of the calibration model, and λ1, λ2…λ_n are the absorbance values at the characteristic wavelengths of the near-infrared spectrum;

[0147] Set the evaluation indicators for the preservation effect. For example, the qualified standard for PV is less than 10 meq / kg, and the qualified standard for TBA is less than 2 mg / kg. According to the test results, count the proportion of PV and TBA exceeding the standard in each batch of aquatic products, and evaluate whether the preservation effect meets the expected goal.

[0148] S602: Feed back the PV and TBA test data obtained from each preservation effect evaluation to the deep learning model as new training samples; sort out the input parameters, including the gas concentration change rate, cold air flow uniformity index, product fat content, and the corresponding output results, including PV and TBA values, to form a new training data set;

[0149] Preprocess the new training data set, including data cleaning and normalization operations, to ensure the quality and consistency of the data;

[0150] Merge the preprocessed new training data set with the previous historical data set and retrain the deep learning model; during the training process, use the same optimization algorithm, the mean square error loss function as before, to update the weight and bias parameters of the model;

[0151] Evaluate the generalization ability and optimization effect of the model by comparing the performance indicators of the model on the training set and the validation set, including loss value, accuracy, etc.; among them, the calculation formula of the mean square error loss function is:

[0152] MSE = \frac{1}{n}\sum_{i = 1}^{n}(Y_{\text{pred},i} - Y_{\text{true},i})

[0153] ^2;

[0154] Wherein, MSE is the mean square error, n is the number of samples, Y_{\text{pred},i} is the PV / TBA value of the i-th sample predicted by the model, and Y_{\text{true},i} is the actual PV / TBA value of the i-th sample.

[0155] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above-described exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to embrace all changes falling within the meaning and scope of the equivalent elements of the claims within the present invention.

Claims

1. A method for ultra-low temperature preservation of aquatic products, characterized in that: It includes the following steps: S1: Collect data from sensors evenly installed in the freezer, and process the collected sensor data. S2: Collect a large amount of historical data to train the oxidation risk prediction model, and perform real-time oxidation risk prediction through the oxidation risk prediction model. S3: According to the predicted oxidation risk results, control the electromagnetic nozzles installed on the top and side walls of the freezer to adjust the oxygen concentration. S4: Integrate an oxygen absorber into the aquatic product packaging, and set a microporous valve on the packaging surface. S5: Embed phase change materials in the freezer wall and shelves, and use an infrared thermal imager to monitor the uniformity of the temperature field distribution in the freezer in real time. S6: After each batch of aquatic products is frozen, use infrared spectroscopy to detect the peroxide value and thiobarbituric acid value on the surface of the aquatic products, quantify the degree of oxidation, and feedback the preservation effect detection data to the deep learning model as new training samples to update the network parameters.

2. The cryogenic preservation method for aquatic products according to claim 1, characterized in that: The step S1 further includes: S101: Evenly install nano oxygen sensors, fiber optic temperature sensors, and 3D air flow velocity sensors in the freezer, covering all areas of the freezer, including corners, edges, and the center position, to comprehensively monitor environmental parameters. Adopt a grid layout according to the freezer structure and air flow distribution characteristics, and appropriately increase the sensor density above, below, and around the product placement area to ensure accurate acquisition of the environmental data around the product. S102: Real-time collect parameter data such as oxygen concentration, temperature, humidity, cold air flow velocity and direction, and the fat thickness on the surface of the aquatic products through each sensor, and transmit the data to the data processing center to build a dynamic database. The data transmission can use a LoRa wireless communication module to ensure stable communication in a low-temperature environment. Use data fusion algorithms such as the Kalman filter method to process multi-source heterogeneous data, and fuse the data of nano oxygen sensors at different positions to improve the accuracy and reliability of the data. The calculation formula is: X_k^'=(X_{k - 1}+Z_k) / 2; where, X_k^' is the estimated value of the fused oxygen concentration, X_{k - 1} is the estimated value at the previous moment, Z_k is the oxygen concentration value measured by the sensor at the current moment; For temperature and humidity parameters, use the weighted average method to fuse the data, and organize and store the fused data to form a complete dynamic database.

3. A method for ultra-low temperature preservation of aquatic products according to claim 1, characterized in that: The step S2 further includes: S201: Collect a large amount of historical data, including the oxygen concentration change rate, cold air flow uniformity index, product fat content parameters, and corresponding oxidation risk level data, as training samples. Use a convolutional neural network to build an oxidation risk prediction model. Extract features and learn patterns from the input data through convolutional layers, pooling layers, and fully connected layers, and automatically identify complex associations in the data. During model training, normalize, denoise, and select features for the data, including normalizing the oxygen concentration change rate data. The formula is: X_{norm}=(X - X_{min}) / (X_{max}-X_{min}); Map the original data X to the [0, 1] interval, where X_{min} and X_{max} are the minimum and maximum values of the data respectively. S202: Input the real-time multi-parameter data into the trained CNN model, update the oxidation risk level every 5 seconds, and divide it into low, medium, and high-risk zones according to the level; According to the prediction result, dynamically adjust the weight coefficients (ω1, ω2) through the reinforcement learning algorithm to meet the preservation requirements of different types of aquatic products. The weight coefficient update formula is as follows: where ω is the weight coefficient, η is the learning rate, r_k is the current reward value, γ is the discount factor, and V(s_k) is the value function of state s_k. is the gradient of the value function with respect to the weight coefficient.

4. A method for ultra-low temperature preservation of aquatic products according to claim 1, characterized in that: The step S3 further includes: S301: Install a micro electromagnetic nozzle array on the top and side walls of the freezer. Each nozzle is equipped with an independent control unit and adopts a rotatable design. It is roughly positioned according to the freezer layout during initial installation and can dynamically adjust the direction and angle during operation; According to the AI-predicted oxidation risk result, control the on / off of the micro electromagnetic nozzle, the direction and speed of the cold air flow. The control signal is adjusted by the control system through the drive circuit to control the nozzle solenoid valve and the fan. Specifically, the nozzles near the high-risk zone will increase the cold air flow speed, adjust the direction, and reduce the oxygen concentration; S302: Use CFD simulation software to set the boundary and initial conditions such as the size, shape, product placement position, cold air inlet and outlet positions of the freezer, solve the Navier-Stokes equation, simulate the air flow distribution in the freezer, and obtain the velocity field, pressure field, and temperature field distributions; Generate an optimal air flow distribution template according to the simulation result, determine the cold air flow speed and direction parameters in each area, optimize the nozzle layout and control parameters, reduce the oxygen accumulation in the dead zone area, and achieve the best air flow distribution effect.

5. A method for ultra-low temperature preservation of aquatic products according to claim 1, characterized in that: The step S4 further includes: S401: Integrate an oxygen absorber (such as iron powder, sulfite, etc.) into the aquatic product packaging. Absorb oxygen through a chemical reaction to reduce the initial oxygen concentration in the packaging to below 5%. The chemical reaction formula is as follows: 4Fe + 3O2 → 2Fe2O3; At the same time, integrate a CO2 slow-release film to slowly release carbon dioxide gas, adjust the gas composition in the packaging, and inhibit the growth of microorganisms and oxidation reactions; S402: Set a micro pore valve on the packaging surface. Its opening and closing state is automatically adjusted according to the oxygen concentration in the external freezer. When the external oxygen concentration increases, the micro pore valve opens appropriately to accelerate the gas exchange rate between the inside and the outside. Conversely, it closes partially to slow down the exchange rate; The gas exchange rate can be expressed as: Q = kAΔP; where Q is the gas exchange rate, k is the gas permeability coefficient, A is the effective area of the micro pore valve, ΔP is the air pressure difference inside and outside the packaging, realizing the "internal and external collaborative oxygen inhibition" mechanism and reducing the oxidation risk in the aquatic product packaging.

6. The cryogenic preservation method for aquatic products according to claim 1, wherein: The step S5 further includes; S501: Embed a paraffin-based phase change material, i.e., PCM, with a phase change temperature of -45°C, in the freezer wall and shelf. The paraffin is encapsulated in an aluminum foil bag and fixed in the corresponding position in a packaged form; Design the thickness and distribution of the phase change material according to the heat load and temperature fluctuation of the freezer. Determine the dosage and distribution by calculating the maximum heat load of the freezer and the latent heat of phase change of the phase change material. The formula is: m = Q / L; where m is the mass of the phase change material, Q is the heat load of the freezer, and L is the latent heat of phase change of the phase change material, to effectively absorb the energy of temperature fluctuation; S502: Use an infrared thermal imager to monitor the uniformity of the temperature field distribution in the freezer in real time. Install it on the top and side of the freezer, scan the inner wall and the surface of the shelves to obtain the temperature distribution image, and evaluate the temperature field uniformity with the standard deviation. When the temperature field is uneven and the temperature fluctuation exceeds the set threshold, automatically trigger the PCM charge / discharge cooling cycle to maintain the stable temperature in the freezer.

7. A method for ultra-low temperature preservation of aquatic products according to claim 1, characterized in that: The said step S6 further includes: S601: After each batch of freezing is completed, use near-infrared spectroscopy to detect the peroxide value and thiobarbituric acid value on the surface of aquatic products to quantify the degree of oxidation. The calculation formula for the peroxide value is: PV = a + bλ1 + cλ2 + … + nλ_n; where PV is the peroxide value, a, b, c…n are calibration model coefficients, and λ1, λ2…λ_n are absorbance values at the characteristic wavelengths of the near-infrared spectrum. Feed the fresh-keeping effect detection data back to the deep learning model as new training samples to update the network parameters. Use the backpropagation algorithm to update the parameters of the convolutional neural network model. According to the error between the predicted value and the actual detected value, adjust the model weights and biases to minimize the error function, which is calculated by the mean square error. The formula is: MSE = (1 / n)∑(Y_pred - Y_true)2; where MSE is the mean square error, n is the number of samples, Y_pred is the predicted value, and Y_true is the actual value. Through the "monitoring - adjustment - verification - optimization" closed-loop mechanism, continuously improve the performance and effect of the fresh-keeping method.

Images