Modeling method of litchi fresh-keeping and unfreezing temperature control model based on thermodynamic model in electromagnetic field

Abstract

The invention provides a modeling method of a litchi fresh-keeping and thawing temperature control model based on a thermodynamic model in an electromagnetic field. The modeling method specifically comprises the following steps: S1, establishing a thermodynamic model of a litchi fresh-keeping and thawing system; s2, obtaining the specific heat capacity of the litchis according to the thermodynamic model; s3, calculating the internal temperature of the litchis; s4, constructing a GRU model, and training the GRU model by using the litchi internal temperature data obtained in the step S3 and the collected unfreezing environment temperature and unfreezing water bath temperature as a training set; s5, a whale optimization algorithm is adopted to optimize the GRU model, the optimized GRU model outputs the litchi thawing temperature Tpred (t), and t is time; and S6, a litchi thawing temperature control strategy based on a fuzzy PID controller is constructed, the Tpred (t) serves as the input of the litchi thawing temperature control strategy, the Tpred (t) is compared with a preset value, and output control is conducted according to the comparison result.

Description

Technical Field

[0001] The present invention belongs to the field of litchi fresh-keeping and thawing, and particularly relates to a method for modeling a litchi fresh-keeping and thawing temperature control model based on a thermodynamic model in an electromagnetic field. Background Art

[0002] In the field of food science, tropical fruits such as litchi are very popular due to their special flavor and nutritional value. However, their marketable period is concentrated, the fresh-keeping is difficult, and the shelf life is extremely short, which severely restricts the market expansion and the improvement of industrial economic benefits. Liquid nitrogen quick-freezing and cold storage is an efficient food freezing technology that can reduce the food temperature below the freezing point in a very short time, forming small and uniform ice crystals, thus maintaining the taste, color, and nutritional value of the food. However, during the freezing and thawing processes, it is easily affected by temperature fluctuations, resulting in poor texture, flavor loss, and degradation of nutritional components. Traditional thawing methods, such as natural room temperature thawing, water bath thawing, or steam thawing, often have problems such as uneven thawing, too long time, and adverse effects on the fruit quality. To solve these problems, researchers have begun to explore methods of using electromagnetic fields and other auxiliary fresh-keeping and thawing to improve the quality of litchi.

[0003] The electromagnetic field resonates with the water molecules in litchi, which will interfere with the formation and stability of hydrogen bonds between water molecules, thereby inhibiting the freezing of internal water in litchi. Based on this principle, the electromagnetic field is applied to litchi ice-temperature fresh-keeping, and the auxiliary thawing equipment can avoid the freezing of water in litchi under the condition of large temperature fluctuations, thereby improving the thawing quality.

[0004] With the in-depth study of the heat conduction theory, it has become possible to establish a thermodynamic model to simulate the heat conduction process. These models can effectively predict the temperature changes of litchi under different environmental conditions and help optimize the thawing strategy. At the same time, the rapid development of deep learning and artificial intelligence technologies has made data-driven methods a reality. By using historical data and real-time monitoring information and combining the output of the thermodynamic model, an intelligent prediction system can be designed to dynamically adjust the temperature control during the thawing process.

[0005] The method of combining the thermodynamic model with deep learning provides a new solution for the thawing process of fruits such as litchi. By accurately simulating the temperature changes of litchi under different environmental conditions through the heat conduction model and combining the non-linear modeling ability of deep learning, the temperature changes inside the fruit during the thawing process can be predicted more accurately; the deep learning algorithm can process and analyze data in real time, enabling the control system to dynamically adjust the heating or cooling strategy and electromagnetic field parameters according to the real-time monitored temperature data to ensure that the fruit is in the best thawing state. Summary of the Invention

[0006] The main object of the present invention is to provide a modeling method for a litchi fresh-keeping and thawing temperature control model based on a thermodynamic model in an electromagnetic field, combining the thermodynamic model with deep learning to establish a more accurate control model.

[0007] To achieve the above object, the technical solution adopted by the present invention is as follows: A modeling method for a litchi fresh-keeping and thawing temperature control model based on a thermodynamic model in an electromagnetic field, specifically including the following steps:

[0008] S1. Establish a thermodynamic model of the litchi fresh-keeping and thawing system;

[0009] S2. Obtain the specific heat capacity of litchi according to the thermodynamic model;

[0010] S3. Calculate the internal temperature of litchi through the following formula:

[0011]

[0012] T lychee (t) is the calculated internal temperature of litchi at time t, q is the heat absorbed by litchi per unit time, calculated from the thermodynamic model, T l ′ ychee is the initial temperature before litchi thawing, obtained by detection, m is the mass of litchi, c lychee is the specific heat capacity of litchi, obtained from step S2;

[0013] S4. Construct a GRU model, and use the internal temperature data of litchi obtained in step S3 and the collected thawing environment temperature and thawing water bath temperature as the training set to train the GRU model;

[0014] S5. Optimize the GRU model using the whale optimization algorithm, and the optimized GRU model outputs the litchi thawing temperature T pred (t), where t is time:

[0015] S6. Construct a litchi thawing temperature control strategy based on a fuzzy PID controller, with T pred (t) as the input of the litchi thawing temperature control strategy, compare T pred (t) with the preset value, and perform output control according to the comparison result.

[0016] Preferably, the step S1 specifically includes the following steps:

[0017] S11: The heater transfers heat to the heat pipe by heat conduction. The heat pipe is a high-efficiency heat-conducting material that quickly transfers heat to the water. The heat output by the heat pipe is obtained through the following formula:

[0018] Q 1 =P(2),

[0019] Among them, Q 1 is the heat transferred from the heat pipe to water per unit time, P is the heating power of the heat pipe, and t is the working time of the heater or the heat pipe;

[0020] S12. The water heated by the heat pipe transfers heat to the air through heat convection. The heat transferred from the water to the air is calculated by the following formula:

[0021] Q 2 = U 2 A water (T water - T air ) (3),

[0022] Among them, Q 2 is the heat transferred from the water to the air per unit time, U 2 is the heat transfer coefficient of the water, A water is the surface area of contact between the water and the air, T water is the current temperature of the water, T airr is the current temperature of the air, T water is the current temperature of the water;

[0023] S13: The air transfers a part of the heat to the litchi. The heat obtained by the litchi from the air transfer is as follows:

[0024] Q 3 = U 3 A air (T air - T lychee ) (4),

[0025] Among them, Q 3 is the heat transferred from the air to the litchi per unit time, U 3 is the heat transfer coefficient of the air, A air is the surface area of contact between the air and the litchi, T airr is the current temperature of the air, T lychee is the current temperature of the litchi;

[0026] Step 14. Since the heat pipe is completely inserted into the water, based on the heat balance relationship, we can obtain:

[0027] Q 1 = Q 2 + Q 3 , and further we can obtain:

[0028] Q 3 = Q 1 - Q 2 (5);

[0029] S15. Substitute Q 3Substitute it into the formula in step 13 to obtain U 3 ;

[0030] S16. Obtain the heat formula of litchi:

[0031] Q 4 =m lychee c lychee (T lychee -T‘ lychee ) (6),

[0032] wherein, Q 4 is the heat absorbed by litchi per unit time, m lychee is the mass of litchi, c lychee is the specific heat capacity of litchi, T lychee is the current temperature of litchi, T‘ lychee is the initial temperature of litchi before thawing;

[0033] Step 17. Since the heat change of litchi is equal to the heat transferred from the air to litchi, therefore, Q 3 =Q 4 , and calculate c lychee through the formula in step S16.

[0034] Preferably, in the step S4, the GRU model includes an input layer, a GRU layer, and an output layer, and the GRU model is controlled by 2 gate structures, specifically including the following steps:

[0035] Step S41. Determine the training set of the GRU model, and the training set is expressed as [x i (t)train, y i (t)train, z i (t)train], wherein,

[0036]

[0037] x i (t)train is the temperature historical data sequence collected by the thawing environment temperature sensor, y i (t)train is the water bath temperature historical data sequence, z i (t)trian is the litchi internal temperature data sequence, m is the sequence length, a is the number of thawing environment temperature data sequences, b is the number of water bath temperature data sequences, and n is the number of litchi temperature data sequences;

[0038] Step S42. Update the 2 control gates and unit information of the GRU model by using the following formula and the training set:

[0039] r t =σ(W r·[h t-1 , x t +B r ) (10),

[0040] z t = σ(W z ·[h t-1 , x t +B z ) (11),

[0041]

[0042] where W r , W z are the weight matrices of the reset gate and the update gate respectively, B r , B z , B h are the bias matrices of the reset gate, the update gate, and the candidate hidden state respectively. r t is the output of the reset gate at time t, zt is the output of the update gate at time t, x t , h t , represent the input, output, and hidden layer update candidate values of the hidden layer nodes at the current time respectively, h t-1 represents the output of the hidden layer nodes at the previous time, ⊙ represents the Hadamard product, represents the matrix element product, σ is the sigmoid activation function with an output range of 0 to 1, tanh is the hyperbolic tangent activation function with an output range of -1 to 1, x t , y t , and z t are the data at time t in the training set [x i (t)train, y i (t)train, z i (t)train];

[0043] Step S43. Obtain the predicted value T t of the output layer of the GRU model as follows:

[0044] T t = (W y h t + b y ) (14),

[0045] where W y is the output weight, and b y is the output bias;

[0046] Preferably, Step S6 specifically includes the following steps:

[0047] S61: Define the input variables and output variable of the fuzzy PID controller. The input variables are the temperature deviation e at time t t and the temperature change rate de / dt:

[0048] e t = T set - T pred (t) (17),

[0049]

[0050] where T set is the set target temperature,

[0051] The output variable u is expressed as follows:

[0052]

[0053] S62: Introduce fuzzy rules: Fuzzify the input quantities. Take the actually measured temperature deviation e t and the temperature change rate de / dt as the inputs of the fuzzy PID controller; Fuzzy inference. Obtain the fuzzy expression of the adjustment parameters of the fuzzy PID controller through fuzzy inference based on the empirical fuzzy control rule table for the temperature deviation and the temperature change rate; Defuzzify. Convert the fuzzy PID controller adjustment parameters ΔK p , ΔK i and ΔK d into explicit numerical values using the centroid method;

[0054] Step S73: According to the obtained adjustment parameters ΔK p , ΔK i and ΔK d update K p , K i , K d as follows:

[0055] K p (n)= K p (n - 1)+ΔK p (n),

[0056] K i (n)= K i (n - 1)+ΔK i (n),

[0057] K d (n)= K d (n - 1)+ΔK d (n),

[0058] In the formula, K p (n - 1), K i(n - 1), K d (n - 1) is the initial proportionality factor, integral factor, and derivative factor, △K p (n), ΔK i (n), ΔK d (n) is the adjusted parameter of the proportionality factor, integral factor, and derivative factor after fuzzy calculation, K p (n), K i (n), K d (n) is the proportionality factor, integral factor, and derivative factor adjusted after fuzzy calculation.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] The model established by the modeling method of the litchi fresh - keeping thawing temperature control model based on the thermodynamic model in the electromagnetic field uses a method combining the thermodynamic model and deep learning to form a hybrid model. The thermodynamic model is used to predict the initial result, which is input into the GRU together with historical data. The GRU is used for further adjustment and optimization, not only improving the accuracy of temperature prediction, but also enhancing the stability and interpretability of the model, and optimizing the thawing temperature control of litchi. At the same time, the whale optimization algorithm (WOA) is used to optimize the training of the GRU model, effectively solving the problem that the model may fall into a local optimal solution in the traditional gradient - descent method, helping to optimize the learning rate scheduling and early - stopping strategy, thereby accelerating the convergence speed and reducing the training time. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 is the modeling flow chart of the present invention;

[0062] Figure 2 is the structural diagram of the GRU model;

[0063] Figure 3 is the flow chart of the whale optimization algorithm;

[0064] Figure 4 is the membership function curve of the fuzzy PID controller;

[0065] Figure 5 is the fuzzy rule table of the fuzzy PID controller. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0066] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variations.

[0067] The modeling method of the litchi fresh - keeping thawing temperature control model based on the thermodynamic model in the electromagnetic field specifically includes the following steps:

[0068] S1. Establish a thermodynamic model to obtain the specific heat capacity of litchi: During thawing, the heater heats the heat pipe, and the heat is inserted into the water to heat the water. The litchi to be thawed is placed directly above the water. The thawing process from the heat pipe through the water, air to the litchi involves multiple heat transfer mechanisms, including heat conduction and heat convection. By obtaining the data of each stage in the heat transfer process, the specific heat capacity of the litchi is calculated:

[0069] S11: The heater transfers heat to the heat pipe by heat conduction. The heat pipe is a high-efficiency heat-conducting material that quickly transfers the heat to the water. The heat output by the heat pipe is obtained through the following formula:

[0070] Q 1 = P (1),

[0071] where Q 1 is the heat transferred from the heat pipe to the water per unit time (W), P is the heating power of the heat pipe (W), and t is the working time of the heater or the heat pipe;

[0072] S12. The water heated by the heat pipe transfers heat to the air by heat convection. The heat transferred from the water to the air is calculated by the following formula:

[0073] Q 2 = U 2 A water (T water - T air ) (3),

[0074] where Q 2 is the heat transferred from the water to the air per unit time (W), U 2 is the heat transfer coefficient of the water (W / (m 2 ·K), A water is the surface area of contact between the water and the air (m 2 ), T water is the current temperature of the water (°C), and T airr is the current temperature of the air (°C);

[0075] S13: The air transfers a part of the heat to the litchi. The heat transferred from the air to the litchi is as follows:

[0076] Q 3 = U 3 A air (T air - T lychee ) (4),

[0077] where Q 3 is the heat transferred from the air to the litchi per unit time (W), U 3 is the heat transfer coefficient of the air (W / (m 2·K), A air is the surface area of air in contact with litchi (m 2 ), T air is the current temperature of air (°C), T lychee is the current temperature of litchi (°C);

[0078] Step 14: Since the heat pipe is completely inserted into water, based on the heat balance relationship, we can obtain:

[0079] Q 1 = Q 2 + Q 3 , and further we can obtain:

[0080] Q 3 = Q 1 - Q 2 ;

[0081] S15: Substitute Q 3 into formula (4) to obtain U 3 ;

[0082] S16: Obtain the heat formula of litchi:

[0083] Q 4 = m lychee c lychee (T lychee - T‘ lychee ) (5),

[0084] where Q 4 is the heat absorbed by litchi per unit time (W), m lychee is the mass of litchi (Kg), c lychee is the specific heat capacity of litchi (J / (Kg·K), T lychee is the current temperature of litchi (°C), T‘ lychee is the initial temperature of litchi before thawing (°C);

[0085] Step 17: Since the heat change of litchi is equal to the heat transferred from air to litchi, therefore, Q 3 = Q 4 , and calculate c lychee through formula (5).

[0086] S2: Establish a temperature prediction model based on the thermodynamic model to calculate the internal temperature T lychee (t):

[0087]

[0088] where t is the thawing time (s), q is the heat absorbed by litchi per unit time (W), calculated from the above thermodynamic model, i.e., q = Q4, T l′ ychee is the initial temperature before litchi thawing, m is the mass of litchi (Kg), and c lychee is the specific heat capacity of litchi (J / (Kg·K)). The input of this temperature prediction model is based on a thermodynamic model, and its main purpose is to generate training data for the GRU model;

[0089] S3: Collect the thawing environment temperature T air , the water bath temperature T water during the thawing process, and the internal temperature T lychee (t) of litchi calculated by formula (6) and other data:

[0090] Perform denoising and normalization processing on the data such as the water bath temperature, thawing environment temperature, and internal temperature of litchi during the thawing process collected. The normalization processing formula is as follows:

[0091]

[0092] Among them, X is the value of the data such as the water bath temperature, thawing environment temperature, and internal temperature of litchi during the thawing process collected above, X′ is the data value after normalization processing, μ is the data mean, and σ is the data standard deviation;

[0093] S4: Construct a GRU model. The GRU model itself uses existing technologies, but relevant data related to litchi preservation and thawing are used for training during training. The GRU model includes an input layer, a GRU layer, and an output layer. Its input is multi-dimensional and the output is one-dimensional, that is, a multi-input single-output model. The GRU unit is controlled by 2 activation gate structures (update gate, reset gate):

[0094] Step S41: Determine the training set of the GRU model, and the training set is expressed as [x i (t)train, y i (t)train, z i (t)train], where

[0095]

[0096]

[0097] x i (t)train is the temperature historical data sequence collected by the thawing environment temperature sensor, y i (t)train is the water bath temperature historical data sequence, that is, both are input values of the model training set, and z i(t) trian is the internal temperature data sequence of litchi, that is, the output value of the model training set. m is the sequence length. The internal temperature data of litchi includes the output temperature data of the thermodynamic model and the historical temperature data collected by the sensor. a is the number of thawing environment temperature data sequences, b is the number of water bath temperature data sequences, and n is the number of litchi internal temperature data sequences.

[0098] Step S42: Update the two control gates and cell information of the GRU model using the following formula and the training set:

[0099] r t = σ(W r · [h t-1 , x t + B r ) (11),

[0100] z t = σ(W z · [h t-1 , x t + B z ) (12),

[0101]

[0102]

[0103] Among them, W r , W z are the weight matrices of the reset gate and the update gate respectively, B r , B z , B h are the bias matrices of the reset gate, the update gate, and the candidate hidden state respectively. r t is the reset gate in the GRU model to control the amplitude of the hidden state of the previous time step flowing into the hidden state of the current time step. z t is the update gate, used to control the amount of state information of the previous moment entering the current state. x t , h t , respectively represent the input, output, and hidden layer update candidate value h t-1 of the hidden layer node at the current moment. h represents the output of the hidden layer node at the previous moment. ⊙ represents the Hadamard product, represents the product of matrix elements. σ is the sigmoid activation function, and the output is from 0 to 1. tanh is the hyperbolic tangent activation function, and the output is from -1 to 1. x t , y t and z t are the training sets [x i (t)train, y (t)train, zi Data at the t-th moment in (t)train

[0104] Step S43: Obtain the predicted value of the output layer of the GRU model as T t :

[0105] T t =(W y h t +b y ) (15),

[0106] where W y is the output weight, determined by the trained W r and W z ; b y is the output bias, determined by the trained B r , B z and B h ; h t is obtained from formula (14), and formulas (11), (12), (13), (14), and (15) constitute the final prediction model, where x t is the input, which is actually the detected water temperature and thawing environment temperature.

[0107] Step S5: Train the GRU model using the whale optimization algorithm, and the whale optimization algorithm adopts the existing technology;

[0108] Step S6: Obtain the thawing temperature T of litchi at the k-th moment finally predicted by the GRU output pred (t):

[0109] Feed the thawing temperature value T of litchi at the t-th moment finally predicted by the GRU output pred (t) back to the fuzzy PID controller and use it as the input value of the fuzzy PID controller;

[0110] Step S7: Design a litchi thawing temperature control strategy based on the fuzzy PID controller, which specifically includes the following steps:

[0111] S71: Define the input variables and output variables of the fuzzy PID controller. The input variables are the temperature deviation e at the t-th moment t and the temperature change rate de / dt:

[0112] e t =T set -T pred (16),

[0113] where T set is the set target temperature,

[0114]

[0115] The output variable u is expressed as follows:

[0116]

[0117] Where K p is the proportionality factor, K i is the integral factor, K d is the differential factor;

[0118] S72: Introduce fuzzy rules: Fuzzify the input variables, taking the actually measured temperature deviation e t and the temperature change rate de / dt as the inputs of the fuzzy PID controller; perform fuzzy inference to obtain the fuzzy representation of the adjustment parameters of the fuzzy PID controller through the fuzzy control rule table based on experience; perform defuzzification to convert the fuzzy adjustment parameters ΔK p 、ΔK i and ΔK d obtained from the fuzzy inference into explicit numerical values using the centroid method;

[0119] Step S73: Update K p 、ΔK i and ΔK d for K p 、K i 、K d as follows:

[0120] K p (n) = K p (n - 1) + ΔK p (n),

[0121] K i (n) = K i (n - 1) + ΔK i (n),

[0122] K d (n) = K d (n - 1) + ΔK d (n),

[0123] In the formula, K p (n - 1), K i (n - 1), K d (n - 1) are the initial proportionality factor, integral factor, and differential factor, and ΔK p (n), ΔK i (n), △K d (n) are the adjusted parameters of the proportionality factor, integral factor, and differential factor obtained through fuzzy calculation, K p (n), Ki (n), K a (n) is the proportionality factor, integral factor, and derivative factor adjusted through fuzzy calculation;

[0124] S8: Use the output u corresponding to the fuzzy PID controller as the control parameter to control the actuator (heating tube) and the electromagnetic generator, and continuously monitor the litchi thawing temperature.

[0125] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection required by the present invention is defined by the appended claims and their equivalents.

Claims

1. A modeling method for a lychee preservation and thawing temperature control model based on a thermodynamic model in an electromagnetic field specifically comprises the following steps: S1. Establish a thermodynamic model of the litchi preservation and thawing system; S2. Obtaining the specific heat capacity of litchi according to a thermodynamic model; S3. Calculate the internal temperature of the litchi using the following formula: T lychee (t) is the calculated internal temperature of litchi at time t, q is the heat absorbed by litchi per unit time, calculated by the thermodynamic model, T l ′ ychee is the initial temperature of litchi before thawing, obtained through testing, m is the mass of litchi, c lychee is the specific heat capacity of litchi, obtained by step S2; S4, constructing a GRU model, using the litchi internal temperature data obtained in step S3 and the collected thawing environment temperature and thawing water bath temperature as a training set to train the GRU model; S5. The whale optimization algorithm is used to optimize the GRU model. The optimized GRU model outputs the thawing temperature T of litchi. pred (t), t is time: S6. Construct a litchi thawing temperature control strategy based on fuzzy PID controller, T pred (t) is used as the input of litchi thawing temperature control strategy. pred (t) Compare with the preset value and perform output control based on the comparison result.

2. The modeling method according to claim 1, characterized in that: The step S1 specifically includes the following steps: S11: The heater transfers heat to the heat pipe by heat conduction. The heat pipe is a highly efficient heat conducting material and quickly transfers heat to water. The heat output by the heat pipe is obtained by the following formula: Q1=P(2), Among them, Q1 is the heat transferred to water by the heat pipe per unit time, P is the heating power of the heat pipe, and t is the working time of the heater or heat pipe; S12. The water heated by the heat pipe transfers heat to the air through thermal convection. The heat transferred from the water to the air is calculated by the following formula: Q2=U2A water (T water -T air ) (3), Among them, Q2 is the heat transferred from water to air per unit time, U2 is the heat transfer coefficient of water, A water is the surface area of ​​water in contact with air, T water is the current temperature of the water, T airr is the current temperature of the air, T water is the current temperature of the water; S13: The air transfers part of the heat to the litchi. The heat transferred by the air to the litchi is as follows: Q3=U3A air (T air -T lychee ) (4), Among them, Q3 is the heat transferred from air to litchi per unit time, U3 is the heat transfer coefficient of air, A air is the surface area of ​​air in contact with litchi, T airr is the current temperature of the air, T lychee is the current temperature of litchi; Step 14: Since the heat pipe is completely inserted into the water, based on the heat balance relationship, it can be obtained that: Q1=Q2+Q3, and then we can get: Q3=Q1-Q2 (5); S15, substituting Q3 into the formula in step 13 to obtain U3; S16. Formula for obtaining the calories of litchi: Q4=m lychee c lychee (T lychee -T l ‘ ychee ) (6), Among them, Q4 is the heat absorbed by litchi per unit time, m lychee is the mass of litchi, c lychee is the specific heat capacity of litchi, T lychee is the current temperature of litchi, T l ‘ ychee is the initial temperature of litchi before thawing; Step 17: Since the heat change of the litchi is equal to the heat transferred from the air to the litchi, Q3 = Q4. The c is calculated by the formula in step S16. lychee .

3. The modeling method according to claim 1, characterized in that: In step S4, the GRU model includes an input layer, a GRU layer, and an output layer. The GRU model is controlled by two gate structures, specifically including the following steps: Step S41: determine the training set of the GRU model, where the training set is represented by [x i (t)train,y i (t)train,z i (t)train], where x i (t)train is the temperature history data sequence collected by the thawing environment temperature sensor, y i (t)train is the historical data sequence of water bath temperature, z i (t)trian is the litchi internal temperature data sequence, m is the sequence length, a is the number of thawing environment temperature data sequences, b is the number of water bath temperature data sequences, and n is the number of litchi temperature data sequences; Step S42: Use the following formula and the training set to update the two control gates and unit information of the GRU model: r t =σ(W r ·[h t-1 ,x t ]+B r ) (10), z t =σ(W z ·[h t-1 ,x t ]+B z ) (11), Among them, W r , W z are the weight matrices of the reset gate and update gate, respectively, r , B z , B h They are the reset gate, update gate, and bias matrix r of the candidate hidden state. t is the output of the reset gate at time t, z t is the output of the update gate at time t, x t ,h t , They represent the input, output and candidate value of the hidden layer node at the current moment, h t-1 represents the output of the hidden layer node at the previous moment, ⊙ represents the Hadamard product, represents the matrix element product, σ is the sigmoid activation function, the output is 0 to 1, tanh is the hyperbolic tangent activation function, the output is -1 to 1, x t ,y t and z t For the training set [x i (t)train,y i (t)train,z i (t)train] in the data at time t; Step S43: Get the predicted value of the output layer of the GRU model as T t : T t =(W y h t +b y ) (14), Among them, W y is the output weight, b y is the output bias.

4. The modeling method according to claim 1, characterized in that ,,Step S6, specifically comprises the following steps: S61: Define the input and output variables of the fuzzy PID controller. The input variable is the temperature deviation e at time t. t And the temperature change rate de / dt: e t =T set -T pred (t) (17), Where T set is the set target temperature, The output variable u is represented as follows: S62: Introduce fuzzy rules: fuzzify the input quantity and convert the actual measured temperature deviation e t and temperature change rate de / dt as the input of the fuzzy PID controller; fuzzy reasoning, through the fuzzy control rule table based on experience, the temperature deviation and temperature change rate are fuzzy inferred to obtain the fuzzy expression of the adjustment parameters of the fuzzy PID controller; Inverse fuzzy, the fuzzy PID controller adjustment parameter ΔK obtained by fuzzy reasoning p , ΔK i and ΔK d The centroid method was used to convert to a clear numerical value; Step S73: According to the obtained adjustment parameter ΔK p , ΔK i and ΔK d To K p , K i , K d To update: K p (n)=K p (n-1)+ΔK p (of), K i (n)=K i (n-1)+ΔK i (of), K d (n)=K d (n-1)+ΔK d (of), In the formula, K p (n-1), K i (n-1), K d (n-1) is the initial proportional factor, integral factor, and differential factor, ΔK p (n), ΔK i (n), ΔK d (n) is the proportional factor, integral factor, and differential factor adjustment parameter after fuzzy calculation, K p (n), K i (n), K d (n) is the proportional factor, integral factor, and differential factor after fuzzy calculation adjustment.

Images