A method for optimizing performance of a lentinus edodes heat pump drying system based on a hybrid driving model

Abstract

The application discloses a kind of based on hybrid drive model's lentinus edodes heat pump drying system performance optimization method, it is related to heat pump drying field.The method includes the following steps: first, establish multi-layer lentinus edodes dryer numerical model.Orthogonal method is carried out numerical experiment sampling, obtains the operating data of dryer outlet air state and moisture content.Then, using the above data trains the dryer import and export parameter proxy model based on LSTM neural network, realizes the fast calculation of dryer import and export parameters.Then, the physical drive model of heat pump system is constructed, and the joint simulation modeling of heat pump system physical drive model and dryer data-driven proxy model is realized.Finally, the performance of heat pump drying system is optimized using genetic algorithm, and the optimal lentinus edodes drying process is obtained.The application can save lentinus edodes heat pump drying system performance optimization test cost and numerical simulation calculation cost, to provide guidance for the performance prediction, selection and operation of lentinus edodes heat pump drying system.

Description

Technical Field

[0001] This invention relates to the field of heat pump drying, and in particular to a method for performance optimization of a shiitake mushroom heat pump drying system based on a hybrid drive model. Background Technology

[0002] To prevent shiitake mushrooms from rotting and turning brown, they need to be dried after harvesting to extend their shelf life. Heat pump drying technology has advantages such as low energy consumption and environmental pollution, a wide range of adjustable temperature and humidity, and less susceptibility to weather conditions, and has been widely used in shiitake mushroom drying unit operations.

[0003] In the actual production of dried shiitake mushrooms, box-type dryers with multiple layers of mushrooms are commonly used. Due to the porosity, heat sensitivity, and biological characteristics of shiitake mushrooms, their drying process involves the coupling of multiple physical fields such as fluid flow, heat transfer, and water vapor transport, resulting in uneven distribution of drying conditions, large dynamic changes in heat and moisture load, and high nonlinearity within the dryer. Therefore, the drying process of multi-layered shiitake mushrooms is difficult to describe using a drying kinetic model based on a single drying condition, while the computational cost of computational fluid dynamics (CFD) numerical methods is too high. Using a data-driven model as a proxy model for the CFD model of the shiitake mushroom dryer can achieve rapid simulation of the heat and moisture load of the multi-layered shiitake mushroom dryer without sacrificing the robustness of the CFD numerical model.

[0004] Optimizing the performance of a shiitake mushroom heat pump drying system requires coupling the inlet and outlet drying medium state parameters of a multi-layer shiitake mushroom dryer to achieve dynamic matching between the energy demand on the dryer side and the energy supply on the heat pump side, thereby improving the drying quality of the shiitake mushrooms and the energy efficiency of the heat pump drying system. Accurate and rapid simulation models can provide low-cost predictions of the drying system's performance under different drying processes, ultimately achieving performance optimization. However, the high computational cost of physical models significantly limits their application in heat pump drying system optimization. Using data-driven surrogate models can effectively reduce model computational costs, thereby overcoming the aforementioned challenges and achieving low-cost performance optimization of the heat pump drying system. Summary of the Invention

[0005] The purpose of this invention is to overcome the above-mentioned shortcomings in the prior art and provide a method for optimizing the performance of a mushroom heat pump drying system based on a hybrid drive model.

[0006] The specific technical solution adopted in this invention is as follows:

[0007] This invention provides a performance optimization method for a mushroom heat pump drying system based on a hybrid driving model, as detailed below:

[0008] Step 1: Construct a data-driven, multi-layered agent model for a mushroom dryer;

[0009] Step 2: Establish a physical model of the heat pump system;

[0010] Step 3: Based on the multi-layer shiitake mushroom dryer proxy model described in Step 1 and the physical model described in Step 2, construct a knowledge and data hybrid driven shiitake mushroom heat pump drying system model, and then use a genetic algorithm to optimize the performance of the shiitake mushroom heat pump drying system model.

[0011] Preferably, step 1 is as follows:

[0012] Step 1-1: Construct a CFD numerical model of a multi-layer shiitake mushroom dryer;

[0013] Steps 1-2: Based on the CFD numerical model of the multi-layer shiitake mushroom dryer, the orthogonal design sampling method is used to determine the sample set;

[0014] Steps 1-3: Based on the sample set, construct a multi-layer mushroom dryer outlet parameter prediction model based on LSTM deep recurrent neural network, i.e., multi-layer mushroom dryer surrogate model.

[0015] Furthermore, in step 1-1, the specific method involves constructing a CFD simulation model of a multi-layer mushroom dryer using a multi-physics coupled porous medium and multiphase structure. The input variables for the CFD numerical model of the multi-layer mushroom dryer include: inlet drying temperature, T... room,in ℃; Inlet dry relative humidity, RH room,in %, %; Inlet wind speed, v room,in ,m / s; uniform density of material, load,m 2 / kg; number of tray layers; initial temperature T0, ℃; initial wet basis moisture content, M 0,wb %, %; Drying time, t, h; Output variables of the CFD numerical model of the multi-layer shiitake mushroom dryer include: outlet air temperature, T room,out ,℃; Relative humidity of outlet air, RH room,out %, %; outlet wind speed, v room,out ,m / s; Real-time wet basis moisture content, M t,wb ,%;

[0016] The specific method for constructing the CFD numerical model of the multi-layer shiitake mushroom dryer is as follows:

[0017] First, a geometric model was established based on the actual multi-layer shiitake mushroom drying box;

[0018] Then, based on the established geometric model, the moisture ratio and volume shrinkage rate of the shiitake mushroom drying process are determined according to the drying kinetic model of a single shiitake mushroom shown in Equation (1) and the shrinkage kinetic model shown in Equation (2). The liquid water evaporation rate of the porous medium domain inside the multi-layer shiitake mushroom dryer is determined using Equation (3), and the porosity of the porous medium domain inside the multi-layer shiitake mushroom dryer is determined using Equation (4).

[0019] (1) MR=exp(-kt) n )

[0020] (2)SR=a+b×MR+c×MR 2

[0021] (3)

[0022] (4)por s =por + SR × (1 - por)

[0023] Where MR is the moisture content of shiitake mushrooms; SR is the volume shrinkage rate of shiitake mushrooms; t is time, in seconds; m evap The evaporation rate of liquid water in the area where shiitake mushrooms are piled up is given by mol / (m²). 3 ·s); por s ρ represents the porosity during the shrinkage process of the accumulated shiitake mushroom area. s The density of dried shiitake mushrooms is kg / m³. 3 ;x s,0 X represents the initial volume fraction of the dry matter inside the shiitake mushroom; m,0 The initial dry basis moisture content of shiitake mushrooms; Mn l ρ is the molar mass of water, kg / mol; por is the initial packing porosity; k, n, a, b, c are model coefficients;

[0024] Based on the principles of conservation of mass, momentum, and energy, the governing equations for turbulent fluid flow, local non-equilibrium heat transfer, and water vapor transport are established. The governing equations for the porous media packing domain of shiitake mushrooms are shown in equations (5), (6), and (7), respectively: (5)

[0026]

[0027]

[0028] (6)

[0030]

[0031]

[0032] (7)

[0033] Where, ρ ma The density of moist air is kg / m³. 3 u is the velocity vector, m / s; p is the pressure, Pa; I is the unit vector; K is the viscous stress tensor, Pa; μ maκ represents the dynamic viscosity of moist air, kg / (m·s); κ represents the permeability of the porous media packing domain, m. 2 dp is the equivalent diameter of a single shiitake mushroom, in meters; Cp ma The specific heat capacity of moist air at constant pressure is J / (kg·K); T ma Temperature of moist air, K; k ma h is the thermal conductivity of moist air, W / (m·K); sf The heat transfer coefficient is the gap heat transfer coefficient, W / (m²). 2 ·K); T m For shiitake mushroom temperature, K; ρ m The density of shiitake mushrooms is kg / m³. 3 ;Cp m The specific heat capacity of shiitake mushrooms under constant pressure, J / (kg·K); k m Here is the thermal conductivity of shiitake mushroom, W / (m·K); H evap The latent heat of vaporization is expressed in J / kg; c v Water vapor concentration, mol / m 3 ;D va Let m be the water vapor-air binary diffusion coefficient. 2 / s;

[0034] Using COMSOL software, the mathematical model of the multi-layer shiitake mushroom dryer was numerically solved using the finite element method, and the model was experimentally verified to obtain the CFD numerical model of the multi-layer shiitake mushroom dryer.

[0035] Furthermore, steps 1-2 are detailed as follows:

[0036] An orthogonal experimental table was generated using Design Expert software. Based on the CFD numerical model of the multi-layer shiitake mushroom dryer described in step 1-1, relevant parameters were modified according to the orthogonal experimental table to perform numerical calculations, and CFD simulation data under different drying conditions and initial conditions were obtained, namely the outlet air state and shiitake mushroom moisture content of the multi-layer shiitake mushroom dryer. A mapping set of one set of inputs corresponding to one set of outputs was determined as the sample set.

[0037] Furthermore, steps 1-3 are detailed below:

[0038] First, the original data of the sample set is preprocessed, and the preprocessed sample set is divided into a training set and a test set. The preprocessing includes operations such as data sorting according to the data collection time, outlier removal and resampling, and data normalization. Then, based on the training set, the hyperparameters of the LSTM model are optimized using methods such as grid search or random search. Finally, the LSTM model with optimal hyperparameters (including but not limited to the number of neurons, input sequence length, learning rate, etc.) is trained using the training set, and the accuracy of the LSTM model is evaluated on the test set to determine whether the error between the output value and the target value meets the requirements, thus obtaining the multilayer mushroom dryer surrogate model.

[0039] Preferably, in step 2, the physical model of the heat pump system includes a compressor model, a condenser model, an expansion valve model, an evaporator model, and a refrigerant charge model.

[0040] The input variables of the compressor model include evaporation temperature, condensation temperature, and evaporator superheat, and the output variables include compressor power, discharge temperature, total mass flow rate at the condenser end, discharge superheat, discharge enthalpy, condensation saturated vapor enthalpy, evaporation saturated vapor enthalpy, evaporation saturated vapor density, suction superheated vapor density, evaporation superheated vapor enthalpy, and suction temperature.

[0041] The input variables of the condenser model include inlet air temperature, air velocity, sum of flow rates of each branch at the condenser end, inlet superheat enthalpy, condensed saturated vapor enthalpy, condenser subcooling, condensing temperature, compressor discharge superheat, and discharge temperature. The output variables include effective pipe length of the condenser branch, pipe length of each phase region, condenser outlet refrigerant subcooled liquid enthalpy, outlet air temperature, and condenser heat exchange.

[0042] The input variables of the expansion valve model include the enthalpy of the refrigerant after mixing at the condenser end, and the output variables include the enthalpy of the refrigerant at the outlet.

[0043] The input variables of the evaporator model include air inlet temperature, face velocity, refrigerant mass flow rate, evaporation temperature, evaporation saturated vapor enthalpy, evaporator inlet refrigerant enthalpy, evaporation saturated vapor density, suction superheated vapor density, evaporation superheated vapor enthalpy, suction temperature, and evaporator air inlet moisture content. The output variables include evaporator branch effective pipe length, superheated zone pipe length, evaporator two-phase zone pipe length, air outlet temperature, evaporator heat exchange, air outlet relative humidity, and evaporator dehumidification capacity.

[0044] In the refrigerant charge model

[0045] Since the refrigerant in a heat pump system is mainly found in the compressor cavity, condenser, evaporator, and refrigerant piping connecting the system, the formula for estimating the refrigerant charge in the system is as follows:

[0046]

[0047] Where M is the refrigerant charge, kg; subscripts TP, sh, and sc represent the two-phase region, superheated region, and subcooled region, respectively; subscripts cond, evap, and comp represent the condenser, evaporator, and compressor, respectively; ρ is the refrigerant density, kg / m³. 3 V is the cavity volume, in meters. 3 Mother is the amount of refrigerant charged in the refrigerant lines excluding the compressor, condenser, and evaporator, expressed in kg.

[0048] Preferably, in step 3, the input variables of the mushroom heat pump drying system model include ambient temperature and humidity, drying temperature, drying humidity, drying wind speed, loading density, number of tray layers and initial moisture content, and the output variables include dehumidification rate per unit energy consumption and heat pump COP.

[0049] The specific method for constructing the model of the shiitake mushroom heat pump drying system is as follows:

[0050] Based on the mathematical relationships between the inlet and outlet parameters of each component, the compressor model, condenser model, multi-layer mushroom dryer proxy model described in step 1, expansion valve model, evaporator model, and refrigerant charge model are connected sequentially. The calculation error of the heat exchanger tube length is reduced by iteratively adjusting the condensing temperature and evaporation temperature. The error between the inlet air state parameters and drying conditions of the multi-layer mushroom dryer proxy model is reduced by adjusting the flow ratio of the flow regulating valve and the compressor frequency. The calculation error of the charge is reduced by adjusting the subcooling. The outlet air state of the dryer is calculated in real time through the multi-layer mushroom dryer proxy model, thereby obtaining the dynamic parameter changes of the heat pump during the mushroom drying process. Finally, a knowledge- and data-driven mushroom heat pump drying system model is constructed.

[0051] Preferably, in step 3, the genetic algorithm optimization process uses the output of the mushroom heat pump drying system model as the fitness function, determines the unit energy consumption dehumidification rate and heat pump COP as optimization objectives, the objective function is to maximize two objectives, and uses the range of values ​​of the input parameters as constraints.

[0052] Furthermore, the specific method for optimizing the performance of the shiitake mushroom heat pump drying system using a genetic algorithm is as follows:

[0053] First, an initial population is determined based on the parameter range, with each individual representing a configuration scheme. The fitness value is calculated using the mushroom heat pump drying system model. All individuals in the initial population are divided into two equal subgroups. Each subgroup corresponds to a sub-objective function, and each sub-objective function performs an independent selection operation within its corresponding subgroup. Then, individuals with high fitness are selected to form a new subgroup, and all new subgroups are merged into a complete population. Crossover and mutation operations are performed on this group to generate the next generation of complete population, and the "splitting-juxtaposition-selection-merging" operation is continuously performed. Based on the fitness function process, an evolutionary search is conducted to finally obtain the Pareto optimal solution.

[0054] Compared with existing technologies, the performance optimization method for a mushroom heat pump drying system based on a hybrid driving model described in this invention has the following advantages:

[0055] First, this invention uses CFD numerical simulation data of a multi-layer shiitake mushroom dryer to construct a fast prediction model for the outlet parameters of the multi-layer shiitake mushroom dryer based on an LSTM deep recurrent neural network, which can reduce the computational cost of the multi-layer shiitake mushroom dryer model.

[0056] Second, the proxy model of the present invention is not affected by the variable and uneven distribution of drying conditions inside the dryer, and uses the derived values ​​of the state parameters of the drying medium at the inlet and outlet of the dryer to calculate the heat and moisture load of the drying process required for the performance optimization of the heat pump drying system.

[0057] Third, this invention utilizes the sequence dependency problem handling capability of LSTM deep recurrent neural networks to comprehensively consider the impact of multiple physical quantities such as drying conditions and mushroom loading on the dynamic performance of the heat pump drying system. It can use genetic algorithms to optimize the drying process that changes over time, effectively improving the dehumidification rate per unit energy consumption and the heat pump COP of the mushroom heat pump drying system. Attached Figure Description

[0058] Figure 1 This is a schematic diagram of the structure of the mushroom heat pump drying system according to an embodiment of the present invention;

[0059] Figure 2 This is a flowchart of the construction process of the proxy model for the inlet and outlet parameters of a multilayer shiitake mushroom dryer based on an LSTM deep recurrent neural network according to an embodiment of the present invention.

[0060] Figure 3 This is a flowchart of a mushroom heat pump drying system based on a hybrid agent model according to an embodiment of the present invention.

[0061] Figure 4 This is a flowchart of the performance optimization process of the mushroom heat pump drying system according to an embodiment of the present invention using a genetic algorithm. Detailed Implementation

[0062] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0063] This invention provides a performance optimization method for a mushroom heat pump drying system based on a hybrid driving model. The method mainly includes the following steps: First, a numerical model of a multi-layer mushroom dryer is established to simulate the changes in outlet air state parameters and moisture content. Then, an orthogonal method is used for numerical experimental sampling to obtain operational data for the multi-layer mushroom dryer under given drying conditions and initial conditions. Next, the above operational data is used to train a surrogate model for the dryer's inlet and outlet parameters based on an LSTM neural network, enabling rapid calculation of the dryer's inlet and outlet parameters. Then, a physical driving model of the heat pump system is constructed, including models for the compressor, condenser, expansion valve, evaporator, and refrigerant charge, etc., and joint simulation modeling of the heat pump system's physical driving model and the dryer's data-driven surrogate model is achieved. Finally, a genetic algorithm is used to optimize the performance of the heat pump drying system to obtain the optimal mushroom drying process. This invention can save on the cost of performance optimization experiments and numerical simulation calculations for mushroom heat pump drying systems, thus providing guidance for performance prediction, selection, and operation of mushroom heat pump drying systems.

[0064] The steps and effects of the optimization method of the present invention will be specifically described below with reference to the embodiments.

[0065] Example

[0066] This invention provides a method for performance optimization of a mushroom heat pump drying system based on a hybrid driving model, specifically including the following steps:

[0067] Step 1: Construct a data-driven, multi-layered agent model for a mushroom dryer, such as... Figure 2 As shown, the specific steps are as follows:

[0068] Step 1-1: Construct a CFD simulation model of a multi-layer mushroom dryer with multi-physics coupling, porous media, and multiphase structure (i.e., a CFD numerical model of a multi-layer mushroom dryer).

[0069] Determine the input and output variables. Model input variables include: inlet drying temperature, T. room,in ℃; Inlet dry relative humidity, RH room,in %, %; Inlet wind speed, v room,in ,m / s; uniform density of material, load,m2 / kg; number of tray layers; initial temperature T0, ℃; initial wet basis moisture content, M 0,wb %, %; drying time, t, h. Model output variables include: outlet air temperature, T. room,out ,℃; Relative humidity of outlet air, RH room,out %, %; outlet wind speed, v room,out ,m / s; Real-time wet basis moisture content, M t,wb The input variable X and output variable Y are shown in the following formula:

[0070] X = [T] room,in ,RH room,in ,v room,in ,load,tray,T0,M 0,wb ,t]

[0071]

[0072] First, a geometric model was established based on the actual dimensions and structure of the multi-layer shiitake mushroom drying box.

[0073] Then, the multi-layered stacked shiitake mushrooms are considered as a porous medium, where the wet substrate of the shiitake mushrooms is the porous medium solid matrix, and the gaps between the stacked shiitake mushrooms are the porous medium pores. The moisture ratio and volume shrinkage rate of the shiitake mushroom drying process are determined according to the drying kinetic model of a single shiitake mushroom shown in Equation (1) and the shrinkage kinetic model shown in Equation (2). The liquid water evaporation rate of the porous medium domain inside the multi-layered shiitake mushroom dryer is determined using Equation (3), and the porosity of the porous medium domain inside the multi-layered shiitake mushroom dryer is determined using Equation (4).

[0074] (1) MR=exp(-kt) n )

[0075] (2)SR=a+b×MR+c×MR 2

[0076] (3)

[0077] (4)por s =por + SR × (1 - por)

[0078] Where MR is the moisture content of shiitake mushrooms; SR is the volume shrinkage rate of shiitake mushrooms; t is time, in seconds; m evap The evaporation rate of liquid water in the area where shiitake mushrooms are piled up is given by mol / (m²). 3 ·s); por s ρ represents the porosity during the shrinkage process of the accumulated shiitake mushroom area. s The density of dried shiitake mushrooms is kg / m³. 3 ;x s,0X represents the initial volume fraction of the dry matter inside the shiitake mushroom; m,0 The initial dry basis moisture content of shiitake mushrooms; Mn l ρ is the molar mass of water, kg / mol; por is the initial packing porosity; k, n, a, b, c are model coefficients.

[0079] Based on the principles of conservation of mass, momentum, and energy, control equations for turbulent fluid flow, local non-equilibrium heat transfer, and water vapor transport are established. The control equations for the porous media packing domain of shiitake mushrooms are shown in equations (5), (6), and (7), respectively: (5)

[0081]

[0082]

[0083] (6)

[0085]

[0086]

[0087] (7)

[0088] Where, ρ ma The density of moist air is kg / m³. 3 u is the velocity vector, m / s; p is the pressure, Pa; I is the unit vector; K is the viscous stress tensor, Pa; μ ma κ represents the dynamic viscosity of moist air, kg / (m·s); κ represents the permeability of the porous media packing domain, m. 2 dp is the equivalent diameter of a single shiitake mushroom, in meters; Cp ma The specific heat capacity of moist air at constant pressure is J / (kg·K); T ma Temperature of moist air, K; k ma h is the thermal conductivity of moist air, W / (m·K); sf The heat transfer coefficient is the gap heat transfer coefficient, W / (m²). 2 ·K); T m For shiitake mushroom temperature, K; ρ m The density of shiitake mushrooms is kg / m³. 3 ;Cp m The specific heat capacity of shiitake mushrooms under constant pressure, J / (kg·K); k m Here is the thermal conductivity of shiitake mushroom, W / (m·K); H evap D is the latent heat of vaporization, J / kg; cv is the water vapor concentration, mol / m3; va Let m be the water vapor-air binary diffusion coefficient. 2 / s.

[0089] Considering the highly nonlinear and multiphysics coupling characteristics of the shiitake mushroom drying process, COMSOL multiphysics simulation software was selected. Four physics modules were used: turbulent flow, rarefied mass transfer, solid heat transfer, and fluid heat transfer. Two additional multiphysics interfaces were added: non-isothermal flow and local thermal nonequilibrium. The mathematical model of the multilayer shiitake mushroom dryer was numerically solved using the finite element method, and experimental verification was then performed.

[0090] Steps 1-2: Determine the sample set using orthogonal design sampling. Generate an orthogonal experimental table using Design Expert software. For the 8 input variables, determine the upper and lower limits and median values, forming a sampling scheme of 120 groups of experiments with 8 factors and 3 levels.

[0091] The value range and level of each factor are shown in Table 1:

[0092] Table 1. Influencing factors and levels of orthogonal array design parameters based on the orthogonal method

[0093]

[0094]

[0095] Based on the numerical model of the multi-layer shiitake mushroom dryer in step 1-1, the relevant parameters were modified according to the experimental design table to perform numerical calculations and obtain CFD simulation data under different drying conditions and initial conditions, namely the outlet air state and shiitake mushroom moisture content of the multi-layer shiitake mushroom dryer. A mapping set of one set of inputs corresponding to one set of outputs was determined as the sample set.

[0096] Steps 1-3: Construct a prediction model for the outlet parameters of a multilayer mushroom dryer based on an LSTM deep recurrent neural network.

[0097] First, all simulation data stored in the simulation database is preprocessed. This includes sorting the data according to the acquisition time, removing outliers and resampling, and normalizing the data. Then, according to the data format requirements of the LSTM model, the dryer operating parameter matrix is ​​reorganized to complete the sample set and obtain the standardized operating parameter matrix.

[0098] First, all simulation data are sorted in chronological order to obtain a simulation data table;

[0099] Second, outliers in the simulation data table are removed. If a row vector is empty, the row data is deleted.

[0100] Third, the data table with outliers is resampled according to the time variable, converting second-level data into minute-level data. The converted parameter value is the mean of the parameter over one minute and is stored in the form of a matrix.

[0101] Fourth, the operating data in the dryer operating parameter matrix X are standardized to obtain the standardized operating parameter matrix. The sampled dataset is transformed so that the mean of each parameter among the eight input variables remains 0 and the standard deviation is 1, thus ensuring a consistent distribution for each parameter. This is illustrated in the following equation:

[0102]

[0103]

[0104]

[0105] The solution dataset is normalized so that the data is bound between 0 and 1, as shown in the following formula:

[0106]

[0107] Next, the preprocessed sample set is divided into a training set and a test set.

[0108] Then, based on the training set, the hyperparameters (including input sequence length, learning rate, number of iterations, number of hidden layers, number of neurons, activation function, batch size, etc.) of the dryer outlet parameter prediction model based on LSTM deep recurrent neural network are initialized. The hyperparameters of the LSTM model are optimized using a grid search method to obtain the optimized hyperparameters.

[0109] Finally, a real-time prediction model for dryer outlet parameters based on an LSTM deep recurrent neural network was constructed. The selected neural network architecture was trained and optimized, accurately capturing the input (product and inlet drying process parameters) - output (outlet air state and real-time moisture content) mapping given by the numerical model. The time series of the independent variables (8 sampling parameters) were used as input to generate the corresponding output variable time series. The LSTM neural network was constructed and trained using the MATLAB platform. The LSTM model was trained using the training set, and its accuracy was evaluated on the test set. Finally, it was determined that the error between the output and target values ​​met the specified requirements.

[0110] Step 2: Establish a physical model of the heat pump system, targeting, for example... Figure 1 The mushroom heat pump drying system shown uses mathematical models and connection relationships for each component. The specific input-output relationships for the compressor model, condenser model, expansion valve model, evaporator model, and refrigerant charge model are as follows:

[0111] Step 2-1: Compressor Model:

[0112] Input variables include evaporation temperature, condensation temperature, and evaporator superheat; output variables include compressor power, discharge temperature, total mass flow rate at the condenser end, discharge superheat, discharge enthalpy, condensation saturated vapor enthalpy, evaporation saturated vapor enthalpy, evaporation saturated vapor density, suction superheated vapor density, evaporation superheated vapor enthalpy, and suction temperature.

[0113] Step 2-2: Condenser Model:

[0114] Input variables include inlet air temperature, air velocity, sum of flow rates of each branch at the condenser end, inlet superheat enthalpy, condensed saturated vapor enthalpy, condenser subcooling, condensing temperature, compressor discharge superheat, and discharge temperature; output variables include effective pipe length of condenser branch, pipe length of each phase region, condenser outlet refrigerant subcooled liquid enthalpy, outlet air temperature, and condenser heat exchange.

[0115] Steps 2-3: Expansion valve model:

[0116] The input variables include the enthalpy of the refrigerant after mixing at the condenser end; the output variables include the enthalpy of the refrigerant at the outlet.

[0117] Steps 2-4: Evaporator Model

[0118] Input variables include air inlet temperature, face velocity, refrigerant mass flow rate, evaporation temperature, evaporation saturated vapor enthalpy, evaporator inlet refrigerant enthalpy, evaporation saturated vapor density, suction superheated vapor density, evaporation superheated vapor enthalpy, suction temperature, and evaporator air inlet moisture content; output variables include evaporator branch effective pipe length, superheated zone pipe length, evaporator two-phase zone pipe length, air outlet temperature, evaporator heat exchange, air outlet relative humidity, and evaporator dehumidification capacity.

[0119] Steps 2-5: Refrigerant charge model:

[0120] Since the refrigerant in a heat pump system is mainly found in the compressor cavity, condenser, evaporator, and refrigerant piping connecting the system, the formula for estimating the refrigerant charge in the system is as follows:

[0121]

[0122] Where M is the refrigerant charge, kg; subscripts TP, sh, and sc represent the two-phase region, superheated region, and subcooled region, respectively; subscripts cond, evap, and comp represent the condenser, evaporator, and compressor, respectively; ρ is the refrigerant density, kg / m³. 3 V is the cavity volume, in meters. 3 Mother is the amount of refrigerant charged in the refrigerant lines excluding the compressor, condenser, and evaporator, expressed in kg.

[0123] In this embodiment, the structural schematic diagram of the shiitake mushroom heat pump drying system is as follows: Figure 1 As shown, the mushroom heat pump drying system includes a heat pump system and an air handling system. In the heat pump system, the refrigerant is compressed by a scroll compressor, increasing its temperature and pressure. After passing through a flow regulating valve, it is divided into two streams at different mass flow rates depending on the valve's opening. These streams flow through an auxiliary condenser and a condenser, respectively, where the refrigerant condenses and releases heat. The refrigerant then merges and flows through a liquid receiver and a drying filter, where it is throttled and depressurized by a throttling valve. It then evaporates and absorbs heat in an evaporator, finally returning to the compressor chamber after passing through a gas-liquid separator, continuously circulating. In the air handling system, air flows through the hot condenser coils under the action of a fan. Heated, the air is then introduced into the multi-layer mushroom dryer. As it flows through the mushrooms, some heat is transferred to them, and the diffused and evaporated moisture is carried away, creating a high-temperature, high-humidity gas. This gas then flows through the cold evaporator coils, where it is cooled and dehumidified, and continues to flow through the condenser, continuously circulating.

[0124] Step 3: Optimize the performance of the above-mentioned mushroom heat pump drying system model using a genetic algorithm. The specific steps are as follows:

[0125] Step 3-1: Construct a knowledge- and data-driven model for a mushroom heat pump drying system. For example... Figure 3 As shown, the input variables of the heat pump drying model include: ambient temperature and humidity, drying temperature, drying humidity, drying wind speed, loading density, number of tray layers, and initial moisture content; the output variables of the heat pump drying model include: dehumidification rate per unit energy consumption and heat pump COP. Based on the mathematical relationships between the inlet and outlet parameters of each component, the compressor model, condenser model, multi-layer mushroom dryer proxy model, expansion valve model, evaporator model, and refrigerant charge model are connected sequentially. The calculation error of the heat exchanger tube length is reduced by iteratively adjusting the condensing temperature and evaporating temperature; the error between the inlet air state parameters and drying conditions in the multi-layer mushroom dryer proxy model is reduced by adjusting the flow ratio of the flow regulating valve and the compressor frequency; and the calculation error of the charge amount is reduced by adjusting the subcooling. The outlet air state of the dryer is calculated in real time through the multi-layer mushroom dryer proxy model, thus enabling dynamic calculation of the mushroom drying process and heat pump performance.

[0126] Step 3-2: As Figure 4 As shown, based on the aforementioned knowledge and data-driven model of the shiitake mushroom heat pump drying system, a genetic algorithm is used to optimize the system's performance. Specifically, the output performance indicators of the dynamic model of the heat pump drying system—namely, the dehumidification rate per unit energy consumption and the heat pump COP—are used as the fitness function. A higher fitness value indicates better performance. The objective function maximizes two optimization objectives, with the range of input parameters serving as constraints.

[0127] First, the initial population is determined based on the parameter range. Each individual in the population represents a configuration scheme, and the fitness value is calculated using the dynamic model of the heat pump drying system in the hybrid surrogate model. All individuals in the initial population are divided into two equal subgroups. Each subgroup corresponds to a sub-objective function, and each sub-objective function performs a selection operation independently within its corresponding subgroup. A roulette wheel selection algorithm is used, with the probability of the i-th individual being selected being Pi. Then:

[0128]

[0129] Where n is the number of individuals in the population, and Fi is the fitness value of the i-th individual.

[0130] Then, individuals with strong adaptability are selected to form a new subgroup, and all new subgroups are merged into a complete population. Crossover and mutation operations are performed on this group to generate the next generation of complete population, with a crossover rate of 0.7 and a mutation rate of 1%, and the "splitting-juxtaposition-selection-merging" operation is continuously performed. Based on the fitness function process, evolutionary search is performed to finally obtain the Pareto optimal solution.

[0131] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, all technical solutions obtained through equivalent substitution or transformation fall within the protection scope of the present invention.

Claims

1. A method for performance optimization of a mushroom heat pump drying system based on a hybrid driving model, characterized in that, Specifically as follows: Step 1: Construct a data-driven, multi-layered agent model for a mushroom dryer; Step 2: Establish a physical model of the heat pump system; the physical model includes a compressor model, a condenser model, an expansion valve model, an evaporator model, and a refrigerant charge model; Step 3: Based on the multi-layer shiitake mushroom dryer proxy model described in Step 1 and the physical model described in Step 2, construct a knowledge and data hybrid driven shiitake mushroom heat pump drying system model, and then use a genetic algorithm to optimize the performance of the shiitake mushroom heat pump drying system model; Step 1 is described in detail as follows: Step 1-1: Construct a CFD numerical model of a multi-layer shiitake mushroom dryer; Steps 1-2: Based on the CFD numerical model of the multi-layer shiitake mushroom dryer, the orthogonal design sampling method is used to determine the sample set; Steps 1-3: Based on the sample set, construct a multi-layer mushroom dryer outlet parameter prediction model based on LSTM deep recurrent neural network, i.e., multi-layer mushroom dryer surrogate model; Steps 1-2 are as follows: Using Design Expert software, an orthogonal experimental table was generated. Based on the CFD numerical model of the multi-layer mushroom dryer described in step 1-1, relevant parameters were modified according to the orthogonal experimental table to perform numerical calculations and obtain CFD simulation data under different drying conditions and initial conditions, namely the outlet air state and mushroom moisture content of the multi-layer mushroom dryer. A mapping set of one set of inputs corresponding to one set of outputs was determined as the sample set. In step 3, the input variables of the mushroom heat pump drying system model include ambient temperature and humidity, drying temperature, drying humidity, drying wind speed, loading density, number of tray layers and initial moisture content, and the output variables include dehumidification rate per unit energy consumption and heat pump COP. The specific method for constructing the model of the shiitake mushroom heat pump drying system is as follows: Based on the mathematical relationships between the inlet and outlet parameters of each component, the compressor model, condenser model, multi-layer mushroom dryer proxy model described in step 1, expansion valve model, evaporator model, and refrigerant charge model are connected sequentially. The calculation error of the heat exchanger tube length is reduced by iteratively adjusting the condensing temperature and evaporation temperature. The error between the inlet air state parameters and drying conditions of the multi-layer mushroom dryer proxy model is reduced by adjusting the flow ratio of the flow regulating valve and the compressor frequency. The calculation error of the charge is reduced by adjusting the subcooling. The outlet air state of the dryer is calculated in real time through the multi-layer mushroom dryer proxy model, thereby obtaining the dynamic parameter changes of the heat pump during the mushroom drying process. Finally, a knowledge- and data-driven mushroom heat pump drying system model is constructed.

2. The method for performance optimization of a mushroom heat pump drying system based on a hybrid driving model according to claim 1, characterized in that, In step 1-1, the input variables of the CFD numerical model of the multi-layer shiitake mushroom dryer include: inlet drying temperature, T. room,in , ℃; Inlet dry relative humidity, RH room,in , %; inlet wind speed, v room,in , m / s; uniform density of material, load, m 2 / kg; number of tray layers; initial temperature, T0, ℃; initial wet basis moisture content, M 0,wb %, %; drying time, t, h; the output variables of the CFD numerical model of the multi-layer shiitake mushroom dryer include: outlet air temperature, T room,out , ℃; Relative humidity of outlet air, RH room,out , %; Exit wind speed, v room,out , m / s; Real-time wet basis moisture content, M t,wb , % The specific method for constructing the CFD numerical model of the multi-layer shiitake mushroom dryer is as follows: First, a geometric model was established based on the actual multi-layer shiitake mushroom drying box; Then, based on the established geometric model, the moisture ratio and volume shrinkage rate of the shiitake mushroom drying process are determined according to the drying kinetic model of a single shiitake mushroom shown in Equation (1) and the shrinkage kinetic model shown in Equation (2). The liquid water evaporation rate of the porous medium domain inside the multi-layer shiitake mushroom dryer is determined using Equation (3), and the porosity of the porous medium domain inside the multi-layer shiitake mushroom dryer is determined using Equation (4). （1） ； （2） ； （3） ； （4） ； Where MR is the moisture content of shiitake mushrooms; SR is the volume shrinkage rate of shiitake mushrooms; t is time, in seconds; m evap The evaporation rate of liquid water in the area where shiitake mushrooms are piled up is given by mol / (m²). 3 ·s); por s The porosity refers to the porosity of the area where the shiitake mushrooms are stacked and shrinking during the process. The density of dried shiitake mushrooms is kg / m³. 3 ;x s,0 X represents the initial volume fraction of the dry matter inside the shiitake mushroom; m,0 The initial dry basis moisture content of shiitake mushrooms; Mn l ρ is the molar mass of water, kg / mol; por is the initial packing porosity; k, n, a, b, c are model coefficients; Based on the principles of conservation of mass, momentum, and energy, control equations for turbulent fluid flow, local non-equilibrium heat transfer, and water vapor transport are established. The control equations for the porous media packing domain of shiitake mushrooms are shown in equations (5), (6), and (7), respectively: （5） ； ； ； （6） ； ； （7） ； in, The density of moist air is kg / m³. 3 u is the velocity vector, m / s; p is the pressure, Pa; I is the unit vector; K is the viscous stress tensor, Pa; μ ma κ represents the dynamic viscosity of moist air, kg / (m·s); κ represents the permeability of the porous media packing domain, m. 2 dp is the equivalent diameter of a single shiitake mushroom, in meters; Cp ma The specific heat capacity of moist air at constant pressure is J / (kg·K); T ma Temperature of moist air, K; k ma h is the thermal conductivity of moist air, W / (m·K); sf The heat transfer coefficient is the gap heat transfer coefficient, W / (m²). 2 ·K); T m The temperature of shiitake mushrooms, in K; The density of shiitake mushrooms is kg / m³. 3 ;Cp m The specific heat capacity of shiitake mushrooms under constant pressure, J / (kg·K); k m Here is the thermal conductivity of shiitake mushroom, W / (m·K); H evap The latent heat of vaporization is expressed in J / kg; c v D represents water vapor concentration, in mol / m³. va Let m be the water vapor-air binary diffusion coefficient. 2 / s; Using COMSOL software, the mathematical model of the multi-layer shiitake mushroom dryer was numerically solved using the finite element method, and the model was experimentally verified to obtain the CFD numerical model of the multi-layer shiitake mushroom dryer.

3. The method for performance optimization of a mushroom heat pump drying system based on a hybrid drive model according to claim 1, characterized in that, Steps 1-3 are described in detail below: First, the original data of the sample set is preprocessed, and the preprocessed sample set is divided into a training set and a test set. The preprocessing includes data sorting according to the data collection time, outlier removal and resampling, and data normalization. Then, based on the training set, the hyperparameters of the LSTM model are optimized using grid search or random search methods. Finally, the LSTM model with optimal hyperparameters is trained using the training set, and the accuracy of the LSTM model is evaluated on the test set to determine whether the error between the output value and the target value meets the requirements, thus obtaining the surrogate model for the multi-layer mushroom dryer.

4. The method for performance optimization of a mushroom heat pump drying system based on a hybrid drive model according to claim 1, characterized in that, The input variables of the compressor model include evaporation temperature, condensation temperature, and evaporator superheat, and the output variables include compressor power, discharge temperature, total mass flow rate at the condenser end, discharge superheat, discharge enthalpy, condensation saturated vapor enthalpy, evaporation saturated vapor enthalpy, evaporation saturated vapor density, suction superheated vapor density, evaporation superheated vapor enthalpy, and suction temperature. The input variables of the condenser model include inlet air temperature, air velocity, sum of flow rates of each branch at the condenser end, inlet superheat enthalpy, condensed saturated vapor enthalpy, condenser subcooling, condensing temperature, compressor discharge superheat, and discharge temperature. The output variables include effective pipe length of the condenser branch, pipe length of each phase region, condenser outlet refrigerant subcooled liquid enthalpy, outlet air temperature, and condenser heat exchange. The input variables of the expansion valve model include the enthalpy of the refrigerant after mixing at the condenser end, and the output variables include the enthalpy of the refrigerant at the outlet. The input variables of the evaporator model include air inlet temperature, face velocity, refrigerant mass flow rate, evaporation temperature, evaporation saturated vapor enthalpy, evaporator inlet refrigerant enthalpy, evaporation saturated vapor density, suction superheated vapor density, evaporation superheated vapor enthalpy, suction temperature, and evaporator air inlet moisture content. The output variables include evaporator branch effective pipe length, superheated zone pipe length, evaporator two-phase zone pipe length, air outlet temperature, evaporator heat exchange, air outlet relative humidity, and evaporator dehumidification capacity. In the refrigerant charge model, the formula for estimating the refrigerant charge is as follows: ； Where M is the refrigerant charge, kg; the subscripts TP, sh, and sc represent the two-phase region, superheated region, and subcooled region, respectively; the subscripts cond, evap, and comp represent the condenser, evaporator, and compressor, respectively. The density of the refrigerant is kg / m³. 3 V is the cavity volume, in meters. 3 Mother is the amount of refrigerant charged in the refrigerant lines excluding the compressor, condenser, and evaporator, expressed in kg.

5. The method for performance optimization of a mushroom heat pump drying system based on a hybrid drive model according to claim 1, characterized in that, In step 3, the genetic algorithm optimization process uses the output of the mushroom heat pump drying system model as the fitness function, determines the unit energy consumption dehumidification rate and heat pump COP as optimization objectives, the objective function is to maximize two objectives, and uses the range of input parameter values ​​as constraints.

6. The method for performance optimization of a mushroom heat pump drying system based on a hybrid drive model according to claim 5, characterized in that, The specific method for optimizing the performance of the shiitake mushroom heat pump drying system using a genetic algorithm is as follows: First, an initial population is determined based on the parameter range, with each individual representing a configuration scheme. The fitness value is calculated using the mushroom heat pump drying system model. All individuals in the initial population are divided into two equal subgroups. Each subgroup corresponds to a sub-objective function, and each sub-objective function performs an independent selection operation within its corresponding subgroup. Then, individuals with high fitness are selected to form a new subgroup, and all new subgroups are merged into a complete population. Crossover and mutation operations are performed on this group to generate the next generation of complete population, and the "splitting-juxtaposition-selection-merging" operation is continuously performed. Based on the fitness function process, an evolutionary search is conducted to finally obtain the Pareto optimal solution.

Images